WEBVTT 00:00:03.200 --> 00:00:07.597 If you have a dog that gets picky about food, you know how frustrating that can 00:00:07.680 --> 00:00:12.477 be. You put the bowl down, they sniff it, walk away, and now you're trying to 00:00:12.560 --> 00:00:16.397 figure out if they're bored, if they don't like it, or if they're just 00:00:16.480 --> 00:00:18.557 holding out for something better. 00:00:18.640 --> 00:00:20.877 Sunday's just launched a brand new fish 00:00:20.960 --> 00:00:22.797 recipe designed for dogs with 00:00:22.880 --> 00:00:26.317 sensitivities, picky eating habits, or 00:00:26.400 --> 00:00:28.397 parents looking to add more variety to 00:00:28.480 --> 00:00:31.117 the bowl. It's made with 80% whole and 00:00:31.200 --> 00:00:33.757 minced fish, including eco-friendly 00:00:33.840 --> 00:00:36.077 invasive carp sourced from American 00:00:36.160 --> 00:00:38.957 rivers. And it's packed with omega-3 00:00:39.040 --> 00:00:41.117 fatty acids to help support healthy 00:00:41.200 --> 00:00:44.397 skin, shiny coats, joints, and brain 00:00:44.480 --> 00:00:49.357 function. Like every Sunday's recipe, it's human-grade, vet developed, and 00:00:49.440 --> 00:00:51.357 gently air dried instead of heavily 00:00:51.440 --> 00:00:54.477 processed. No synthetic additives, no 00:00:54.560 --> 00:00:57.437 meat meals, no artificial flavors, and 00:00:57.520 --> 00:00:59.757 no fillers. Stanley's been loving it, 00:00:59.840 --> 00:01:02.797 too. What do you think, Stanley? You 00:01:02.880 --> 00:01:05.517 like Sundays? For dog owners, it's 00:01:05.600 --> 00:01:08.157 simple, too. No fridge, no thawing, no 00:01:08.240 --> 00:01:10.397 prep. Just scoop and serve. Make the 00:01:10.480 --> 00:01:12.957 switch to Sundays. Go right now to 00:01:13.040 --> 00:01:16.477 sundaysfordd dogs.com/srs50 00:01:16.560 --> 00:01:19.757 and get 50% off your first order. Or you 00:01:19.840 --> 00:01:22.477 can use code SRS50 at checkout. That's 00:01:22.560 --> 00:01:27.480 50% off your first order at sundaysfordogs.com/srs50. 00:01:27.840 --> 00:01:30.840 sundaysfordogs.com/srs50 00:01:31.040 --> 00:01:35.117 or use code srs50 at checkout. 00:01:35.200 --> 00:01:37.357 When you run a business, you track every 00:01:37.440 --> 00:01:39.837 dollar and your bank shouldn't make that 00:01:39.920 --> 00:01:42.557 harder or hold you back. Chime is 00:01:42.640 --> 00:01:44.397 changing the way people bank by offering 00:01:44.480 --> 00:01:47.597 the most rewarding fee free banking. 00:01:47.680 --> 00:01:49.997 It's built for you, not like these old 00:01:50.080 --> 00:01:52.797 banks. No overdraft fees, no monthly 00:01:52.880 --> 00:01:55.517 fees, and access to thousands of feef 00:01:55.600 --> 00:01:59.277 free ATMs. With Chime, you can get up to 00:01:59.360 --> 00:02:01.437 $1,150 00:02:01.520 --> 00:02:05.117 in annual rewards fee free. You get 5% 00:02:05.200 --> 00:02:07.117 cash back on your Chime card in a 00:02:07.200 --> 00:02:09.997 category of your choice, like gas or 00:02:10.080 --> 00:02:12.557 groceries. Your savings grow faster with 00:02:12.640 --> 00:02:17.037 a 3.75% APY. That's nine times higher than the 00:02:17.120 --> 00:02:19.437 national average. You also get premium 00:02:19.520 --> 00:02:22.237 travel perks like airport lounge access 00:02:22.320 --> 00:02:27.357 and a 24/7 concierge and access up to 00:02:27.440 --> 00:02:30.957 $500 of your pay early with my pay. 00:02:31.040 --> 00:02:33.677 Chime is rated five stars by USA Today 00:02:33.760 --> 00:02:37.357 for customer service, real humans 24/7. 00:02:37.440 --> 00:02:39.917 And you're not just switching banks, 00:02:40.000 --> 00:02:44.557 you're upgrading to America's number one choice for banking with a Chime checking 00:02:44.640 --> 00:02:49.037 account. Chime is not just smarter banking. It is the most rewarding way to 00:02:49.120 --> 00:02:53.677 bank. Join the millions who are already banking feef free today. Head to 00:02:53.760 --> 00:02:55.917 chime.comsrs. 00:02:56.000 --> 00:03:00.877 That's chime.com/srs. It only takes a few minutes to sign up. 00:03:00.960 --> 00:03:04.557 Chime is a fintech, not a bank. Banking services for my pay and Chime card 00:03:04.640 --> 00:03:05.757 provided by Chime's bank partners. 00:03:05.840 --> 00:03:08.557 Optional products and services may have fees or charges. Stated annual 00:03:08.640 --> 00:03:11.517 percentage yield and cash back for Chime Prime only. No minimum balance required. 00:03:11.600 --> 00:03:14.717 Checking account ranking based on a JD Power survey published October 20th, 00:03:14.800 --> 00:03:18.077 2025. For more information on APY rates, my pay, spot me, and travel perks, go to 00:03:18.160 --> 00:03:23.760 chime.com/disclosures. [music] 00:03:25.200 --> 00:03:27.757 >> Alex Wang, welcome to the show, man. 00:03:27.840 --> 00:03:29.837 >> Yeah, thanks for having me. I'm excited. 00:03:29.920 --> 00:03:33.757 >> So am I. I like I was telling you at breakfast, I don't I don't know a whole 00:03:33.840 --> 00:03:37.357 lot about tech, but ever since Joe came on, I've been trying to wrap my head 00:03:37.440 --> 00:03:41.837 around it all, and it's just fascinating subject. I I love talking about this 00:03:41.920 --> 00:03:44.237 subject now. So, thank you for coming. 00:03:44.320 --> 00:03:48.397 Well, it's becoming so critical to national security and all the stuff that 00:03:48.480 --> 00:03:52.317 you're very passionate about. So, I mean, I think I think fundamentally tech 00:03:52.400 --> 00:03:54.157 is like we got to get it right. 00:03:54.240 --> 00:03:56.877 Otherwise, stuff gets really dangerous. 00:03:56.960 --> 00:03:59.597 >> Yeah. Yeah. Scares the [__] out of me. 00:03:59.680 --> 00:04:01.037 In fact, we were just having a 00:04:01.120 --> 00:04:03.837 conversation downstairs about about you 00:04:03.920 --> 00:04:07.037 having kids and you were waiting and and 00:04:07.120 --> 00:04:11.277 and Neurolink came up and I had to I had to I had to pause the conversation. 00:04:11.360 --> 00:04:15.597 Dude, I'm like I'm worried about Neurolink, but it sounds like you're 00:04:15.680 --> 00:04:20.477 pretty gung-ho about it. >> So, yeah, a few things. So, yeah. I I 00:04:20.560 --> 00:04:22.477 mean, what I mention is basically I want 00:04:22.560 --> 00:04:25.677 to wait to have kids until we figure out 00:04:25.760 --> 00:04:29.117 how Neurolink or other it's called brain 00:04:29.200 --> 00:04:31.517 computer interfaces. So, other ways for 00:04:31.600 --> 00:04:33.677 brains to interlink with with a with a 00:04:33.760 --> 00:04:36.797 computer um until they start working. Uh 00:04:36.880 --> 00:04:38.957 because so there's a few reasons for 00:04:39.040 --> 00:04:42.397 this. First is in your first like seven 00:04:42.480 --> 00:04:45.277 years of life, your brain is more 00:04:45.360 --> 00:04:49.277 neuroplastic than at any other point in 00:04:49.360 --> 00:04:51.117 your life like by by an order of 00:04:51.200 --> 00:04:54.157 magnitude. So there have been examples 00:04:54.240 --> 00:04:59.277 where you know for example if somebody if a kid is born like you have a newborn 00:04:59.360 --> 00:05:02.397 that has let's say they have cataracts in their eyes so they so they can't see 00:05:02.480 --> 00:05:07.597 through um uh the cataracts and then they live their first seven life seven 00:05:07.680 --> 00:05:10.877 years of their life with those cataracts and then you have them removed when 00:05:10.960 --> 00:05:13.677 they're like eight or nine then even 00:05:13.760 --> 00:05:15.277 with those removed they're not going to 00:05:15.360 --> 00:05:18.797 learn how to see because they're it's so 00:05:18.880 --> 00:05:23.357 important in those first seven years of your development that you're able to 00:05:23.440 --> 00:05:26.317 you're you're able to see that your brain can like learn how to read the 00:05:26.400 --> 00:05:31.197 signals coming off of your eyes. And if you if that's not if you don't have that 00:05:31.280 --> 00:05:35.517 until you're like eight or nine, then you won't learn how to see. So because 00:05:35.600 --> 00:05:39.357 it's so important that your neuroplast your your neuroplasticity is so high in 00:05:39.440 --> 00:05:43.117 that early stage of life. I think when we get neural link and we get these 00:05:43.200 --> 00:05:45.437 other technologies kids who are born 00:05:45.520 --> 00:05:47.437 with them are going to learn how to use 00:05:47.520 --> 00:05:51.277 them in like crazy crazy ways like it'll 00:05:51.360 --> 00:05:56.157 be a actually like a part of their brain in a way that it'll never be true for an 00:05:56.240 --> 00:05:59.837 adult who gets like a neural link or whatever hooked into their hooked into 00:05:59.920 --> 00:06:02.797 their brain. Um, so that's why to wait. 00:06:02.880 --> 00:06:06.237 Now, Neurolink as a as a concept or like 00:06:06.320 --> 00:06:07.837 taking hooking your brain up to a 00:06:07.920 --> 00:06:11.917 computer. Um, I kind of take a a 00:06:12.000 --> 00:06:16.877 pragmatic view on this, which is, you know, my day job, I work on AI. I 00:06:16.960 --> 00:06:20.157 believe a lot in AI. I think AI is going to continue becoming smarter and 00:06:20.240 --> 00:06:25.197 smarter, more and more capable, more and more powerful. Um, AI is going to is 00:06:25.280 --> 00:06:28.237 going to continue being able to do more and more and more and more. We're going 00:06:28.320 --> 00:06:31.597 to have robots. we're going to have other forms for that AI to take over 00:06:31.680 --> 00:06:35.197 time. And so, and humans, we're only 00:06:35.280 --> 00:06:38.077 evolving at a certain rate. Like humans 00:06:38.160 --> 00:06:42.557 are, you know, we are, humans will get smarter over time. It's just on the time 00:06:42.640 --> 00:06:45.997 scale of like millions of years because natural selection and evolution is 00:06:46.080 --> 00:06:48.637 really slow. >> I don't know. Are we getting smarter? 00:06:48.720 --> 00:06:52.157 [laughter] >> I don't know about recently, but uh a 00:06:52.240 --> 00:06:56.237 little setback. >> Yeah. A little a little blip. Um, so if 00:06:56.320 --> 00:06:59.837 you play this forward, right, like you're going to have AIs that are going 00:06:59.920 --> 00:07:02.557 to continue getting smarter, continue 00:07:02.640 --> 00:07:04.397 improving, like they're going to keep 00:07:04.480 --> 00:07:07.277 improving really quickly. And, you know, 00:07:07.360 --> 00:07:09.917 biology is going to improve only so 00:07:10.000 --> 00:07:13.197 fast. And so what what we need at some 00:07:13.280 --> 00:07:16.797 point is the ability to tap into AI 00:07:16.880 --> 00:07:17.997 ourselves. It's like we're going to need 00:07:18.080 --> 00:07:21.677 to bring biological life uh alongside 00:07:21.760 --> 00:07:24.397 all of the siliconbased or artificial 00:07:24.480 --> 00:07:25.917 intelligence and we're going to want to 00:07:26.000 --> 00:07:28.717 be able to tap into that for for our own 00:07:28.800 --> 00:07:31.437 sake, for humanity's sake. And so 00:07:31.520 --> 00:07:34.317 eventually I think we're going to need 00:07:34.400 --> 00:07:37.437 some interlink or hookup between our 00:07:37.520 --> 00:07:40.397 brains directly to AI and the internet 00:07:40.480 --> 00:07:45.117 and all these things. Um, and it's a it is potentially dangerous and it's 00:07:45.200 --> 00:07:49.197 potentially, you know, to your point, terrifying and scary, but we just are 00:07:49.280 --> 00:07:52.717 going to have to do it. Like AI is going to go like this. Humans are going to 00:07:52.800 --> 00:07:55.997 improve at a much slower rate and we're going to need to hook into that 00:07:56.080 --> 00:07:58.957 capability. >> I mean, what 00:07:59.040 --> 00:08:03.357 you know that I've already expressed fear in this and and so I'm I'm curing 00:08:03.440 --> 00:08:05.837 my own fears, I'm just curious like what 00:08:05.920 --> 00:08:11.117 what in your mind what could go wrong? I mean the there's like the obvious thing 00:08:11.200 --> 00:08:14.397 is that some corporation hacks your brain. Well, it's a corporation hacks 00:08:14.480 --> 00:08:18.317 your brain which even that's pretty bad, but that'll be like what they'll like 00:08:18.400 --> 00:08:22.557 they'll like send ads directly to your brain or they'll like make it so you 00:08:22.640 --> 00:08:24.157 want to buy their products or whatnot. 00:08:24.240 --> 00:08:27.197 But then even worse, obviously a you 00:08:27.280 --> 00:08:29.997 know foreign actor, a terrorist, an 00:08:30.080 --> 00:08:35.037 adversary, a state actor you know hacks into your brain and um and takes your 00:08:35.120 --> 00:08:38.157 memories or takes you know like manipulates you or all these things. I 00:08:38.240 --> 00:08:40.637 mean that is that's obviously pretty 00:08:40.720 --> 00:08:43.357 bad. [laughter] 00:08:43.440 --> 00:08:47.837 Yeah. Um and I think that's I to like 00:08:47.920 --> 00:08:50.877 it's definitely a huge risk. I mean for 00:08:50.960 --> 00:08:54.957 sure if you have a direct link into 00:08:55.040 --> 00:09:00.237 someone's brain um and you have the ability to like read their memories, 00:09:00.320 --> 00:09:03.757 control their thoughts, read their thoughts, like um you know that's pretty 00:09:03.840 --> 00:09:08.557 bad. Uh I've I've talked to a lot of scientists in this space and a lot of 00:09:08.640 --> 00:09:09.837 people working on this stuff including 00:09:09.920 --> 00:09:15.677 the folks at Nurling and um you know 00:09:15.760 --> 00:09:18.237 mind reading and mind control are like 00:09:18.320 --> 00:09:20.637 those are the that is where the 00:09:20.720 --> 00:09:22.877 technology will go over time right and 00:09:22.960 --> 00:09:26.717 so um it is like it's something that we 00:09:26.800 --> 00:09:28.637 have to you know like any advanced 00:09:28.720 --> 00:09:31.437 technology we have to not [__] that up 00:09:31.520 --> 00:09:35.197 um but it's going to be pretty critical 00:09:35.280 --> 00:09:39.357 if we want if we want humans to remain 00:09:39.440 --> 00:09:41.517 relevant as AI keeps getting better. 00:09:41.600 --> 00:09:45.597 >> I mean, I interviewed um Andrew Huberman. Do you know who that is? 00:09:45.680 --> 00:09:50.317 >> Yeah. Yeah. Yeah. >> And and um and talked to Ben Car Dr. Ben 00:09:50.400 --> 00:09:52.077 Carson about it too as a kind of a 00:09:52.160 --> 00:09:54.797 follow-on discussion. But what humorin 00:09:54.880 --> 00:09:58.077 was telling me is that cuz this whole 00:09:58.160 --> 00:10:02.317 thing is it sounds like it's it's I don't know a whole lot about neural link 00:10:02.400 --> 00:10:04.157 but from what I've gathered it's it's 00:10:04.240 --> 00:10:07.517 going to help the blind see and it 00:10:07.600 --> 00:10:10.317 sounds like it helps with some um 00:10:10.400 --> 00:10:14.557 connectivity in your joints and bones and stuff for people that are paralyzed. 00:10:14.640 --> 00:10:16.397 But something that Huberman brought up 00:10:16.480 --> 00:10:19.757 is that I was like, well, if if it could 00:10:19.840 --> 00:10:22.637 is going to help the blind see, then 00:10:22.720 --> 00:10:26.317 could they project a total false reality 00:10:26.400 --> 00:10:29.517 into your head? Meaning, you're seeing 00:10:29.600 --> 00:10:33.597 who knows what [__] in the skies everywhere? Sounds like they could 00:10:33.680 --> 00:10:37.837 recreate an entire false reality. He said, "Yes, they will have that ability, 00:10:37.920 --> 00:10:41.357 but not only will they have that ability, they can they can manipulate 00:10:41.440 --> 00:10:43.517 every one of your senses. touch, smell, 00:10:43.600 --> 00:10:45.117 taste, 00:10:45.200 --> 00:10:48.557 uh, insert emotions into your into your 00:10:48.640 --> 00:10:52.237 brain, fear, what, whatever it is. And I 00:10:52.320 --> 00:10:56.957 was like, "Holy shit." Like, they they could manipulate your entire reality 00:10:57.040 --> 00:11:01.117 into a false reality. I mean, you think that's And then I asked Dr. Ben Carson 00:11:01.200 --> 00:11:04.717 about it and he said, you know, who's a a worldrenowned neurosurgeon, he said, 00:11:04.800 --> 00:11:09.757 "Yes, absolutely." They he goes, "Or, you know, they could use it for good." 00:11:09.840 --> 00:11:13.037 But he goes, "What?" [laughter] He was he kind of put it on me. He's like, 00:11:13.120 --> 00:11:14.557 "Well, what do you think would happen?" 00:11:14.640 --> 00:11:18.237 And and like would it be used for good eventually? Uh or would it be used for 00:11:18.320 --> 00:11:22.797 evil? And I mean, what are your what are your thoughts on that? You think that's a 00:11:22.880 --> 00:11:27.197 real possibility? >> I mean, yeah. So, first of all, like we 00:11:27.280 --> 00:11:30.157 don't understand the brain too much today, but eventually we will. Like, 00:11:30.240 --> 00:11:34.077 we're like science is going to solve this problem, right? And um everything 00:11:34.160 --> 00:11:35.997 you just mentioned is ultimately going 00:11:36.080 --> 00:11:39.917 to be on the table, you know? um 00:11:40.000 --> 00:11:43.437 manipulating your emotions, manipulating your senses. The senses thing is already 00:11:43.520 --> 00:11:45.597 happening where I think in monkeys 00:11:45.680 --> 00:11:51.197 they've shown that like um they can, you know, they don't know what it's like 00:11:51.280 --> 00:11:52.557 from the monkeykey's perspective, but 00:11:52.640 --> 00:11:57.117 they're able to project like on a grid 00:11:57.200 --> 00:11:59.757 of a monkey and get them to like like 00:11:59.840 --> 00:12:04.237 click on the right button really reliably. Wow. So they they somehow they 00:12:04.320 --> 00:12:08.557 hook into basically the neural circuits that are doing the visual processing 00:12:08.640 --> 00:12:10.477 visual like image processing in the 00:12:10.560 --> 00:12:13.837 brain and they're able to project like 00:12:13.920 --> 00:12:17.837 things into their into their vision such that the monkey will like always click 00:12:17.920 --> 00:12:21.437 the button that you want it to collect to to click. Um and then you know you 00:12:21.520 --> 00:12:23.357 give it you know a treat or something. 00:12:23.440 --> 00:12:26.237 >> Damn. And so yeah manipulating vision, 00:12:26.320 --> 00:12:28.157 manipulating your senses, manipulating 00:12:28.240 --> 00:12:31.357 your emotions. Um uh this is this will 00:12:31.440 --> 00:12:33.277 be longer term but like leveraging your 00:12:33.360 --> 00:12:36.637 memories um manipulating your memories, 00:12:36.720 --> 00:12:39.917 manipulating like uh those are that 00:12:40.000 --> 00:12:41.197 stuff is on the table. The other stuff 00:12:41.280 --> 00:12:43.757 that is I think more exciting is like 00:12:43.840 --> 00:12:47.837 being able to hook into AI and like all 00:12:47.920 --> 00:12:50.477 of a sudden I have encyclopedic 00:12:50.560 --> 00:12:52.957 knowledge about everything and just like 00:12:53.040 --> 00:12:55.677 you know chat or other AI systems do. I 00:12:55.760 --> 00:12:58.237 can like think at superhuman speeds. I 00:12:58.320 --> 00:13:02.157 can um all of a sudden I can like uh I 00:13:02.240 --> 00:13:05.597 have like way more information I can process like I can like understand 00:13:05.680 --> 00:13:09.117 everything that's going on in the world and process that instantaneously. Like I 00:13:09.200 --> 00:13:11.037 think there's there's an element here 00:13:11.120 --> 00:13:13.117 where it'll legitimately turn us 00:13:13.200 --> 00:13:16.157 superhuman um from a just cognitive 00:13:16.240 --> 00:13:18.477 standpoint. But then to your point, like 00:13:18.560 --> 00:13:22.797 the the flip side of that is the um is 00:13:22.880 --> 00:13:24.637 the risk the other way, which is that 00:13:24.720 --> 00:13:27.437 you're going to have uh it's a huge 00:13:27.520 --> 00:13:32.477 attack vector. Um, yeah. I mean, in 00:13:32.560 --> 00:13:35.517 like I [clears throat] said, I'm not I'm not super tech, but your company Scale 00:13:35.600 --> 00:13:37.997 AI, you basically, correct me if I'm 00:13:38.080 --> 00:13:41.437 wrong, Scale Scale AI is basically the 00:13:41.520 --> 00:13:45.837 the database that the AI uses to come up 00:13:45.920 --> 00:13:49.437 with its answers and answer your prompts and all of that. Correct. 00:13:49.520 --> 00:13:51.997 >> Yeah. So, we do we do a few things. So 00:13:52.080 --> 00:13:55.437 we help large companies and governments 00:13:55.520 --> 00:14:00.877 uh deploy safe and secure advanced AI systems. Um we help with basically every 00:14:00.960 --> 00:14:04.877 step of the process. But the first thing that we were known for and we've done 00:14:04.960 --> 00:14:06.317 very well is exactly what you're saying 00:14:06.400 --> 00:14:09.677 which is creating largecale data sets 00:14:09.760 --> 00:14:13.437 and creating data foundry is what we call it but creating the large scale 00:14:13.520 --> 00:14:18.237 data production that goes into fueling every single one of the major AI models. 00:14:18.320 --> 00:14:23.357 And you know if you ask questions in chat GPT you know that question uh you 00:14:23.440 --> 00:14:25.837 know it's able to answer a lot of those questions well because of data that 00:14:25.920 --> 00:14:30.077 we're able to provide it. And as AI gets more and more advanced you know we're 00:14:30.160 --> 00:14:32.717 continually fueling more advanced 00:14:32.800 --> 00:14:35.037 scientific advanced information and data 00:14:35.120 --> 00:14:37.517 into those models. And then we also work 00:14:37.600 --> 00:14:40.157 with you know the largest you know 00:14:40.240 --> 00:14:42.877 enterprises and and governments like the 00:14:42.960 --> 00:14:46.477 DoD and and other agencies in the US to 00:14:46.560 --> 00:14:49.117 deploy and build full AI systems 00:14:49.200 --> 00:14:53.517 leveraging their own data. And our strategy as a company has been you know 00:14:53.600 --> 00:14:58.957 how do we focus on we have how do we focus on a small number of customers who 00:14:59.040 --> 00:15:02.397 where we can have like a really big impact. So we work with the number one 00:15:02.480 --> 00:15:05.117 bank, we work with the number one pharma company, the number one healthcare 00:15:05.200 --> 00:15:07.197 system, the number one telco, um the 00:15:07.280 --> 00:15:10.877 number one country, America. Uh and uh 00:15:10.960 --> 00:15:13.597 and we work with all of them to like how 00:15:13.680 --> 00:15:16.877 can you no kidding take how you are 00:15:16.960 --> 00:15:21.117 operating today and take sort of the workflows that you're doing today or the 00:15:21.200 --> 00:15:25.597 operations that you have today and use AI to fundamentally transform them. So 00:15:25.680 --> 00:15:29.517 if you're like the largest health care system in the world, how do you and you 00:15:29.600 --> 00:15:33.997 have to, you know, provide care to all of these patients, you know, millions of 00:15:34.080 --> 00:15:37.597 patients, how do you do so in the most effective manner? How do you do it 00:15:37.680 --> 00:15:41.437 logistically better? How do you improve your diagnosis? How do you improve the 00:15:41.520 --> 00:15:45.197 the overall health outcomes of all of your patients? Like that's a problem 00:15:45.280 --> 00:15:47.357 that we help solve with them. or for the 00:15:47.440 --> 00:15:49.597 DoD, you know, there's so much that we 00:15:49.680 --> 00:15:52.557 can do to operate more efficiently and 00:15:52.640 --> 00:15:56.717 uh and ultimately in a more automated way. I mean, you'll know this, I think, 00:15:56.800 --> 00:15:59.997 better than anyone. And so, how do you how do you start implementing those 00:16:00.080 --> 00:16:03.997 systems with AI? >> Mhm. We we'll dive way more in the weeds 00:16:04.080 --> 00:16:05.517 of that later in the interview. kind of 00:16:05.600 --> 00:16:09.197 where I was going with this was if so if 00:16:09.280 --> 00:16:13.357 originally it was it was feeding the AI 00:16:13.440 --> 00:16:14.957 you're given the data center you're 00:16:15.040 --> 00:16:19.837 giving the data to the AI to you come up 00:16:19.920 --> 00:16:22.077 with the answers and and and and answer 00:16:22.160 --> 00:16:25.357 the prompts and so where I was going is 00:16:25.440 --> 00:16:27.597 if if [clears throat] you have Neurolink 00:16:27.680 --> 00:16:30.717 in your head and it's accessing your 00:16:30.800 --> 00:16:33.997 data centers how easy would it be to 00:16:34.080 --> 00:16:38.717 just feed [__] into the data center 00:16:38.800 --> 00:16:41.037 that then feeds all everybody that has a 00:16:41.120 --> 00:16:43.437 neural link in their head. So it could 00:16:43.520 --> 00:16:46.397 be I mean it could be anything. I mean 00:16:46.480 --> 00:16:49.197 here here's an example. I'm a Christian. 00:16:49.280 --> 00:16:53.037 A lot of people think that AI is going to manipulate the Bible and change a lot 00:16:53.120 --> 00:16:55.997 of things. And so how easy would it be 00:16:56.080 --> 00:16:58.797 to just feed that into the AI data 00:16:58.880 --> 00:17:01.517 center and then 00:17:01.600 --> 00:17:05.997 that's that's the new whatever you feed it that becomes the new truth because 00:17:06.080 --> 00:17:08.157 that's what everybody's accessing is 00:17:08.240 --> 00:17:13.357 that specific data. >> Yeah. I mean I think a yes for sure 00:17:13.440 --> 00:17:17.357 that's a huge risk and this is one of the reasons why I think it's really 00:17:17.440 --> 00:17:20.317 important that US or other democratic 00:17:20.400 --> 00:17:23.197 countries lead on AI versus the CCP like 00:17:23.280 --> 00:17:25.597 the Chinese Communist Party or other or 00:17:25.680 --> 00:17:27.997 Russia or other autocratic countries 00:17:28.080 --> 00:17:32.237 because the potential to utilize even AI 00:17:32.320 --> 00:17:36.077 today by the way you can use it to propagandize to a dramatic degree but 00:17:36.160 --> 00:17:37.997 yeah once you get towards you know you 00:17:38.080 --> 00:17:39.917 have neural link or other brain computer 00:17:40.000 --> 00:17:43.677 interfaces that are that can directly, 00:17:43.760 --> 00:17:46.717 you know, um insert thoughts into into 00:17:46.800 --> 00:17:49.597 people's brains. I mean, it's it's 00:17:49.680 --> 00:17:52.317 extreme power that has never existed 00:17:52.400 --> 00:17:55.677 before. And so, who governs that power? 00:17:55.760 --> 00:17:57.757 Who governs that technology? Who makes 00:17:57.840 --> 00:17:59.677 sure that, you know, it's used for the 00:17:59.760 --> 00:18:03.037 right purposes? Um those are like some 00:18:03.120 --> 00:18:07.357 of the most important societal questions that we'll have to deal with. Man, I 00:18:07.440 --> 00:18:09.357 mean, where do you even where do you 00:18:09.440 --> 00:18:12.477 even start with that? Who do you trust 00:18:12.560 --> 00:18:15.747 to control your [__] mind? 00:18:15.830 --> 00:18:20.317 [clears throat] >> Yeah. I mean, I think Well, it's 00:18:20.400 --> 00:18:25.277 interesting. I think the one thing that I think has been um I think a lot of 00:18:25.360 --> 00:18:28.317 people kind of understand it now and we were talking a little bit about this at 00:18:28.400 --> 00:18:30.157 breakfast is like even the degree to 00:18:30.240 --> 00:18:34.397 which even just general media today kind 00:18:34.480 --> 00:18:38.237 of controls your mind or controls the like opinions you have or the beliefs 00:18:38.320 --> 00:18:41.277 you have. And you know, um, you know, we 00:18:41.360 --> 00:18:44.957 were talking about like, you know, is 00:18:45.040 --> 00:18:49.357 does does the media prop up certain military forces to make them seem far 00:18:49.440 --> 00:18:50.877 more fearsome than they actually are? 00:18:50.960 --> 00:18:52.957 And like, you know, there's like some 00:18:53.040 --> 00:18:54.877 lowgrade you can kind of view like some 00:18:54.960 --> 00:18:59.437 lowgrade um forms of like, you know, um, 00:18:59.520 --> 00:19:02.557 >> propaganda, >> propaganda, manipulation, all that stuff 00:19:02.640 --> 00:19:04.957 is like happening like let's say like on 00:19:05.040 --> 00:19:09.757 a scale of 1 to 10 at the one or two level today. And then once you have 00:19:09.840 --> 00:19:11.277 Neuralink or other devices, it's going 00:19:11.360 --> 00:19:15.677 to be like a nine or a 10. And um and I 00:19:15.760 --> 00:19:17.357 think it's really hard. I mean, I think 00:19:17.440 --> 00:19:19.997 I don't think 00:19:20.080 --> 00:19:23.677 any country is prepared to govern 00:19:23.760 --> 00:19:27.437 technology as powerful as a technology that we're going to be developing over 00:19:27.520 --> 00:19:32.877 the next few decades. Like AI, I don't know if we're prepared. Um brain 00:19:32.960 --> 00:19:35.517 computer interfaces, I don't know if we're prepared. Large scale robotics, I 00:19:35.600 --> 00:19:39.677 don't know if we're prepared. Like these are technologies that are just so much 00:19:39.760 --> 00:19:41.357 more powerful than anything that has 00:19:41.440 --> 00:19:43.997 come before. Sometimes people will say 00:19:44.080 --> 00:19:47.997 like, you know, AI is the new mobile. 00:19:48.080 --> 00:19:49.917 It'll be as big as mobile phones and 00:19:50.000 --> 00:19:52.717 it's just no, it's going to be like a 00:19:52.800 --> 00:19:58.157 thousand times bigger and more important and like more impactful. Um, and it's 00:19:58.240 --> 00:20:01.837 not clear that we did the best job regulating mobile phones even. So, um, 00:20:01.920 --> 00:20:04.317 there's it's going to be uh it's going 00:20:04.400 --> 00:20:09.597 to be really important that we get it right. Yeah. I mean, [sighs and gasps] 00:20:09.680 --> 00:20:12.397 everybody that gets one, I mean, you could you could basically 00:20:12.480 --> 00:20:15.597 instantaneously have your have an entire 00:20:15.680 --> 00:20:20.317 army, an entire nation that's linked into your thoughts, your way of 00:20:20.400 --> 00:20:24.797 thinking, and manipulate that entire population to do who the hell knows 00:20:24.880 --> 00:20:29.517 what. Hopefully something for good, but you know how things wind how things 00:20:29.600 --> 00:20:31.357 generally wind up going. But you're 00:20:31.440 --> 00:20:33.117 gung-ho about this stuff. Would you put 00:20:33.200 --> 00:20:37.677 it in? I would I would put it in, but I 00:20:37.760 --> 00:20:40.077 would I would be, you know, I there's a 00:20:40.160 --> 00:20:41.677 few things that need to happen before I' 00:20:41.760 --> 00:20:44.957 I'd be willing to put it in. First, I 00:20:45.040 --> 00:20:49.037 would need to really feel good about the 00:20:49.120 --> 00:20:52.157 cyber offense defense posture. Like, I 00:20:52.240 --> 00:20:57.197 need to have really good confidence that I would be able to defend from uh any 00:20:57.280 --> 00:20:59.117 attacks like any sort of cyber attacks 00:20:59.200 --> 00:21:02.637 into, you know, my brain interface. Um 00:21:02.720 --> 00:21:05.997 uh and that's like that's one big bar. 00:21:06.080 --> 00:21:09.437 Um and then I would need to feel pretty 00:21:09.520 --> 00:21:10.557 confident. I would need to feel 00:21:10.640 --> 00:21:15.437 confident that there were um that uh it 00:21:15.520 --> 00:21:20.077 wouldn't deeply alter my consciousness in any major way like and that I think 00:21:20.160 --> 00:21:23.677 you would see from data of other people who like use it and you you know you'd 00:21:23.760 --> 00:21:27.597 kind of get a sense just from like other people adopting it. Um those would be 00:21:27.680 --> 00:21:31.837 the two things I would need to like feel really really confident about. 00:21:31.920 --> 00:21:34.077 It's a bad thing. It's a bad thing. 00:21:34.160 --> 00:21:36.877 >> Well, the last thing, you know, and then we should talk about other stuff, but 00:21:36.960 --> 00:21:40.557 the last thing about this is um 00:21:40.640 --> 00:21:44.877 uh you know, one of the things that you know, people are there's a lot of talk 00:21:44.960 --> 00:21:48.317 right now about how humans will live forever, right? Or like can humans live 00:21:48.400 --> 00:21:50.957 forever? How do you not die? And a lot 00:21:51.040 --> 00:21:53.517 of that's a lot of that's focused on 00:21:53.600 --> 00:21:56.077 keeping our human bodies healthy and 00:21:56.160 --> 00:22:00.717 keeping our, you know, how do you like how do you take care of yourself? How do 00:22:00.800 --> 00:22:04.877 you take care of your human body? How do we cure diseases um such that like 00:22:04.960 --> 00:22:07.197 humans can live to hundreds and hundreds 00:22:07.280 --> 00:22:12.557 of years? But I think what's the the actual endgame is that we figure out how 00:22:12.640 --> 00:22:14.077 to 00:22:14.160 --> 00:22:16.317 upload our consciousnesses 00:22:16.400 --> 00:22:19.837 from our from our meat brains into a 00:22:19.920 --> 00:22:24.637 computer. And I kind of think about um 00:22:24.720 --> 00:22:28.317 Neuralink or other like other bridges 00:22:28.400 --> 00:22:31.997 between your brain and and computers as like the first step there. 00:22:32.080 --> 00:22:34.237 >> Well, hold on. What? There's a whole 00:22:34.320 --> 00:22:36.557 another rabbit hole we can So, you're 00:22:36.640 --> 00:22:39.837 saying that we we we should be able to 00:22:39.920 --> 00:22:45.277 upload our consciousness or you want to be able to upload our consciousness into 00:22:45.360 --> 00:22:48.397 whatever. >> Yeah, I think I mean now we're like 00:22:48.480 --> 00:22:50.957 we're on like deep end of sci-fi, but um 00:22:51.040 --> 00:22:53.197 but yeah, I mean I think I think there 00:22:53.280 --> 00:22:56.797 will over time be 00:22:56.880 --> 00:23:00.237 uh there. So, one I think the technology 00:23:00.320 --> 00:23:05.037 will exist at some point. We're we're not close today, right? We barely have 00:23:05.120 --> 00:23:08.237 neural link, you know, kind of working, right? So, we're not close, but the 00:23:08.320 --> 00:23:13.437 technology will exist to upload your consciousness onto a computer. 00:23:13.520 --> 00:23:17.437 >> Holy [__] >> And then, okay, let's say, let's say 00:23:17.520 --> 00:23:19.917 we're sitting here, you know, it's like 00:23:20.000 --> 00:23:22.157 50 years from now, this technology 00:23:22.240 --> 00:23:25.277 exists. Um, and you're asking the 00:23:25.360 --> 00:23:29.197 question, uh, you know, are people going 00:23:29.280 --> 00:23:33.437 to upload their consciousness? Well, first off, there's a lot of people who 00:23:33.520 --> 00:23:35.277 who naturally would like people with 00:23:35.360 --> 00:23:38.317 terminal illnesses, um people near 00:23:38.400 --> 00:23:43.037 death, um you know, uh people who are like very fringe and you know, like 00:23:43.120 --> 00:23:45.997 experimenting with this new technology, there will be a class of people who will 00:23:46.080 --> 00:23:47.197 just initially do it. 00:23:47.280 --> 00:23:50.877 >> Mhm. And then um and then as that starts 00:23:50.960 --> 00:23:53.837 to happen and they upload their 00:23:53.920 --> 00:23:56.797 consciousness like the if you have a 00:23:56.880 --> 00:23:59.197 digital you have these sort of like 00:23:59.280 --> 00:24:04.397 digital intelligences um they're uh you know that's true 00:24:04.480 --> 00:24:05.757 immortality. That's the closest thing 00:24:05.840 --> 00:24:08.717 you'll get to to true immortality. Um 00:24:08.800 --> 00:24:13.437 and so the uh I think it's going to become 00:24:13.520 --> 00:24:16.957 like once the technology exists, you know, when it exists, it's going to 00:24:17.040 --> 00:24:21.517 become quite uh uh it's probably going to become a very natural path for most 00:24:21.600 --> 00:24:24.077 humans to go down. >> So what do you what do you think what do 00:24:24.160 --> 00:24:27.837 you think happens if you get your consciousness uploaded and what would it 00:24:27.920 --> 00:24:30.557 even be uploaded into like a cloud or something? 00:24:30.640 --> 00:24:32.477 >> Yeah, it' be uploaded to a cloud. 00:24:32.560 --> 00:24:37.117 >> What do you think? Do you think that you can experience life by uploading your 00:24:37.200 --> 00:24:39.437 consciousness to a cloud? 00:24:39.520 --> 00:24:44.397 >> Yeah. So, so uh yeah, this is uh few 00:24:44.480 --> 00:24:47.757 things. So, first um I'm a big believer 00:24:47.840 --> 00:24:51.837 in robotics. I think we're basically at 00:24:51.920 --> 00:24:55.277 the start of a robotics revolution. Um 00:24:55.360 --> 00:24:59.437 and we're in the very early innings of it, but people are starting to make 00:24:59.520 --> 00:25:02.797 humanoid robots. They're going to get really, really good. people are starting 00:25:02.880 --> 00:25:06.957 to apply them to manufacturing and industrialization and other contexts. Um 00:25:07.040 --> 00:25:11.837 I think the costs are going to come down dramatically and so eventually yeah if 00:25:11.920 --> 00:25:13.917 you you would believe that if you upload 00:25:14.000 --> 00:25:16.397 it and then you could download or down 00:25:16.480 --> 00:25:19.437 link down to a down to a humanoid robot 00:25:19.520 --> 00:25:22.557 then you would kind of experience the 00:25:22.640 --> 00:25:26.877 real world like any other world. Um, or 00:25:26.960 --> 00:25:30.077 you would you could continue in some 00:25:30.160 --> 00:25:35.197 kind of like simulated universe in uh you could almost like play a video game 00:25:35.280 --> 00:25:38.877 in the cloud kind of thing and that could be like the other alternative. 00:25:38.960 --> 00:25:41.357 >> Wow. What do what do you think happens 00:25:41.440 --> 00:25:46.760 when you die? uh 00:25:47.440 --> 00:25:51.517 you know the as AI has gotten 00:25:51.600 --> 00:25:55.037 so um so Elon always talks about how 00:25:55.120 --> 00:25:57.757 we're in a we live in a simulation right 00:25:57.840 --> 00:26:01.277 >> um uh and I remember when I first heard 00:26:01.360 --> 00:26:04.877 him talk about this I was like ah no this is like I don't believe that I 00:26:04.960 --> 00:26:07.117 don't believe we're in a simulation but 00:26:07.200 --> 00:26:09.757 um but as AI has gotten better and 00:26:09.840 --> 00:26:13.357 better at simulating the world like I 00:26:13.440 --> 00:26:15.757 don't know if you've seen these AI video 00:26:15.840 --> 00:26:19.917 um generation models like uh Sora or VO 00:26:20.000 --> 00:26:22.877 or some of these models but you know 00:26:22.960 --> 00:26:26.717 they can produce videos that are totally 00:26:26.800 --> 00:26:31.117 realistic um you would most people could not tell the difference between AI well 00:26:31.200 --> 00:26:36.157 we're seeing this AI generated video and uh and and real video and as that's 00:26:36.240 --> 00:26:41.517 happening it's making me think more and more that we probably live in a simulation 00:26:41.600 --> 00:26:44.957 >> no [__] it. >> Yeah. >> How do you just This is already 00:26:45.040 --> 00:26:48.477 fascinating. We haven't even got to the interview yet. How How do you think 00:26:48.560 --> 00:26:50.797 we're living in a simulation? I mean, I 00:26:50.880 --> 00:26:53.597 know they they say they they cannot 00:26:53.680 --> 00:26:56.717 disprove it. Yeah. You can't like It's 00:26:56.800 --> 00:27:01.757 kind of one of these things. There's there's no way to prove or disprove 00:27:01.840 --> 00:27:03.677 that that you live in a simulation. And 00:27:03.760 --> 00:27:08.877 so, but it's like it's like it's like any, you know, afterlife thought or 00:27:08.960 --> 00:27:12.957 religious thought like all these things that are like fundamentally unprovable. 00:27:13.040 --> 00:27:15.517 But the reason I think it's the case is 00:27:15.600 --> 00:27:18.717 I think in our lifetime we are going to 00:27:18.800 --> 00:27:22.637 be able to create simulations of reality 00:27:22.720 --> 00:27:26.157 that will be hyper realistic. Like I 00:27:26.240 --> 00:27:30.237 think we are gonna create the ability to 00:27:30.320 --> 00:27:33.277 um simulate different versions of our 00:27:33.360 --> 00:27:37.677 world with hyperrealistic accuracy. Um 00:27:37.760 --> 00:27:42.957 and uh and that will happen over the next few decades. And if if we can like 00:27:43.040 --> 00:27:44.477 it's kind of like that Rick and Morty 00:27:44.560 --> 00:27:48.637 episode where if we have the ability as 00:27:48.720 --> 00:27:51.037 an intelligent race to produce, you 00:27:51.120 --> 00:27:54.957 know, millions of simulated worlds, um 00:27:55.040 --> 00:27:58.877 then the likelihood is that we're, you 00:27:58.960 --> 00:28:01.037 know, we're probably also the simulation 00:28:01.120 --> 00:28:05.917 of some other uh more intelligent or more capable species. 00:28:06.000 --> 00:28:09.757 >> Where do you think consciousness goes right now when you die? 00:28:09.840 --> 00:28:11.597 Uh, 00:28:11.680 --> 00:28:16.477 >> what if we are what if we are the super 00:28:16.560 --> 00:28:21.197 advanced robotics? >> Yeah, I think um 00:28:21.280 --> 00:28:22.637 >> and your consciousness just gets 00:28:22.720 --> 00:28:27.277 downloaded into another body generation. 00:28:27.360 --> 00:28:31.837 Yeah, that's true. That would be that's that's something like one way to think 00:28:31.920 --> 00:28:36.077 about it which is like yeah it's all this big simulation that's running and 00:28:36.160 --> 00:28:38.797 as soon as like you know you get you get 00:28:38.880 --> 00:28:43.677 kind of like downloaded or like taken off or like decommissioned from you know 00:28:43.760 --> 00:28:47.117 one entity you get like you know uploaded to another entity kind of 00:28:47.200 --> 00:28:50.397 thing. Um it's kind of that that's 00:28:50.480 --> 00:28:52.397 plausible. I think there's another world 00:28:52.480 --> 00:28:57.197 where like consciousness is like is 00:28:57.280 --> 00:29:00.637 consciousness may not like be that big a 00:29:00.720 --> 00:29:05.357 deal so to speak. like it could be the case that you know I definitely as as 00:29:05.440 --> 00:29:07.517 the models have gotten better and better 00:29:07.600 --> 00:29:09.037 as the AI models have gotten better and 00:29:09.120 --> 00:29:13.597 better um you look at them and uh 00:29:13.680 --> 00:29:16.557 you know you definitely wonder if at some point you're just going to have 00:29:16.640 --> 00:29:21.597 models that are properly conscious and it may just be the fact that like you 00:29:21.680 --> 00:29:24.637 know it's something that can be engineered and if it's something that 00:29:24.720 --> 00:29:28.797 can be engineered then um then all bets are off I think 00:29:28.880 --> 00:29:31.917 >> damn it's pretty wild to think about. 00:29:32.000 --> 00:29:35.917 >> Yeah. Yeah. >> But uh well, let's move into the interview. You ready? 00:29:36.000 --> 00:29:38.237 >> Yeah. >> All right. Everybody starts off with an 00:29:38.320 --> 00:29:44.477 introduction here. So, [clears throat] here we go. 00:29:44.560 --> 00:29:47.517 Alex Wang, founder and CEO of Scale AI, 00:29:47.600 --> 00:29:51.997 a company that's backbone of the AI revolution, providing the data and 00:29:52.080 --> 00:29:53.917 infrastructure that powers the AI 00:29:54.000 --> 00:29:56.397 revolution. Child prodigy who grew up in 00:29:56.480 --> 00:29:59.357 Los Alamos, New Mexico, surrounded by 00:29:59.440 --> 00:30:01.357 scientists with parents who were 00:30:01.440 --> 00:30:04.237 physicists working on military projects. 00:30:04.320 --> 00:30:07.357 Coding wizard who by age 15 was already 00:30:07.440 --> 00:30:09.997 solving AI problems at Kora that stumped 00:30:10.080 --> 00:30:13.357 P PhDs. Visionary entrepreneur who 00:30:13.440 --> 00:30:16.397 dropped out of MIT at 19, turning a Y 00:30:16.480 --> 00:30:18.477 combinator startup into a national 00:30:18.560 --> 00:30:20.877 security powerhouse that's helping the 00:30:20.960 --> 00:30:23.997 US stay ahead in the global AI race. 00:30:24.080 --> 00:30:25.997 Youngest self-made billionaire in the 00:30:26.080 --> 00:30:29.437 world by age 24 built a company valued 00:30:29.520 --> 00:30:33.037 at nearly 25 billion while staying laser 00:30:33.120 --> 00:30:37.117 focused on solving the biggest bottleneck in AI highquality data. 00:30:37.200 --> 00:30:39.597 Unafraid to call the US China AI 00:30:39.680 --> 00:30:44.717 competition and AI warning that the Chinese startups like Deepseek are 00:30:44.800 --> 00:30:46.397 closing the gap faster than most 00:30:46.480 --> 00:30:49.517 realize. guided by your mission to build 00:30:49.600 --> 00:30:52.157 future where AI drives progress, 00:30:52.240 --> 00:30:56.397 security, and opportunity. And so 00:30:56.480 --> 00:31:00.877 there's a big question right now that everybody's that 00:31:00.960 --> 00:31:03.117 everybody's thinking about. Is AI the 00:31:03.200 --> 00:31:05.437 next oil? 00:31:05.520 --> 00:31:10.717 Yeah, I think uh few thoughts there. Um 00:31:10.800 --> 00:31:13.837 in some ways, yes, in some ways no. So 00:31:13.920 --> 00:31:18.957 uh AI is definitely the next um some 00:31:19.040 --> 00:31:20.957 ways in which it's it is the next oil. 00:31:21.040 --> 00:31:23.917 AI will fundamentally 00:31:24.000 --> 00:31:28.397 be the lifeblood of any future economy, 00:31:28.480 --> 00:31:30.397 any future military, any future 00:31:30.480 --> 00:31:33.357 government. Like if you play it out, 00:31:33.440 --> 00:31:36.637 you're like the degree to which a 00:31:36.720 --> 00:31:39.357 country or economy is able to utilize AI 00:31:39.440 --> 00:31:41.837 to make its economy more efficient, to 00:31:41.920 --> 00:31:46.797 automate parts of its economy, um to do automated research and development, 00:31:46.880 --> 00:31:51.277 automate R&D, like you know, push forward in science um using AI. All that 00:31:51.360 --> 00:31:53.277 stuff is going to mean that countries 00:31:53.360 --> 00:31:55.597 that adopt AI effectively will have 00:31:55.680 --> 00:32:00.717 like, you know, nearly infinite GDP growth and countries that don't adopt it 00:32:00.800 --> 00:32:02.077 are going to get are going to get left 00:32:02.160 --> 00:32:05.997 behind. Um, so it is it is sort of the 00:32:06.080 --> 00:32:10.157 the fuel that will power the future of 00:32:10.240 --> 00:32:13.997 of every country. And by the way, I think the same is true of of hard power. 00:32:14.080 --> 00:32:16.317 Like if you look at what the militaries 00:32:16.400 --> 00:32:18.797 of the future are going to be like or or 00:32:18.880 --> 00:32:21.837 what war looks like in the future, AI is 00:32:21.920 --> 00:32:25.837 at the at the core of of what that is going to look like. I'm sure we'll get 00:32:25.920 --> 00:32:28.637 into that. Um and then the ways that 00:32:28.720 --> 00:32:33.357 it's not like oil is, you know, oil is 00:32:33.440 --> 00:32:37.437 this finite resource. you know, we we 00:32:37.520 --> 00:32:42.397 you know, countries that stumble upon large oil reserves, they they have that 00:32:42.480 --> 00:32:45.837 large oil reserve. At some point, it's going to run out. Like in Norway, you 00:32:45.920 --> 00:32:49.037 know, it runs out at some point. And um 00:32:49.120 --> 00:32:52.557 and so it it lends the country power and 00:32:52.640 --> 00:32:55.917 economic riches for a time period. Um 00:32:56.000 --> 00:32:59.597 and then you exhaust it and then you're looking for more oil. Whereas AI is 00:32:59.680 --> 00:33:01.917 going to be a technology that will just 00:33:02.000 --> 00:33:04.637 keep compounding upon itself and will 00:33:04.720 --> 00:33:10.237 keep, you know, um the smarter AI, the more economic power you're going to get, 00:33:10.320 --> 00:33:12.317 which means you're going to build smarter AIs, which means you have more 00:33:12.400 --> 00:33:14.157 economic power and so on and so forth. 00:33:14.240 --> 00:33:15.437 And so it's going to there's going to be 00:33:15.520 --> 00:33:18.637 a flywheel that keeps going on AI, which 00:33:18.720 --> 00:33:22.717 means that um it's not going to be a 00:33:22.800 --> 00:33:26.957 timebased a time limited resource, let's say. It's 00:33:27.040 --> 00:33:32.077 going to be something that that will just continue racing and accelerating 00:33:32.160 --> 00:33:37.197 for >> the entire per perpetuity >> and and data is part of that. Data is a 00:33:37.280 --> 00:33:39.037 big part of that. >> Data is the core part of it. Yeah. So, 00:33:39.120 --> 00:33:42.397 so a lot of times actually um uh I like 00:33:42.480 --> 00:33:45.277 to compare data to oil versus AI. 00:33:45.360 --> 00:33:49.197 >> That's actually what I meant. I [__] that up. I meant to say data. 00:33:49.280 --> 00:33:51.197 >> Yeah. Yeah. Well, I mean I think that's 00:33:51.280 --> 00:33:56.877 totally true. like data. If you think about AI, it boils down to like how do 00:33:56.960 --> 00:34:01.277 you make AI? Well, there's like three pieces. There's um the algorithms like 00:34:01.360 --> 00:34:03.757 the actual code that goes into the the 00:34:03.840 --> 00:34:08.477 AI systems that you know really smart people have to write. Um I used to you 00:34:08.560 --> 00:34:13.997 know write some of these algorithms um back in the day. Uh then there's the um 00:34:14.080 --> 00:34:17.757 then there's the compute the computational power which boils down to 00:34:17.840 --> 00:34:21.677 large scale data centers. you know, do you have the power to fuel them? Do you 00:34:21.760 --> 00:34:23.837 have the chips um to go inside them? 00:34:23.920 --> 00:34:25.197 Like that's like a large scale 00:34:25.280 --> 00:34:28.717 industrial project in question. And then 00:34:28.800 --> 00:34:31.917 um and then data. Do you have all of the 00:34:32.000 --> 00:34:35.997 the lifeblood or do you have all the data that feeds into these algorithms 00:34:36.080 --> 00:34:40.077 that they learn off of? And it's really kind of like the raw material for a lot 00:34:40.160 --> 00:34:42.717 of this intelligence. And and so that's 00:34:42.800 --> 00:34:45.517 why I think data is is the closest thing 00:34:45.600 --> 00:34:48.077 to oil because it is what what gets fed 00:34:48.160 --> 00:34:50.237 into these algorithms fed into the chips 00:34:50.320 --> 00:34:53.677 to make AI so powerful. Um and 00:34:53.760 --> 00:34:55.997 everything we know about AI is that you 00:34:56.080 --> 00:34:58.797 know the better you are at all three of 00:34:58.880 --> 00:35:03.677 these things algorithms computational power data the better your AI get. And 00:35:03.760 --> 00:35:05.437 it's just all about racing ahead on all 00:35:05.520 --> 00:35:08.477 three of these. So when when we see like 00:35:08.560 --> 00:35:11.917 chat GPT gro these types of things are 00:35:12.000 --> 00:35:15.037 they sharing a data center or are they 00:35:15.120 --> 00:35:16.877 all are they completely separate data 00:35:16.960 --> 00:35:20.237 centers? They all use they they all have 00:35:20.320 --> 00:35:23.037 separate data centers. Um this is 00:35:23.120 --> 00:35:27.037 actually one of the one of the major 00:35:27.120 --> 00:35:29.757 uh lanes of competition between the 00:35:29.840 --> 00:35:33.037 companies is who has the ability to 00:35:33.120 --> 00:35:36.077 secure more power and build bigger data 00:35:36.160 --> 00:35:40.397 centers um because ultimately 00:35:40.480 --> 00:35:42.077 you know as AI gets more and more 00:35:42.160 --> 00:35:44.877 powerful the question then becomes how 00:35:44.960 --> 00:35:48.317 many AIs can you run so let's say for a 00:35:48.400 --> 00:35:50.077 second that we get to you know a really 00:35:50.160 --> 00:35:54.557 powerful AI that can do automated cyber 00:35:54.640 --> 00:35:58.157 hacking. So it can do like it can log 00:35:58.240 --> 00:36:01.117 into any kind of server or log into 00:36:01.200 --> 00:36:03.597 another you know or or try to hack some 00:36:03.680 --> 00:36:05.837 website or try to hack some other um try 00:36:05.920 --> 00:36:09.597 to hack some some system. Um then then 00:36:09.680 --> 00:36:13.997 the question is just okay if I have that how many of those can I run? Can I run 00:36:14.080 --> 00:36:18.077 a,000 copies of that? Can I run 10,000 copies of that? Can I run a 100 million 00:36:18.160 --> 00:36:22.317 copies of that? >> Wow. And that all just boils down to how 00:36:22.400 --> 00:36:25.357 many data centers do you have up and running. And that then that boils down 00:36:25.440 --> 00:36:29.357 to okay, how much power do you have to fuel those data centers? How many chips 00:36:29.440 --> 00:36:32.797 do you have to run in those data centers? And how do you keep those 00:36:32.880 --> 00:36:35.597 online for as long as possible? And what 00:36:35.680 --> 00:36:39.837 data is constantly fueling those models to keep getting them to become better 00:36:39.920 --> 00:36:42.397 and better and better. And so this is 00:36:42.480 --> 00:36:47.117 one of the reasons why one of the major ways that the AI companies compete, you 00:36:47.200 --> 00:36:48.957 know, between 00:36:49.040 --> 00:36:52.237 XAI, Elon's company, and OpenAI and um 00:36:52.320 --> 00:36:55.437 and Google and Amazon and and Meta and 00:36:55.520 --> 00:36:56.877 all these companies. One of the major 00:36:56.960 --> 00:37:00.157 ways they compete is just who right now 00:37:00.240 --> 00:37:03.997 is securing more power and more real 00:37:04.080 --> 00:37:06.557 estate for data centers 5 years from now 00:37:06.640 --> 00:37:09.757 and six years from now. Um, and so the 00:37:09.840 --> 00:37:11.997 the battles five, six years down the 00:37:12.080 --> 00:37:14.317 line are being fought literally today. 00:37:14.400 --> 00:37:17.677 >> Wow, man. That's fascinating stuff. 00:37:17.760 --> 00:37:22.317 Well, couple more things before we get into your life story here. Got you a gift. 00:37:22.400 --> 00:37:25.277 >> Oh, man. >> Everybody gets one. 00:37:25.360 --> 00:37:28.237 >> Love it. >> Vigilance Elite Gummy Bears. 00:37:28.320 --> 00:37:31.917 >> There you go. >> Legal in all 50 states. 00:37:32.000 --> 00:37:35.197 No funny business, just candy. Made here in the USA. 00:37:35.280 --> 00:37:40.317 >> Yeah. And um and then one other thing, got a Patreon account. It's a 00:37:40.400 --> 00:37:44.877 subscription account. It's turned into quite the community. And um they've been 00:37:44.960 --> 00:37:48.157 here with me since the beginning when I was running this thing out of my attic. 00:37:48.240 --> 00:37:51.517 And then we moved here. And now we're moving to a new studio. And the team's 00:37:51.600 --> 00:37:56.957 10 times bigger than what it was, which was just me and my wife. But um it's all 00:37:57.040 --> 00:38:00.557 because of them. And so they're the reason I get to sit here with you today. 00:38:00.640 --> 00:38:04.877 And uh so one of the things I do is I offer them the opportunity to ask every 00:38:04.960 --> 00:38:07.037 guest a question. 00:38:07.120 --> 00:38:10.317 This is from Kevin Omali. 00:38:10.400 --> 00:38:12.957 With AI now able to essentially 00:38:13.040 --> 00:38:15.437 replicate so many facets in our of our 00:38:15.520 --> 00:38:18.237 reality, do you see a future where all 00:38:18.320 --> 00:38:20.717 video or photographic evidence presented 00:38:20.800 --> 00:38:25.997 in trials become suspect based on the ability for any of it to have been 00:38:26.080 --> 00:38:29.757 replicated through artificial intelligence tools? 00:38:29.840 --> 00:38:31.677 Uh yeah, so this goes back to what we're 00:38:31.760 --> 00:38:33.197 just talking about. I do think AI is 00:38:33.280 --> 00:38:37.037 going to enable you to do crazy levels 00:38:37.120 --> 00:38:40.477 of simulation and um I don't think our 00:38:40.560 --> 00:38:42.957 courts are ready for it. I think that 00:38:43.040 --> 00:38:46.797 like the like like Kevin was saying, AI 00:38:46.880 --> 00:38:49.197 will be able to generate very convincing 00:38:49.280 --> 00:38:53.597 video, very convincing images um in a 00:38:53.680 --> 00:38:56.957 way at a like we we're not even really at that point yet. Like right now, you 00:38:57.040 --> 00:39:00.077 can still tell when these videos or images are AI generated. That's going to 00:39:00.160 --> 00:39:04.317 keep getting better and it's going to be indistinguishable um from from real 00:39:04.400 --> 00:39:04.877 video. So, 00:39:04.960 --> 00:39:06.797 >> how the hell are we going to discern 00:39:06.880 --> 00:39:11.517 what's real and what's AI generated? 00:39:11.600 --> 00:39:14.237 >> I think that there's two things. I think 00:39:14.320 --> 00:39:20.567 first people are going to need really good [__] detectors, like 00:39:20.650 --> 00:39:25.517 [laughter] like insanely good. And I think um I 00:39:25.600 --> 00:39:28.797 think kids today, by the way, already have much better [__] detectors 00:39:28.880 --> 00:39:32.237 because they grow up on the internet where there's just so much there's so 00:39:32.320 --> 00:39:35.997 much of everything that they have they they already kind of like learn to have 00:39:36.080 --> 00:39:38.557 better and better um [__] detectors. 00:39:38.640 --> 00:39:42.237 Um but uh so that's one. And then the 00:39:42.320 --> 00:39:47.357 second is I mean I think there's going to be um there this is an area where I I 00:39:47.440 --> 00:39:51.037 know there's a lot of push for for various forms of policy and regulation 00:39:51.120 --> 00:39:54.157 but um this is going to I mean it's 00:39:54.240 --> 00:39:56.317 going to be a major question like hey if 00:39:56.400 --> 00:40:00.877 if there's fabricated video or um or 00:40:00.960 --> 00:40:03.757 imagery used in a trial and it's 00:40:03.840 --> 00:40:06.477 discovered that it was fabricated like 00:40:06.560 --> 00:40:10.797 you know what what are the what are the consequences of that and I think it's 00:40:10.880 --> 00:40:13.037 about tuning that such that if you 00:40:13.120 --> 00:40:14.797 fabricate evidence or you fabricate 00:40:14.880 --> 00:40:20.077 things then um then you know that's maybe a worst defense then maybe that's 00:40:20.160 --> 00:40:23.997 the worst offense of all >> uh then I think people would then you 00:40:24.080 --> 00:40:28.717 deter a lot of usage of those tools then if if you set up the incentives in the 00:40:28.800 --> 00:40:31.437 right way. [snorts] >> Yeah. I mean what you know first thing 00:40:31.520 --> 00:40:35.037 that goes to my mind is the US government. I mean, just showing you 00:40:35.120 --> 00:40:39.037 around the studio and stuff talking about, hey, this is what the what the 00:40:39.120 --> 00:40:42.317 government did to those Blackwater guys I was telling you about. They deleted 00:40:42.400 --> 00:40:44.317 the evidence. Well, instead of deleting 00:40:44.400 --> 00:40:45.517 the evidence, 00:40:45.600 --> 00:40:49.357 >> they could make new evidence that is a 00:40:49.440 --> 00:40:51.597 fake gunfight in the source square, 00:40:51.680 --> 00:40:54.557 Baghdad, that proves they're guilty. And 00:40:54.640 --> 00:40:57.277 and then it's the government behind it. 00:40:57.360 --> 00:40:59.357 You know, we've seen it with Brad Giri. 00:40:59.440 --> 00:41:00.877 We've seen it with Eddie Gallagher. 00:41:00.960 --> 00:41:02.797 We've seen it with the Blackwater guys. 00:41:02.880 --> 00:41:05.357 We've seen it a ton just just in my 00:41:05.440 --> 00:41:09.277 small network circle and I could I mean 00:41:09.360 --> 00:41:10.637 you see what's going on with the 00:41:10.720 --> 00:41:12.557 elections all over Europe. They they 00:41:12.640 --> 00:41:17.117 they pulled Georgescu calling him uh 00:41:17.200 --> 00:41:20.077 what was it? I don't know some under 00:41:20.160 --> 00:41:23.597 Russian influence Marie Le Pen in France 00:41:23.680 --> 00:41:26.397 done. I mean, they were talking about 00:41:26.480 --> 00:41:28.397 pulling somebody in Germany not too 00:41:28.480 --> 00:41:30.717 long, maybe about 6 months ago, and it's 00:41:30.800 --> 00:41:33.917 it's just, man, it's [__] crazy, you 00:41:34.000 --> 00:41:36.877 know, and and um scares the hell out of 00:41:36.960 --> 00:41:41.037 me. Scares the hell out of me because then they can just frame anybody they 00:41:41.120 --> 00:41:45.517 want. >> Yeah. I think uh 00:41:45.600 --> 00:41:47.517 definitely one of the one of the 00:41:47.600 --> 00:41:50.877 outcomes of AI is that institutions that 00:41:50.960 --> 00:41:54.877 have power today will gain way more power. 00:41:54.960 --> 00:41:58.717 >> Yeah. >> Um it will it's not naturally 00:41:58.800 --> 00:42:00.957 democratizing. It's a centralizing um 00:42:01.040 --> 00:42:05.597 kind of technology and so uh 00:42:05.680 --> 00:42:08.957 and so yeah we need to build mechanisms so that we can trust those institutions 00:42:09.040 --> 00:42:11.517 otherwise um it doesn't end well. 00:42:11.600 --> 00:42:13.757 >> Yeah. Well, let's get to your story. 00:42:13.840 --> 00:42:17.197 >> Well, I I have gifts, too. Do I do I love gifts? Okay, great. 00:42:17.280 --> 00:42:20.477 >> Um, so a few things I mean, we're going to talk about this, but I grew up in Los 00:42:20.560 --> 00:42:23.997 Alamos, New Mexico. So, my uh my parents 00:42:24.080 --> 00:42:28.077 were both physicists who worked on the n at the national lab there. This is the 00:42:28.160 --> 00:42:31.837 birthplace of the atomic bomb. Um, I don't know if you saw Oppenheimer, but 00:42:31.920 --> 00:42:33.757 uh half of that movie set in Los Alamos, 00:42:33.840 --> 00:42:36.397 where I'm from. So, we got a Los Alamos 00:42:36.480 --> 00:42:41.117 hat. Uh, Los Alamos National Laboratory hat. 00:42:41.200 --> 00:42:42.877 Dude, >> it's uh 00:42:42.960 --> 00:42:48.877 >> very cool. >> We have some Los Alamos coins. 00:42:48.960 --> 00:42:51.837 >> So, >> oh man, >> about the there's one about the atom 00:42:51.920 --> 00:42:54.477 bomb. Uh one about the the Norris 00:42:54.560 --> 00:42:57.037 Bradberry who's a lab director. Um and 00:42:57.120 --> 00:42:59.677 then uh and then also coin about the you 00:42:59.760 --> 00:43:02.397 know the father of the atomic bomb. We 00:43:02.480 --> 00:43:07.197 go. We have a uh 00:43:07.280 --> 00:43:10.237 a like a copy like a basically a copy of 00:43:10.320 --> 00:43:15.117 all the the manual that they that they gave to the scientists uh that got 00:43:15.200 --> 00:43:16.717 declassified 00:43:16.800 --> 00:43:18.157 um 00:43:18.240 --> 00:43:22.397 >> uh from the from the uh actual uh from 00:43:22.480 --> 00:43:26.077 the actual Manhattan project. >> Wow. 00:43:26.160 --> 00:43:28.237 >> And this is cool as [__] 00:43:28.320 --> 00:43:33.517 >> And uh this one's just a fun one. It's a uh it's a rocket kit for you and your 00:43:33.600 --> 00:43:35.517 kids. >> Oh, man. They're going to love that. 00:43:35.600 --> 00:43:37.997 >> Yeah. Thank you, 00:43:38.080 --> 00:43:42.717 >> dude. Thank you. This is going to look awesome in the studio. 00:43:42.800 --> 00:43:43.917 >> That's very cool. 00:43:44.000 --> 00:43:46.717 >> Yeah, it's been kind of surreal. I mean, 00:43:46.800 --> 00:43:50.717 everybody calls uh AI the the next 00:43:50.800 --> 00:43:55.197 Manhattan Project. And so, it's been uh it's been funny cuz that's where I grew 00:43:55.280 --> 00:43:58.237 up. Uh it's like I don't know. Feels weird. 00:43:58.320 --> 00:44:02.237 >> I'll bet it does. Yeah, >> I'll bet it does. So, what were you into 00:44:02.320 --> 00:44:05.437 as a kid? >> So, 00:44:05.520 --> 00:44:07.837 uh, yeah. So, again, both my parents are 00:44:07.920 --> 00:44:11.597 physicists and my and my dad's dad was a 00:44:11.680 --> 00:44:14.477 physicist as well. So, I grew up in this 00:44:14.560 --> 00:44:16.077 like 00:44:16.160 --> 00:44:20.157 pure physics family. Um, so science, 00:44:20.240 --> 00:44:24.877 technology, physics, math, these were uh these were the things I was like I was 00:44:24.960 --> 00:44:27.037 like I was really excited about as a 00:44:27.120 --> 00:44:32.237 kid. Um, and uh I remember like around 00:44:32.320 --> 00:44:33.757 the dinner table we would talk about 00:44:33.840 --> 00:44:37.037 black holes and wormholes and you know 00:44:37.120 --> 00:44:40.797 alien life and uh supernova and you know 00:44:40.880 --> 00:44:42.717 far away galaxies and all that stuff. 00:44:42.800 --> 00:44:44.237 That stuff was all very captivating to 00:44:44.320 --> 00:44:47.677 me. I was thinking about kind of like 00:44:47.760 --> 00:44:50.877 basically like you know understanding 00:44:50.960 --> 00:44:52.877 the universe so for lack of a better 00:44:52.960 --> 00:44:57.117 term. Um, and then I I really like math 00:44:57.200 --> 00:44:59.917 and uh I realized 00:45:00.000 --> 00:45:02.397 kind of, you know, in about four in 00:45:02.480 --> 00:45:05.277 fourth grade I entered my very first 00:45:05.360 --> 00:45:07.197 math competition, 00:45:07.280 --> 00:45:11.997 which is a thing. Uh uh and I I like uh 00:45:12.080 --> 00:45:13.357 it was in it was in the whole state of 00:45:13.440 --> 00:45:16.557 of New Mexico. Uh and I scored the best 00:45:16.640 --> 00:45:19.597 out of any fourth grader in New Mexico. 00:45:19.680 --> 00:45:23.997 Um which uh and then that like activated 00:45:24.080 --> 00:45:27.117 this like competitive gene in me and 00:45:27.200 --> 00:45:32.317 then I just started like you know I got consumed by math competitions, science 00:45:32.400 --> 00:45:33.917 competitions, physics competitions. 00:45:34.000 --> 00:45:36.877 >> What kind of math are you doing in fourth grade? 00:45:36.960 --> 00:45:39.277 >> You math are you doing? 00:45:39.360 --> 00:45:41.917 >> Yeah. Yeah. fourth. I remember let's see 00:45:42.000 --> 00:45:45.357 my parents taught me algebra 00:45:45.440 --> 00:45:51.117 in I want to say it was second grade maybe between >> Are you serious? 00:45:51.200 --> 00:45:54.237 >> Yeah. >> You mastered algebra in second grade? 00:45:54.320 --> 00:45:58.717 >> I don't know if I mastered it but I was Yeah, I was playing around with algebra. 00:45:58.800 --> 00:46:01.517 They they taught me the basics of algebra and I would just like spend all 00:46:01.600 --> 00:46:03.037 time thinking about it in second grade. 00:46:03.120 --> 00:46:05.517 >> It's like seven 8 years old, right? 00:46:05.600 --> 00:46:07.597 >> Yeah. Like eight eight. Yeah. 00:46:07.680 --> 00:46:12.237 >> Holy [__] Um and then and so by the 00:46:12.320 --> 00:46:14.477 time I was by the time I was in fourth 00:46:14.560 --> 00:46:16.557 grade, I could do kind of like I could 00:46:16.640 --> 00:46:19.997 do some basic algebra. I could do um 00:46:20.080 --> 00:46:22.557 some basic geometry stuff like that. And 00:46:22.640 --> 00:46:26.237 then uh let's see where where did I do 00:46:26.320 --> 00:46:28.877 from there? By the time I was in middle 00:46:28.960 --> 00:46:31.437 school, I was doing calculus 00:46:31.520 --> 00:46:37.037 and then um and I was and then I was doing college level math in middle 00:46:37.120 --> 00:46:40.077 school as well. So, those are the two things I was doing in middle school. And 00:46:40.160 --> 00:46:43.677 then, um, in high school, I just became 00:46:43.760 --> 00:46:46.397 obsessed with computers and I just spent 00:46:46.480 --> 00:46:50.477 all day programming. Um, and I realized 00:46:50.560 --> 00:46:54.077 like science and math are cool, but um, 00:46:54.160 --> 00:46:58.157 but with computers and programming, you could actually make stuff. Uh, and that 00:46:58.240 --> 00:47:01.677 was that ended up, you know, becoming the the major obsession. 00:47:01.760 --> 00:47:03.837 >> Back to the dinner table conversations. 00:47:03.920 --> 00:47:05.997 >> Yeah. I mean, Los Alamos, there's like a 00:47:06.080 --> 00:47:09.997 lot of conspiracies and all kinds of 00:47:10.080 --> 00:47:12.237 stuff going on about that place. Remote 00:47:12.320 --> 00:47:15.597 viewing, all all this stuff come seems 00:47:15.680 --> 00:47:21.400 to stem to Los Alamos. But 00:47:21.680 --> 00:47:24.957 >> I have two parents that are physicists in Los Alamos. You guys are talking 00:47:25.040 --> 00:47:28.317 about black holes and aliens and [__] What do you think? Are we are there 00:47:28.400 --> 00:47:32.557 aliens? Uh so there's there's this famous 00:47:32.640 --> 00:47:35.117 paradox the Fermy paradox which is you 00:47:35.200 --> 00:47:37.597 know what are the odds that we live in 00:47:37.680 --> 00:47:42.237 this like vast vast vast universe and uh 00:47:42.320 --> 00:47:45.437 and there's like you know there's 00:47:45.520 --> 00:47:47.997 there's billions hundreds of billions 00:47:48.080 --> 00:47:52.477 trillions of other of other stars and planets and um you know what are the 00:47:52.560 --> 00:47:57.117 chances that like none of them have intelligent life I mean I think like 00:47:57.200 --> 00:48:01.517 definitely somewhere else in our universe there has to be intelligent life. 00:48:01.600 --> 00:48:05.117 >> I think so. >> For sure. The the but the benefit or I 00:48:05.200 --> 00:48:08.557 don't know if the benefit but the but like part of the issue is if we're 00:48:08.640 --> 00:48:12.797 really really really far apart like like millions of light years apart, hundreds 00:48:12.880 --> 00:48:15.277 of millions of light years apart, there's no way we're ever going to 00:48:15.360 --> 00:48:18.317 communicate with each other. Like we're just like super duper far away from each 00:48:18.400 --> 00:48:18.957 other. 00:48:19.040 --> 00:48:21.597 >> Um so I think that's plausible. And then 00:48:21.680 --> 00:48:26.237 the there's the um you know uh there's 00:48:26.320 --> 00:48:30.637 the what's called the dark forest hypothesis. Um I think this is one of 00:48:30.720 --> 00:48:32.637 the things I I actually believe the most 00:48:32.720 --> 00:48:35.597 in probably. So you have the firmy 00:48:35.680 --> 00:48:38.557 paradox that says basically like hey 00:48:38.640 --> 00:48:41.117 what are what are the odds that there's 00:48:41.200 --> 00:48:45.597 no intelligent life out there in the universe? Um there h it's probably zero. 00:48:45.680 --> 00:48:48.557 There has to be some intelligent life somewhere else in the universe. And then 00:48:48.640 --> 00:48:51.597 the question is like why aren't we seeing any like why aren't we seeing any 00:48:51.680 --> 00:48:55.517 aliens? Why aren't we like coming into contact with them? And so then there's 00:48:55.600 --> 00:49:00.557 all these like how do you explain why that is? And there was this um there's 00:49:00.640 --> 00:49:04.957 this hypothesis called the dark forest hypothesis which originally came out of 00:49:05.040 --> 00:49:09.197 a sci-fi novel actually uh but is the one that like jives the most with my 00:49:09.280 --> 00:49:11.917 thoughts which is um the reason you 00:49:12.000 --> 00:49:15.677 don't run into other intelligent life is 00:49:15.760 --> 00:49:18.957 uh if you play the game theory out 00:49:19.040 --> 00:49:20.877 uh if you're an intelligent life you 00:49:20.960 --> 00:49:24.477 don't actually want to be like blaring 00:49:24.560 --> 00:49:26.397 to every other intelligent life that you 00:49:26.480 --> 00:49:29.997 exist because if If you do that, then 00:49:30.080 --> 00:49:33.677 they're just going to come and take you out. Like you're basically like a you 00:49:33.760 --> 00:49:38.797 become like a huge target for other forms of intelligent life. And there, 00:49:38.880 --> 00:49:41.197 you know, some intelligent lives out there are going to be hyperaggressive 00:49:41.280 --> 00:49:44.317 and are going to want to take out, you know, other uh other forms of 00:49:44.400 --> 00:49:46.237 intelligent life. So the dark forest 00:49:46.320 --> 00:49:49.837 hypothesis is that once you become an 00:49:49.920 --> 00:49:53.277 intelligent life form and you become a multiplanary species and all that, you 00:49:53.360 --> 00:49:55.677 realize that you're kind of best off 00:49:55.760 --> 00:50:00.317 minding your own business and not, you know, sending all these sorts of signals 00:50:00.400 --> 00:50:04.157 and trying to like make contact with other life because um it's higher risk 00:50:04.240 --> 00:50:08.957 to do that than to just kind of like, you know, stay isolated. And so there is 00:50:09.040 --> 00:50:12.077 intelligent life out there. There are aliens out there, but everybody's 00:50:12.160 --> 00:50:14.317 incentive is just to stay isolated. 00:50:14.400 --> 00:50:19.037 Interesting. I don't know. I used to believe in it. 00:50:19.120 --> 00:50:22.317 Then I interviewed a bunch of guys. I don't know. I don't know. I think all 00:50:22.400 --> 00:50:25.997 this shit's a big distraction to be honest with you. Yeah, there's 00:50:26.080 --> 00:50:28.637 definitely I mean there's definitely the 00:50:28.720 --> 00:50:31.037 the other portion of this which is you 00:50:31.120 --> 00:50:35.117 know um UFOs are a conspiracy such that 00:50:35.200 --> 00:50:39.837 you know the military can do all sorts of airborne testing and uh and it gets 00:50:39.920 --> 00:50:41.437 discredited because you know people say 00:50:41.520 --> 00:50:44.637 it's UFOs and then uh and then nobody 00:50:44.720 --> 00:50:48.877 believes it. Like there's just no I I'm of all the people I've talked to there's 00:50:48.960 --> 00:50:53.277 just no hard evidence and then and then and then it's the well that's 00:50:53.360 --> 00:50:55.517 classified. It's like I mean is it 00:50:55.600 --> 00:51:00.117 you're on a podcast tour you know but 00:51:00.200 --> 00:51:03.837 [clears throat] I don't know sometimes I think you know this is like all I watch 00:51:03.920 --> 00:51:07.037 is the expanding you all the black holes 00:51:07.120 --> 00:51:09.117 all the this is what I fall asleep to at 00:51:09.200 --> 00:51:11.997 night and uh I don't know I mean they 00:51:12.080 --> 00:51:16.637 they found what like Saturn's rings are all water they think they may have found 00:51:16.720 --> 00:51:18.397 you know there's a possibility of life 00:51:18.480 --> 00:51:20.797 on some of the moons on Saturn that 00:51:20.880 --> 00:51:24.797 would Neptune I think it's made of is 00:51:24.880 --> 00:51:29.837 meant is it Neptune that's made of water like a lot of oceans that are frozen and 00:51:29.920 --> 00:51:33.437 so there may have once been life then there's a they think they found a 00:51:33.520 --> 00:51:35.197 pyramid on Mars or something I don't 00:51:35.280 --> 00:51:38.797 know I sometimes I think maybe 00:51:38.880 --> 00:51:43.277 maybe at any at any one given at any 00:51:43.360 --> 00:51:46.637 particular given point in time there 00:51:46.720 --> 00:51:51.437 there is only one planet that holds life 00:51:51.520 --> 00:51:54.397 as we know at a time and then maybe when 00:51:54.480 --> 00:51:58.157 that planet, you know, becomes obsolete, 00:51:58.240 --> 00:52:00.557 everything goes extinct, maybe it moves, 00:52:00.640 --> 00:52:03.277 you know, maybe it was Mars, I don't 00:52:03.360 --> 00:52:05.597 know, 5 billion years ago and that's 00:52:05.680 --> 00:52:08.717 where life was and then somehow, you 00:52:08.800 --> 00:52:12.717 know, [__] changed and then it developed on Earth. I don't I don't know. That's 00:52:12.800 --> 00:52:16.477 that's that's where I'm at right now. I go back and forth on this [__] all the time. 00:52:16.560 --> 00:52:18.157 >> Yeah, totally. Well, cuz the because our 00:52:18.240 --> 00:52:20.797 star has a life cycle, right? And as it 00:52:20.880 --> 00:52:25.997 goes through that life cycle, different points of our solar system become 00:52:26.080 --> 00:52:28.877 different temperatures, have different conditions, you know, all that kind of 00:52:28.960 --> 00:52:31.677 stuff. And so, um, that's a plausible 00:52:31.760 --> 00:52:36.397 theory. I mean, I think, uh, I think it's I mean, I think both that and what 00:52:36.480 --> 00:52:38.797 we were talking about before in terms of like consciousness in the afterlife, 00:52:38.880 --> 00:52:42.397 these are like some of the some of the great questions because you just, you 00:52:42.480 --> 00:52:44.077 know, we'll probably never know the answers. 00:52:44.160 --> 00:52:48.797 >> Yep. Yep. What what were your parents working on 00:52:48.880 --> 00:52:53.517 at Los Alamos? >> They were um Are they still working 00:52:53.600 --> 00:52:56.157 there? >> Yeah, my mom still's working. Uh my my 00:52:56.240 --> 00:52:57.837 dad's not working, but uh but my mom's 00:52:57.920 --> 00:53:01.357 still working. And so they were part of 00:53:01.440 --> 00:53:04.477 uh of the divisions in Los Alamos 00:53:04.560 --> 00:53:08.877 National Lab that were that worked on classified work. Um that uh they were 00:53:08.960 --> 00:53:11.037 they had clearance. Uh my mom sells 00:53:11.120 --> 00:53:15.517 clearance um with the DOE. Uh, and I 00:53:15.600 --> 00:53:19.837 actually remember like when I grew up, I 00:53:19.920 --> 00:53:23.197 just assumed they were working on cool 00:53:23.280 --> 00:53:26.957 physics research because I was like a 00:53:27.040 --> 00:53:32.157 kid and I I didn't put two and two together. Um and so I I remember when I 00:53:32.240 --> 00:53:33.597 grew up I thought the Los Alamos 00:53:33.680 --> 00:53:36.557 National Lab like used to be the place 00:53:36.640 --> 00:53:39.997 where the atomic bomb was built and then 00:53:40.080 --> 00:53:42.717 uh decades later is just like this like 00:53:42.800 --> 00:53:45.597 advanced scientific research area where 00:53:45.680 --> 00:53:48.557 they're doing research into you know all 00:53:48.640 --> 00:53:51.037 of the you know the frontier of human 00:53:51.120 --> 00:53:52.637 knowledge and it's just this like great 00:53:52.720 --> 00:53:56.477 scientific research area and then um and 00:53:56.560 --> 00:54:00.077 it wasn't until I I it wasn't until I literally 00:54:00.160 --> 00:54:02.557 got to college where I was talking to a 00:54:02.640 --> 00:54:05.117 friend about it and it like dawned on me 00:54:05.200 --> 00:54:08.397 that oh wait, Los Alam's probably still 00:54:08.480 --> 00:54:12.820 mostly weapons research. [laughter] 00:54:13.200 --> 00:54:17.037 Um and uh and oh that's why you would 00:54:17.120 --> 00:54:20.157 need a clearance to [laughter] 00:54:20.240 --> 00:54:23.037 stuff in New Mexico. And then uh since I 00:54:23.120 --> 00:54:26.237 left they actually restarted um they 00:54:26.320 --> 00:54:30.477 restarted uh uh what's called nuclear pit production but they restarted 00:54:30.560 --> 00:54:33.117 basically manufacturing the cores of of 00:54:33.200 --> 00:54:36.397 nuclear weapons. Um this was this must 00:54:36.480 --> 00:54:38.477 have been like 2018 00:54:38.560 --> 00:54:41.917 2019 in Los Alamos. Um, and then I was 00:54:42.000 --> 00:54:44.717 like, "Oh, yeah, no, it's it's mostly a 00:54:44.800 --> 00:54:47.517 research facility to to 00:54:47.600 --> 00:54:50.717 research new nuclear warheads and new 00:54:50.800 --> 00:54:55.997 new nucle new nuclear weapons." Um, uh, and so that dawned on me that didn't 00:54:56.080 --> 00:54:59.277 dawn on me until I was like all the way in college, but yeah. 00:54:59.360 --> 00:55:03.437 >> Wow. >> So, I my guess is my parents worked on that, but uh, >> probably. 00:55:03.520 --> 00:55:08.317 >> Yeah. >> Damn, that's crazy. Wow. 00:55:08.400 --> 00:55:09.997 What else were you into as a kid other 00:55:10.080 --> 00:55:12.637 than other than mathematics? 00:55:12.720 --> 00:55:15.517 >> I loved math. I loved I loved coding. I 00:55:15.600 --> 00:55:17.197 loved science. I loved all that stuff. 00:55:17.280 --> 00:55:22.637 Um I uh I uh was really into violin. I 00:55:22.720 --> 00:55:25.757 was really into to um I uh I would like 00:55:25.840 --> 00:55:27.437 I would practice like you know an hour 00:55:27.520 --> 00:55:30.957 of violin a day. Um, a lot of that was 00:55:31.040 --> 00:55:33.837 because there was sort of like uh, you 00:55:33.920 --> 00:55:36.557 know, in some in some, you know, fields 00:55:36.640 --> 00:55:42.157 or some areas there's like there's just a real beauty to perfection. 00:55:42.240 --> 00:55:45.117 >> Mhm. >> Um, and I think this is true in like a 00:55:45.200 --> 00:55:48.397 lot of arts, um, a lot of music, a lot 00:55:48.480 --> 00:55:50.877 of a lot of frankly everything. I mean, 00:55:50.960 --> 00:55:52.877 I see it even in my current life, in my 00:55:52.960 --> 00:55:55.517 current day-to-day job, but um but there 00:55:55.600 --> 00:55:58.237 was just like, hey, if you could if you 00:55:58.320 --> 00:56:01.997 practice enough to get to play a piece 00:56:02.080 --> 00:56:04.077 perfectly, then it would like it would 00:56:04.160 --> 00:56:07.197 be beautiful. And if you like along the 00:56:07.280 --> 00:56:12.477 way, it's like total dog [__] until you get to the point of like perfection. Um 00:56:12.560 --> 00:56:16.317 there's kind there's a lot of beauty to that concept to me, which is like, you 00:56:16.400 --> 00:56:20.877 know, once you get something totally perfect, it becomes beautiful. Um that 00:56:20.960 --> 00:56:23.037 was that was captivating when I was a when I was a kid. 00:56:23.120 --> 00:56:26.877 >> So you were a perfectionist from a young age and you're still a perfectionist 00:56:26.960 --> 00:56:28.237 today. 00:56:28.320 --> 00:56:32.077 >> Yeah. I see a lot of beauty in like you 00:56:32.160 --> 00:56:36.557 know now I would say I I don't think I don't think we have the luxury to be 00:56:36.640 --> 00:56:38.397 perfectionists. I'm much more pragmatic 00:56:38.480 --> 00:56:42.077 now. Like, you know, 00:56:42.160 --> 00:56:43.597 like we were talking about, the world is 00:56:43.680 --> 00:56:46.957 extremely messy. Like, like the the 00:56:47.040 --> 00:56:50.237 reality is, you know, stuff is super 00:56:50.320 --> 00:56:53.037 chaotic. There's a lot of bad [__] going 00:56:53.120 --> 00:56:54.957 on constantly. Um, there's a lot of good 00:56:55.040 --> 00:56:57.357 [__] going on constantly, but perfection 00:56:57.440 --> 00:57:01.677 is not really a like plausible 00:57:01.760 --> 00:57:05.997 objective. Like, we're never going to get perfection. Um, so I'm a lot more 00:57:06.080 --> 00:57:08.797 pragmatic now, but I do see a lot of beauty and perfection. 00:57:08.880 --> 00:57:11.757 >> I mean, I I'm also a perfectionist. I 00:57:11.840 --> 00:57:17.917 battle it every [__] day. Like I it 00:57:18.000 --> 00:57:21.277 I'm OCD. I did it. But, you know, and 00:57:21.360 --> 00:57:24.637 I've I've read about it. I've watched 00:57:24.720 --> 00:57:29.357 talks about it. It's And I came to the conclusion, which I hate saying this 00:57:29.440 --> 00:57:31.517 because I am a perfectionist at heart. 00:57:31.600 --> 00:57:36.077 you know that perfectionism can get in the way of success. Did you find that? I 00:57:36.160 --> 00:57:40.237 mean, it sounds it sounds weird even like asking you the [__] question 00:57:40.320 --> 00:57:42.317 because you're the youngest billionaire 00:57:42.400 --> 00:57:46.157 in the world at age 24 and I mean you're 00:57:46.240 --> 00:57:49.117 28 years old now. So it sounds weird 00:57:49.200 --> 00:57:51.597 saying did perfectionism hold you back 00:57:51.680 --> 00:57:56.637 but did it? I think um yeah, at some 00:57:56.720 --> 00:57:59.437 point I just like I I like some bit 00:57:59.520 --> 00:58:02.477 flipped and I realized like you got to 00:58:02.560 --> 00:58:07.517 just do the 8020 lots of times. Like you got to do 20% of the effort. That's 80% 00:58:07.600 --> 00:58:09.837 as good and you just have to be okay with that. 00:58:09.920 --> 00:58:12.637 >> Mhm. >> Um and you just have to do that over and 00:58:12.720 --> 00:58:17.357 over and over again. Um so at some point I internalize that and it's like it's 00:58:17.440 --> 00:58:21.597 like anathema to perfectionism. It's like the exact opposite. And so now I 00:58:21.680 --> 00:58:24.717 think about as like hey there's some things where perfectionism really is the 00:58:24.800 --> 00:58:27.917 right answer and there's some things where you just got to you just got to 00:58:28.000 --> 00:58:30.077 like be okay with imperfection and just 00:58:30.160 --> 00:58:32.237 like speed is the objective 00:58:32.320 --> 00:58:36.077 >> versus perfection is the objective. Um 00:58:36.160 --> 00:58:39.197 >> so and yeah I would say now honestly I 00:58:39.280 --> 00:58:43.837 think more things like most things are speed is the objective not not 00:58:43.920 --> 00:58:47.917 perfection. Um so yeah I would say I've kind of had like a whole journey with it. 00:58:48.000 --> 00:58:50.717 >> What what was it that flipped you? 00:58:50.800 --> 00:58:53.677 I think what like uh 00:58:53.760 --> 00:58:56.557 so um 00:58:56.640 --> 00:59:01.117 there's this thing that um Elon says to 00:59:01.200 --> 00:59:02.797 uh people at his company when they're in 00:59:02.880 --> 00:59:04.317 like when they're like a crisis 00:59:04.400 --> 00:59:09.357 situation um and he says like uh 00:59:09.440 --> 00:59:12.877 hey like you know let's say you're in a crisis situation like people are like 00:59:12.960 --> 00:59:14.557 not figuring out how to deal with it and 00:59:14.640 --> 00:59:17.117 then he asked like imagine there was a 00:59:17.200 --> 00:59:20.557 bomb strapped to your body that will go 00:59:20.640 --> 00:59:23.357 off if you don't come up with a solution 00:59:23.440 --> 00:59:26.477 to this problem. Like then what are you 00:59:26.560 --> 00:59:30.397 going to do? Um and then you know most time when people actually like think 00:59:30.480 --> 00:59:33.997 through that scenario they like focus and they get their act together and like 00:59:34.080 --> 00:59:37.357 figure out um like like something to do. 00:59:37.440 --> 00:59:42.397 Um and I think a lot of times startups are like that. like you're like there's 00:59:42.480 --> 00:59:47.197 so many moments that are so life and death and so high pressure that um 00:59:47.280 --> 00:59:51.197 you're just in these situations all the time where you're like you have to act 00:59:51.280 --> 00:59:54.797 and you have to like do something otherwise you're toast and you just have 00:59:54.880 --> 00:59:58.397 to like figure out what the best plan of action is and the best course of action 00:59:58.480 --> 00:59:59.677 and just do it. 00:59:59.760 --> 01:00:04.477 >> Um so I think that that the realities of 01:00:04.560 --> 01:00:06.877 you know having to operate quickly I 01:00:06.960 --> 01:00:10.637 think just over time remolded my brain. 01:00:10.720 --> 01:00:14.797 Interesting. [snorts] Do you have any brothers? Do you have any siblings? 01:00:14.880 --> 01:00:20.237 >> Yeah, I have two brothers. Two older brothers. Um they uh they're both I 01:00:20.320 --> 01:00:21.597 dropped out of college and both my 01:00:21.680 --> 01:00:26.637 brothers have PhDs. So um uh uh but my 01:00:26.720 --> 01:00:29.277 my oldest brother is an economist. Um 01:00:29.360 --> 01:00:34.637 and my my uh my other brother is PhD in neuroscience. So they're uh 01:00:34.720 --> 01:00:37.037 >> Jeez, >> they're smart. Yeah, they're smart guys. 01:00:37.120 --> 01:00:39.837 >> Whole lineage of geniuses, huh? 01:00:39.920 --> 01:00:44.637 Yeah, I think uh my uh yeah, I think my 01:00:44.720 --> 01:00:49.517 parents are are are uh are probably still a little a little myiffed that 01:00:49.600 --> 01:00:53.690 none of us became physicists, but [laughter] 01:00:53.920 --> 01:00:58.957 Oh, man. Well, I'm I'm sure they're they got to be happy with how everything 01:00:59.040 --> 01:01:03.277 turned out. I mean, wow. >> Yeah. Yeah. No, I think I my uh my 01:01:03.360 --> 01:01:04.317 parents are super proud of me. 01:01:04.400 --> 01:01:05.757 >> So, where do you go to where did you go 01:01:05.840 --> 01:01:08.397 to school? I mean, where do you were you 01:01:08.480 --> 01:01:12.157 homeschooled? I went to uh Los Alamos uh 01:01:12.240 --> 01:01:13.757 public high school. Los Alamos public 01:01:13.840 --> 01:01:15.757 middle school. There's there's like the 01:01:15.840 --> 01:01:19.677 town is 10,000 or so people. Um now it's 01:01:19.760 --> 01:01:22.877 more cuz they do pit they do 01:01:22.960 --> 01:01:27.117 manufacturing of these like nuclear cores. So now there's a lot more people 01:01:27.200 --> 01:01:28.477 there. But when I was growing up there 01:01:28.560 --> 01:01:30.157 was like 10 to 15,000 people. So pretty 01:01:30.240 --> 01:01:34.317 small town. And um and there's like one 01:01:34.400 --> 01:01:35.757 public middle school, one public high 01:01:35.840 --> 01:01:39.837 school, a few elementary schools, and uh 01:01:39.920 --> 01:01:42.557 and yeah, that's the you know, I went to 01:01:42.640 --> 01:01:45.997 I went to public school. Um I was lucky 01:01:46.080 --> 01:01:49.677 like I I think those are those are amazing public schools. Uh but it's like 01:01:49.760 --> 01:01:50.957 it is public school like any other 01:01:51.040 --> 01:01:52.717 public school. And then I would just get 01:01:52.800 --> 01:01:57.597 home every day and um and effectively 01:01:57.680 --> 01:02:03.000 like do math and science like every day. 01:02:04.080 --> 01:02:06.077 What 01:02:06.160 --> 01:02:10.317 like what how do you go what what is the 01:02:10.400 --> 01:02:12.957 average second grader? I mean you said 01:02:13.040 --> 01:02:15.997 you had learned algebra in second grade. 01:02:16.080 --> 01:02:19.517 What what is an average? It's been a long time since I've been in second 01:02:19.600 --> 01:02:23.917 grade. things may have changed, but I'm pretty sure it's basic addition. 01:02:24.000 --> 01:02:27.277 >> Yeah, I think it's like addition. Maybe you get to your times tables. Not sure. 01:02:27.360 --> 01:02:29.277 >> Yeah, maybe some multiplication tables. 01:02:29.360 --> 01:02:33.037 >> Yeah. Yeah. >> I mean, so how do you 01:02:33.120 --> 01:02:36.237 Dude, what's what is that like to go to 01:02:36.320 --> 01:02:38.877 go from the night before studying 01:02:38.960 --> 01:02:41.917 algebra to 01:02:42.000 --> 01:02:44.397 2 plus 2 is four? 01:02:44.480 --> 01:02:47.597 >> Yeah. I uh 01:02:47.680 --> 01:02:53.037 I like M I definitely remember in school 01:02:53.120 --> 01:02:56.317 like I think like a lot of a lot of kids 01:02:56.400 --> 01:02:59.357 in general just sort of like 01:02:59.440 --> 01:03:03.277 generally kind of um buying out of the 01:03:03.360 --> 01:03:08.077 whole thing if that makes sense like like kind of just um tuning out and 01:03:08.160 --> 01:03:10.237 daydreaming and just kind of like 01:03:10.320 --> 01:03:12.957 ignoring what was happening in classes. 01:03:13.040 --> 01:03:17.597 Um uh that definitely that definitely started happening and then I what you 01:03:17.680 --> 01:03:21.197 know what what I would actually do or focus on is like go back and then do 01:03:21.280 --> 01:03:25.677 math at home. I mean you're more you're more advanced than the teacher. 01:03:25.760 --> 01:03:27.997 >> There were I remember one time there was 01:03:28.080 --> 01:03:31.677 like uh there was the good thing about 01:03:31.760 --> 01:03:33.517 what you know this the school of the I 01:03:33.600 --> 01:03:36.237 went to is like the teachers 01:03:36.320 --> 01:03:39.757 were really like also invested in my 01:03:39.840 --> 01:03:43.037 education. like I think they um many of 01:03:43.120 --> 01:03:45.837 my teachers wanted to see me like thrive 01:03:45.920 --> 01:03:50.477 and continue learning and um and that was that was awesome. Like I could I can 01:03:50.560 --> 01:03:52.877 imagine a totally separate school where it's like 01:03:52.960 --> 01:03:57.757 >> the teachers don't care because you know >> um 01:03:57.840 --> 01:04:01.277 you know it's just like their lives are chaotic, the classroom's chaotic, all 01:04:01.360 --> 01:04:04.317 that kind of stuff. But but I'd lucky to have teachers who really cared. 01:04:04.400 --> 01:04:07.917 >> Yeah. I mean 01:04:08.000 --> 01:04:10.877 seems like it worked out well. I mean, 01:04:10.960 --> 01:04:12.717 for all the success that you have 01:04:12.800 --> 01:04:17.997 amassed in 28 years, I mean, you're a very grounded person, and I never really 01:04:18.080 --> 01:04:22.397 know what I'm going to get with you guys. At breakfast, I'm I'm I was super 01:04:22.480 --> 01:04:26.957 impressed. I'm like, "Wow, this guy's like really grounded person." And seems 01:04:27.040 --> 01:04:29.277 like a really good person. So, >> Oh, too nice. 01:04:29.360 --> 01:04:31.037 >> Kudos to you, man. 01:04:31.120 --> 01:04:37.720 >> But hey, let's take a quick break. When we come back, we'll get into MIT. 01:04:50.160 --> 01:04:53.277 All right, Alex, we're back from the break. We're getting ready to move into 01:04:53.360 --> 01:04:56.717 you going to college. So, you you started at MIT, correct? 01:04:56.800 --> 01:04:58.717 >> Yep. >> How did that go? 01:04:58.800 --> 01:05:03.277 >> Yeah. So, let's see. I was uh so I'll 01:05:03.360 --> 01:05:05.037 say the first the few years before that. 01:05:05.120 --> 01:05:07.517 So, I dropped out of of high school, actually. 01:05:07.600 --> 01:05:08.957 >> Oh, you dropped out of high school? 01:05:09.040 --> 01:05:11.437 >> Yeah. Yeah, I dropped out of high school. Um, 01:05:11.520 --> 01:05:13.917 >> why not? Why was wasn't challenging 01:05:14.000 --> 01:05:15.837 enough for you? 01:05:15.920 --> 01:05:19.517 >> Uh, I dropped out a year early to to go 01:05:19.600 --> 01:05:22.877 work at uh Quora at this tech company. 01:05:22.960 --> 01:05:24.957 Um, I think a lot of people run into 01:05:25.040 --> 01:05:27.037 Quora, like the question answer website. 01:05:27.120 --> 01:05:29.917 Uh, but uh but I went to go work at a 01:05:30.000 --> 01:05:33.597 tech company for a year. Um and uh and 01:05:33.680 --> 01:05:37.277 then after a year of that I decided okay it's time to go to college. So I went to 01:05:37.360 --> 01:05:42.237 I I went to MIT. >> Yeah. 15 your stump in PhDs. 01:05:42.320 --> 01:05:44.237 >> It was maybe not quite that maybe not 01:05:44.320 --> 01:05:46.397 quite that early but uh but yeah like by 01:05:46.480 --> 01:05:50.317 16 17. Um yeah was uh I was more I was 01:05:50.400 --> 01:05:51.997 more competent by that point. 01:05:52.080 --> 01:05:54.157 >> What are you stumping these guys on? 01:05:54.240 --> 01:05:58.717 >> So um well at that point that was like 01:05:58.800 --> 01:06:02.157 early early AI. It wasn't even called AI yet. It was called machine learning. 01:06:02.240 --> 01:06:05.197 That was like the more popular term. Um 01:06:05.280 --> 01:06:07.037 and it was about training different 01:06:07.120 --> 01:06:11.277 algorithms that would you know uh rerank 01:06:11.360 --> 01:06:13.837 content. It was just like all the like 01:06:13.920 --> 01:06:15.837 um all the algorithms for like these 01:06:15.920 --> 01:06:18.797 social media style um style things and 01:06:18.880 --> 01:06:23.677 it's like okay what algorithm creates the most engagement or what algorithm 01:06:23.760 --> 01:06:28.237 like gets people you know the most hooked on on these feeds. That's what I 01:06:28.320 --> 01:06:29.757 that's what I was working on back then. 01:06:29.840 --> 01:06:33.997 >> Gotcha. Um, and so, uh, so I went, so I 01:06:34.080 --> 01:06:36.557 I worked I worked for a bit and then I 01:06:36.640 --> 01:06:40.317 went to MIT and, um, when 01:06:40.400 --> 01:06:42.797 >> what are you, sorry to interrupt, couple 01:06:42.880 --> 01:06:44.877 more questions. 01:06:44.960 --> 01:06:49.437 What is it like for you to be 16, 17 01:06:49.520 --> 01:06:53.917 years old stumping PhDs? 01:06:54.000 --> 01:06:57.997 I mean, is that is that just like normal life for you? 01:06:58.080 --> 01:06:59.597 I mean, you you know what I mean? like 01:06:59.680 --> 01:07:02.717 does it does it set in like holy [__] 01:07:02.800 --> 01:07:07.277 I'm really [__] smart you know or 01:07:07.360 --> 01:07:10.717 >> I think um I think something that I 01:07:10.800 --> 01:07:16.477 internalized pretty early on was that um 01:07:16.560 --> 01:07:19.757 was that focus was really really 01:07:19.840 --> 01:07:21.837 critical and so I didn't think 01:07:21.920 --> 01:07:25.357 necessarily I mean like I think a lot of 01:07:25.440 --> 01:07:29.837 people are really smart and I don't know if necessarily I'm like way smarter 01:07:29.920 --> 01:07:31.517 fundamentally than a lot of these other 01:07:31.600 --> 01:07:34.957 people, but I was like hyperfocused on 01:07:35.040 --> 01:07:38.317 math as a kid and then hyperfocused on 01:07:38.400 --> 01:07:40.157 phys physics and then in high school 01:07:40.240 --> 01:07:42.237 just hyperfocused on programming and 01:07:42.320 --> 01:07:46.157 then um and so if you if you're like 01:07:46.240 --> 01:07:49.437 hyperfocused and you're just like you like really 01:07:49.520 --> 01:07:52.637 invest the time and the effort you can 01:07:52.720 --> 01:07:55.037 make really really fast progress. Mhm. 01:07:55.120 --> 01:07:58.157 >> So, one of the things that I always like 01:07:58.240 --> 01:08:00.797 I I've believed in for a long time is 01:08:00.880 --> 01:08:05.517 that if you if you overdo things, like 01:08:05.600 --> 01:08:08.557 you like really like invest lots of 01:08:08.640 --> 01:08:12.557 time, lots of effort, you go the extra mile, you go the extra 10 miles, and 01:08:12.640 --> 01:08:14.797 you're like constantly overdoing things, 01:08:14.880 --> 01:08:19.277 then you will improve faster than 01:08:19.360 --> 01:08:24.157 anybody else by many times. Um, and a lot of other people, maybe they're just 01:08:24.240 --> 01:08:26.877 not going the extra mile or maybe they're just not as focused or, you 01:08:26.960 --> 01:08:28.717 know, they're like meandering a bit 01:08:28.800 --> 01:08:32.317 more. And so that's really like I I 01:08:32.400 --> 01:08:35.757 definitely like for me, I think a lot of 01:08:35.840 --> 01:08:38.797 a lot of what I attribute um being able 01:08:38.880 --> 01:08:42.077 to accomplish so much to is really about 01:08:42.160 --> 01:08:45.437 focus and and overdoing it, going the 01:08:45.520 --> 01:08:50.157 extra mile. Um, that's that's what I think it boils down to. What did your 01:08:50.240 --> 01:08:52.797 parents think when you dropped out of school? 01:08:52.880 --> 01:08:56.877 >> Um, you know, they my parents I think 01:08:56.960 --> 01:08:58.637 still probably really want me to get a 01:08:58.720 --> 01:09:02.637 PhD and and do scientific research. So, 01:09:02.720 --> 01:09:05.757 um, uh, they 01:09:05.840 --> 01:09:08.957 I think they view and I respect this 01:09:09.040 --> 01:09:11.917 belief uh, you know, I think they view 01:09:12.000 --> 01:09:13.597 the pursuit of science, the pursuit of 01:09:13.680 --> 01:09:18.397 knowledge as above all else. Um and and 01:09:18.480 --> 01:09:22.077 so I would always tell them, hey, I'm just, you know, this is like a little 01:09:22.160 --> 01:09:26.077 detour, but ultimately I'm going to come back and, you know, finish my degree and 01:09:26.160 --> 01:09:28.797 finish my, you know, get a PhD and I'll 01:09:28.880 --> 01:09:33.357 be on the straight and narrow. Um uh so that's what I always what I was always 01:09:33.440 --> 01:09:37.357 tell them, but uh and then at some point it just didn't be it wasn't believable, 01:09:37.440 --> 01:09:40.237 so [laughter] I just stopped telling them that. 01:09:40.320 --> 01:09:43.197 >> Why did you decide to go to school? I 01:09:43.280 --> 01:09:46.797 went to school because um 01:09:46.880 --> 01:09:50.557 uh well there were two things. One was 01:09:50.640 --> 01:09:53.517 like genuinely I wanted to learn a lot 01:09:53.600 --> 01:09:57.277 about AI very quickly and I knew I could 01:09:57.360 --> 01:10:00.477 kind of do that while working maybe but 01:10:00.560 --> 01:10:02.157 um the best thing to do really would be 01:10:02.240 --> 01:10:05.597 to like go to school like invest all my 01:10:05.680 --> 01:10:08.717 time into it and and uh and try to learn 01:10:08.800 --> 01:10:11.517 learn very very quickly. Um and then the 01:10:11.600 --> 01:10:14.397 second thing was like you know almost 01:10:14.480 --> 01:10:17.197 anyone you'll not anyone but like many 01:10:17.280 --> 01:10:21.757 many people if you ask them like what were the best years of your life like a 01:10:21.840 --> 01:10:27.437 lot of people will say their college years and so I was like [__] I can't I'm 01:10:27.520 --> 01:10:28.637 not going to sacrifice the college 01:10:28.720 --> 01:10:33.597 years. Um so so uh yeah I went to school 01:10:33.680 --> 01:10:36.957 I like I decided to just go really 01:10:37.040 --> 01:10:39.437 really deep into AI. I took all of the 01:10:39.520 --> 01:10:42.077 AI courses I could um uh while I was at 01:10:42.160 --> 01:10:44.877 MIT. I was only there for a year, but I 01:10:44.960 --> 01:10:48.397 I started out I remember I took a um I 01:10:48.480 --> 01:10:50.477 wanted to take the the sort of like 01:10:50.560 --> 01:10:55.597 hardest machine learning course the first semester I got there and the my 01:10:55.680 --> 01:10:59.357 freshman adviser, the person who was like I get all my courses approved with 01:10:59.440 --> 01:11:03.357 was the professor of that course. Um, this just like happened to be the case. 01:11:03.440 --> 01:11:05.677 And uh, I like signed up for her course 01:11:05.760 --> 01:11:10.397 and then she she said like, "You're a freshman. You're you're not going to, 01:11:10.480 --> 01:11:13.117 you know, this is going to be this is going to be too much for you." And I was 01:11:13.200 --> 01:11:16.317 like, "Ah, just give me a chance. Like, you know, I I I just want to try it. 01:11:16.400 --> 01:11:19.597 Like, I'm really passionate about the topic." And she's like, "Okay, well, 01:11:19.680 --> 01:11:23.517 we'll let you uh we'll let you go till the first, you know, for the first few 01:11:23.600 --> 01:11:25.837 weeks and see how you do." And so then I 01:11:25.920 --> 01:11:29.517 get in and then uh I remember I was like 01:11:29.600 --> 01:11:33.437 I felt uh I felt like the stakes were really high cuz like I wanted to like 01:11:33.520 --> 01:11:35.357 prove that I could do this. And so the 01:11:35.440 --> 01:11:39.437 first test rolls around and uh I think 01:11:39.520 --> 01:11:43.517 by like sheer luck it just happened to mostly be about things that like like 01:11:43.600 --> 01:11:45.517 there were a lot of things in the course I didn't understand but happened to be 01:11:45.600 --> 01:11:49.917 about stuff that I did understand in the course pretty well and I got like one of 01:11:50.000 --> 01:11:52.397 the top marks in that course and there were like hundreds of people in this 01:11:52.480 --> 01:11:57.117 class. Um, and so then after that point, the professor let me do whatever I 01:11:57.200 --> 01:11:58.877 wanted. And then 01:11:58.960 --> 01:12:01.837 >> uh, and so then I did all of these I I 01:12:01.920 --> 01:12:04.157 was I went really deep into AI and all 01:12:04.240 --> 01:12:06.797 the and all the AI course work at MIT. 01:12:06.880 --> 01:12:10.717 Um, and then this was the year when uh, 01:12:10.800 --> 01:12:13.197 DeepMind, the this like AI company out 01:12:13.280 --> 01:12:15.917 of London, came out with Alph Go, which 01:12:16.000 --> 01:12:19.677 was the first AI that beat um, the best 01:12:19.760 --> 01:12:23.757 Go players in the world, which was viewed at that point as like probably 01:12:23.840 --> 01:12:27.757 the hardest strategy game or the hardest sort of like um, yeah, the hardest 01:12:27.840 --> 01:12:29.997 strategy game for AIs to beat. And that 01:12:30.080 --> 01:12:32.957 was a big deal. And then I started 01:12:33.040 --> 01:12:37.597 tinkering with AI on my own. So, I built I wanted to build like a camera inside 01:12:37.680 --> 01:12:41.607 my fridge that would tell me when my roommates were stealing my food. 01:12:41.690 --> 01:12:46.477 [laughter] And uh and and so I started tinkering 01:12:46.560 --> 01:12:52.077 with it. And then I pretty quickly realized um uh kind of what we're just 01:12:52.160 --> 01:12:56.717 what we were talking about earlier that data was going to be like that 01:12:56.800 --> 01:12:58.477 everything was going to be blocked on 01:12:58.560 --> 01:13:01.597 data. like if we no matter what you 01:13:01.680 --> 01:13:04.397 wanted AI to do that was that was going 01:13:04.480 --> 01:13:09.517 to rely on data to make the AI do those things. Um and so I and I looked around 01:13:09.600 --> 01:13:12.557 I was like nobody's working on this problem. You know you have plenty of 01:13:12.640 --> 01:13:15.437 guys working on building great algorithms. You have plenty of people 01:13:15.520 --> 01:13:17.597 working on building the chips and the 01:13:17.680 --> 01:13:20.317 computational capacity and and um and 01:13:20.400 --> 01:13:25.917 all that. Nobody working on data. So I was you know I was impatient. [snorts] 01:13:26.000 --> 01:13:28.237 You know I was 19 years old. I was kind of impatient. And I was like, "Well, if 01:13:28.320 --> 01:13:32.077 nobody's going to do it, I might as well do it." Dropped out, started the 01:13:32.160 --> 01:13:34.877 company, and was off to the races. Um, 01:13:34.960 --> 01:13:39.677 >> Damn. So, did you uh perfect the the refrigerator AI to tell you if your 01:13:39.760 --> 01:13:41.357 roommates are stealing your food? 01:13:41.440 --> 01:13:43.597 >> I uh that was part of the problem. I was 01:13:43.680 --> 01:13:48.477 like I was I I was trying to build it and then I realized I didn't have 01:13:48.560 --> 01:13:52.397 anywhere near enough data. So, it always like fire incorrectly and always have 01:13:52.480 --> 01:13:54.317 false positives, false negatives, etc. 01:13:54.400 --> 01:13:59.197 And then um and I realized like uh then 01:13:59.280 --> 01:14:01.997 that was like the light bulb moment. I was like, "Oh [__] if I really want to 01:14:02.080 --> 01:14:05.357 make this, I need like like a million 01:14:05.440 --> 01:14:09.437 times more data than I have now." And that's going to be true for like every 01:14:09.520 --> 01:14:12.157 AI thing that anyone ever wants to build. 01:14:12.240 --> 01:14:16.797 >> Um and so that was kind of the the genesis of the >> the idea really. 01:14:16.880 --> 01:14:18.557 >> So you left MIT. 01:14:18.640 --> 01:14:22.397 >> Left MIT. I remember I moved I I flew 01:14:22.480 --> 01:14:27.037 straight from Boston to San Francisco 01:14:27.120 --> 01:14:30.957 um to start the company. Uh and um 01:14:31.040 --> 01:14:34.717 basically immediately went from like at 19 years old 01:14:34.800 --> 01:14:37.357 >> 19 years old. Yeah. I immediately left 01:14:37.440 --> 01:14:40.477 and then I started coding [laughter] um 01:14:40.560 --> 01:14:42.717 in San Francisco and I was part of this 01:14:42.800 --> 01:14:44.877 uh this like accelerator like I was part 01:14:44.960 --> 01:14:47.677 of this program called Y Combinator and 01:14:47.760 --> 01:14:49.757 um and it's kind of like the Hunger 01:14:49.840 --> 01:14:54.077 Games for startups. Um so there's like 01:14:54.160 --> 01:14:58.477 there's like it starts out there's a 100 startups at the start of the summer and 01:14:58.560 --> 01:15:02.717 you're all like grinding away. You're all working. you're all trying to like 01:15:02.800 --> 01:15:05.277 show milestones and show progress and 01:15:05.360 --> 01:15:08.637 then it culminates at the end of at the 01:15:08.720 --> 01:15:12.237 end of the uh of of Y Com at the end of 01:15:12.320 --> 01:15:16.317 it all there's a demo day where everybody presents their companies, 01:15:16.400 --> 01:15:20.957 presents their progress and tries to get investment and uh and it so it literally 01:15:21.040 --> 01:15:22.317 it quite literally is the Hunger Games. 01:15:22.400 --> 01:15:25.037 It's like you go through this whole thing at the end if you get investment 01:15:25.120 --> 01:15:27.037 you get money you've won. If you didn't 01:15:27.120 --> 01:15:30.877 um you've lost. Uh and uh and so that 01:15:30.960 --> 01:15:34.317 was like that was that was the beginning of the company. We ended up getting good 01:15:34.400 --> 01:15:37.277 investment. >> What what did you do? 01:15:37.360 --> 01:15:40.237 >> Uh well at that time we were we were it 01:15:40.320 --> 01:15:41.997 was it was around data for AI. So it was 01:15:42.080 --> 01:15:45.197 all around like uh how do we fuel data 01:15:45.280 --> 01:15:47.997 for for um what people want to build 01:15:48.080 --> 01:15:52.797 with AI. But at that time it was like so early that like the use cases were 01:15:52.880 --> 01:15:54.877 pretty stupid. Like [clears throat] we 01:15:54.960 --> 01:15:59.597 were helping one company try to detect 01:15:59.680 --> 01:16:03.357 like it was like a t-shirt company. They made like custom t-shirt designs and 01:16:03.440 --> 01:16:04.957 we're trying to help them detect when 01:16:05.040 --> 01:16:08.477 people um were like like used a t-shirt 01:16:08.560 --> 01:16:11.757 design that was like that was like um 01:16:11.840 --> 01:16:15.117 like uh that was like unfit for to print 01:16:15.200 --> 01:16:17.757 like you know it had like gore or or 01:16:17.840 --> 01:16:20.877 like um you know you all sorts of like 01:16:20.960 --> 01:16:24.797 illegal stuff like if basically like identifying illegal t-shirt designs. 01:16:24.880 --> 01:16:28.157 It's kind of like stupid now they say it. And then we're helping another 01:16:28.240 --> 01:16:31.117 company um it was like a furniture 01:16:31.200 --> 01:16:33.597 marketplace. We're helping them like 01:16:33.680 --> 01:16:36.237 improve their search algorithm with AI. 01:16:36.320 --> 01:16:38.637 And then maybe a few months in maybe 3 01:16:38.720 --> 01:16:41.677 months in we started working with 01:16:41.760 --> 01:16:43.677 autonomous vehicle companies and 01:16:43.760 --> 01:16:46.957 self-driving companies. Um and then that 01:16:47.040 --> 01:16:50.877 and that ended up being like the real uh 01:16:50.960 --> 01:16:52.877 the real meat behind our effort for the 01:16:52.960 --> 01:16:55.917 first 3 4 years. So, we worked with, you 01:16:56.000 --> 01:17:01.597 know, General Motors and Toyota and um Whimo and, you know, all of the major 01:17:01.680 --> 01:17:05.277 automakers. Wow. >> In helping them build self-driving cars. 01:17:05.360 --> 01:17:08.077 >> How many people are you competing against? 01:17:08.160 --> 01:17:11.597 >> I mean, um I think in anything you do in 01:17:11.680 --> 01:17:16.717 startup plan like you have like tens of competitors. Um you know, and and there 01:17:16.800 --> 01:17:21.997 were there were definitely tens of competitors at that time. Um and uh and 01:17:22.080 --> 01:17:23.037 so it's like you know these are 01:17:23.120 --> 01:17:28.237 competitive spaces but um where uh as we 01:17:28.320 --> 01:17:30.467 described I don't mind competition 01:17:30.550 --> 01:17:35.437 [laughter] um from math competition days and so um and so we we were just like 01:17:35.520 --> 01:17:38.557 really focused on the problem really focused on how do you what are the best 01:17:38.640 --> 01:17:40.797 possible data sets um for these 01:17:40.880 --> 01:17:43.117 self-drivening cars a lot of that had to 01:17:43.200 --> 01:17:46.397 do with it's called sensor fusion so you 01:17:46.480 --> 01:17:47.997 know there's so many different kinds of 01:17:48.080 --> 01:17:51.037 sensors And how do you combine all these 01:17:51.120 --> 01:17:54.797 different sensors to get, you know, one 01:17:54.880 --> 01:17:58.877 output? So like if multiple sensors sense a person, how do you like collect 01:17:58.960 --> 01:18:02.237 all that together to say that's one person right there and that's one car 01:18:02.320 --> 01:18:04.317 right there and that's one, you know, 01:18:04.400 --> 01:18:06.557 bicycle over there. Um, so that was kind 01:18:06.640 --> 01:18:10.637 of our specialty as a company. And then 01:18:10.720 --> 01:18:12.397 um then we're kind of off to the races 01:18:12.480 --> 01:18:18.077 just on that. We we grew the company to like 100 or so people. Let's go back 01:18:18.160 --> 01:18:22.477 just a little bit. >> Okay. >> So, you you go to San Francisco by 01:18:22.560 --> 01:18:24.637 yourself as a 19-year-old kid who had 01:18:24.720 --> 01:18:29.757 just dropped out of MIT. How do you 01:18:29.840 --> 01:18:34.157 mean you're immature at that point? And 01:18:34.240 --> 01:18:36.397 so, how do you develop leadership 01:18:36.480 --> 01:18:39.357 skills? And I mean, how do you have how 01:18:39.440 --> 01:18:42.477 do you have the knowhow and make the 01:18:42.560 --> 01:18:44.637 connections to build a company as a 01:18:44.720 --> 01:18:47.517 19-year-old kid? 01:18:47.600 --> 01:18:51.197 Yeah, you so um 01:18:51.280 --> 01:18:53.837 let's see what happened. So basically 01:18:53.920 --> 01:18:57.517 early on like um it's about who you get 01:18:57.600 --> 01:19:01.517 investment from. And so if you get >> So it was just you with the competition. 01:19:01.600 --> 01:19:02.877 There was no team. 01:19:02.960 --> 01:19:07.357 >> No team. No team. And then um and so I 01:19:07.440 --> 01:19:10.077 and I was coding every day. And then I 01:19:10.160 --> 01:19:13.917 got um we got Y Cominator to invest in 01:19:14.000 --> 01:19:18.717 us. And then we got um this this investment firm called Excel which was 01:19:18.800 --> 01:19:20.317 who were one of the early investors into 01:19:20.400 --> 01:19:24.397 Facebook to invest. And so we got um 01:19:24.480 --> 01:19:28.717 some some good investors and then they 01:19:28.800 --> 01:19:31.757 helped me build the team like find 01:19:31.840 --> 01:19:36.557 people to hire. I also hired you know what actually happened is I mostly hired 01:19:36.640 --> 01:19:38.317 people I knew from school. >> Really? 01:19:38.400 --> 01:19:42.957 >> Yeah. So like >> cuz you could trust them. I think more 01:19:43.040 --> 01:19:44.317 that they could trust me because I think 01:19:44.400 --> 01:19:48.477 if like at the time if I went to like a 01:19:48.560 --> 01:19:53.917 a 25-year-old engineer in San Francisco and I was like, "Hey, we should we 01:19:54.000 --> 01:19:58.077 should work together." I had no credibility. Like um I remember I w I 01:19:58.160 --> 01:20:00.477 like I would get coffee with these people and I would say like, "Yeah, this 01:20:00.560 --> 01:20:01.677 is what we're working on. It's super 01:20:01.760 --> 01:20:03.517 cool. You should join us." And then they 01:20:03.600 --> 01:20:08.157 would all just be like, "Okay, um cool. 01:20:08.240 --> 01:20:12.157 I guess I'm going to go back to my job now. [laughter] 01:20:12.240 --> 01:20:14.557 Um, so early on I had no credibility 01:20:14.640 --> 01:20:17.197 except for with people I went to uh 01:20:17.280 --> 01:20:20.237 college with uh who we were just like 01:20:20.320 --> 01:20:24.397 friends and we liked each other and so I managed to recruit a bunch of them over. 01:20:24.480 --> 01:20:26.797 Um, and >> they dropped out too. 01:20:26.880 --> 01:20:29.517 >> Some of them dropped out. Some [clears throat] of them just happened 01:20:29.600 --> 01:20:32.717 to, you know, were like seniors or whatever, finished school and then 01:20:32.800 --> 01:20:35.357 joined. Um, it was like a mix. It was a 01:20:35.440 --> 01:20:39.037 mix. Um and uh and that was like the 01:20:39.120 --> 01:20:42.797 early nucleus of the team, the early 01:20:42.880 --> 01:20:47.837 sort of like cohort of the team. And then um and then we started picking up 01:20:47.920 --> 01:20:51.597 momentum because we're starting to work with large automotive companies. We're 01:20:51.680 --> 01:20:56.237 starting to work with you know these very futuristic autonomous driving companies. 01:20:56.320 --> 01:21:00.397 >> And then as momentum started to pick up like you know we were able to grow and 01:21:00.480 --> 01:21:04.157 build out the team over time. I mean, so so where did you get your business sense 01:21:04.240 --> 01:21:05.917 or did you hire somebody to run all of 01:21:06.000 --> 01:21:07.597 that? You were you were the mastermind 01:21:07.680 --> 01:21:12.877 behind everything? Um I uh maybe about a 01:21:12.960 --> 01:21:14.957 year in I hired somebody literally with 01:21:15.040 --> 01:21:19.277 the title head of business. Um but uh 01:21:19.360 --> 01:21:22.557 but until then I was just kind of like I 01:21:22.640 --> 01:21:24.317 was just trying to like learn it all. 01:21:24.400 --> 01:21:26.797 >> How did you get the product out there? 01:21:26.880 --> 01:21:31.517 Uh I uh I just coded it all up and then 01:21:31.600 --> 01:21:35.677 there like I like put it out on one of these there's all these like websites 01:21:35.760 --> 01:21:37.677 where you can launch startups and I put 01:21:37.760 --> 01:21:39.437 it out on we put it out on one of those 01:21:39.520 --> 01:21:44.077 websites and uh it went like microviral 01:21:44.160 --> 01:21:49.677 you know um like viral among like people 01:21:49.760 --> 01:21:53.037 who were on Twitter to look for new 01:21:53.120 --> 01:21:55.917 startup ideas. Um, and then it was kind 01:21:56.000 --> 01:22:00.557 of that was like the early seed that just that ended up uh enabling 01:22:00.640 --> 01:22:05.197 everything to grow. But it was like I mean 01:22:05.280 --> 01:22:09.757 at the time it was I mean it was it was tough going, you know? You you're like 01:22:09.840 --> 01:22:13.677 like I would just like I would just spend all my time coding. Then every 01:22:13.760 --> 01:22:16.637 once in a while I would like post 01:22:16.720 --> 01:22:20.877 something to the internet and just like and then I would beg all of my friends. 01:22:20.960 --> 01:22:24.237 I was like I would say like please go upote this, please go like this, like 01:22:24.320 --> 01:22:27.757 please like you know give me some ounce 01:22:27.840 --> 01:22:31.437 of traction and uh yeah that was the early days. 01:22:31.520 --> 01:22:34.157 >> Damn. Was it scale AI at the beginning? 01:22:34.240 --> 01:22:36.957 >> Yeah, scale AI actually it was called it 01:22:37.040 --> 01:22:40.797 was uh it was scale API at first and 01:22:40.880 --> 01:22:44.317 then because that was just like that website was available and then it became 01:22:44.400 --> 01:22:47.757 scale AI like a year and a half later. 01:22:47.840 --> 01:22:52.237 Um but uh yeah so so the whole the whole 01:22:52.320 --> 01:22:56.877 I mean early startups are so gnarly. Um it's I mean it's really crazy if you 01:22:56.960 --> 01:22:59.597 look at like all these big companies and you like you know think about what they 01:22:59.680 --> 01:23:01.437 were like in the early days. They're all 01:23:01.520 --> 01:23:03.757 they're all pretty pretty uh pretty 01:23:03.840 --> 01:23:07.037 rough and tumble. Um but but the coolest 01:23:07.120 --> 01:23:09.117 thing like we cuz we started working 01:23:09.200 --> 01:23:12.317 with all these automotive companies and 01:23:12.400 --> 01:23:15.437 working on self-driving. Um 01:23:15.520 --> 01:23:18.237 uh it quickly became 01:23:18.320 --> 01:23:19.997 hyperinteresting [snorts] 01:23:20.080 --> 01:23:22.717 um because you know this was like one of 01:23:22.800 --> 01:23:25.997 the great scientific and and and 01:23:26.080 --> 01:23:29.037 engineering challenges of the time um 01:23:29.120 --> 01:23:31.837 and uh and we ultimately ended up being 01:23:31.920 --> 01:23:33.997 successful like Whimo one of our 01:23:34.080 --> 01:23:38.557 customers is now launched and driving um 01:23:38.640 --> 01:23:43.677 large scale robo taxi services in you know San Francisco, LA, Phoenix, they're 01:23:43.760 --> 01:23:45.437 launching in more cities. Like, >> wow. 01:23:45.520 --> 01:23:47.197 >> Um, it's pretty amazing. 01:23:47.280 --> 01:23:51.037 >> Wow. Damn. And the company grew how 01:23:51.120 --> 01:23:55.757 fast? >> So, let's see. I think the numbers are 01:23:55.840 --> 01:23:59.837 something like >> 5 years. You are the youngest 5 years 01:23:59.920 --> 01:24:01.997 from when you started it. You become the 01:24:02.080 --> 01:24:04.477 youngest billionaire in the world. 01:24:04.560 --> 01:24:08.637 >> Yeah, that's crazy to think about. That did not feel obvious. The first year it 01:24:08.720 --> 01:24:11.197 was like it was like 01:24:11.280 --> 01:24:17.037 for the first ye first 12 months it was like 01:24:17.120 --> 01:24:21.517 one to three people like it was like it was like almost nobody. It was like me 01:24:21.600 --> 01:24:24.637 and like one or two other people working on it for the first year. 01:24:24.720 --> 01:24:26.717 >> That's it for the first one year. 01:24:26.800 --> 01:24:30.237 >> And then after the second year we go 01:24:30.320 --> 01:24:32.477 from that like 01:24:32.560 --> 01:24:35.757 one to three people and we start hiring 01:24:35.840 --> 01:24:39.597 more people. we get to maybe I think 01:24:39.680 --> 01:24:42.157 like like 15 or so people 01:24:42.240 --> 01:24:47.197 and then that third year we went from 15 01:24:47.280 --> 01:24:52.237 or so people to like maybe 100 01:24:52.320 --> 01:24:53.677 and then we were kind of off the then it 01:24:53.760 --> 01:24:55.437 was like 100 and then we and then we 01:24:55.520 --> 01:24:59.837 were like 200 and then 500 and then we 01:24:59.920 --> 01:25:01.917 kept growing and now we're up to like,00 01:25:02.000 --> 01:25:05.117 people um but the f it was like really 01:25:05.200 --> 01:25:09.917 slow going at first and um 01:25:10.000 --> 01:25:12.637 yeah and we we we 01:25:12.720 --> 01:25:15.997 uh we focused on first it was autonomous 01:25:16.080 --> 01:25:17.517 driving 01:25:17.600 --> 01:25:22.477 and then and then starting um starting 01:25:22.560 --> 01:25:26.477 about 3 years in we started focusing on defense 01:25:26.560 --> 01:25:28.157 um and working with the DoD. 01:25:28.240 --> 01:25:30.717 >> What are you guys doing in defense? 01:25:30.800 --> 01:25:35.037 >> So uh we do we do a few things. So, one 01:25:35.120 --> 01:25:38.317 of the the first things we did was help 01:25:38.400 --> 01:25:42.717 the DoD with its its own data problem to 01:25:42.800 --> 01:25:46.317 help them be able to train AI systems. 01:25:46.400 --> 01:25:49.277 So, um you know, one of the first things 01:25:49.360 --> 01:25:51.037 that we worked on was like, you know, 01:25:51.120 --> 01:25:53.757 they wanted to the DoD wanted to do 01:25:53.840 --> 01:25:57.277 image recognition on satellite imagery, 01:25:57.360 --> 01:26:02.397 SAR imagery, you know, other like all forms of overhead imagery. But they had 01:26:02.480 --> 01:26:07.517 this huge data problem, you know, just like me with the fridge. Um they had the 01:26:07.600 --> 01:26:10.717 same problem like how, you know, they need to be able to have data that lets 01:26:10.800 --> 01:26:13.197 them detect things and all this imagery. 01:26:13.280 --> 01:26:15.677 And so we the first thing we did was 01:26:15.760 --> 01:26:18.557 fuel um the data sets and data 01:26:18.640 --> 01:26:21.037 capabilities for the DoD. That was true 01:26:21.120 --> 01:26:24.157 for the first few years. And then um 01:26:24.240 --> 01:26:28.877 more recently, we've been working with them to do large scale fielding of AI 01:26:28.960 --> 01:26:30.877 capabilities. what what what what kind 01:26:30.960 --> 01:26:34.637 of stuff is DoD looking for in imagery? 01:26:34.720 --> 01:26:36.637 >> Uh so I mean 01:26:36.720 --> 01:26:41.987 >> so let me let me let me also so basically the way I understand this is 01:26:42.070 --> 01:26:45.677 [clears throat] >> you don't need a human to detect 01:26:45.760 --> 01:26:48.317 something maybe like a nuclear reactor. 01:26:48.400 --> 01:26:50.957 Is that is it am I on the right track here? 01:26:51.040 --> 01:26:55.517 >> So they look >> or a missile silo or Yeah. And so AI is 01:26:55.600 --> 01:26:57.757 detecting all these which drastically 01:26:57.840 --> 01:27:00.077 reduces human error, human manpower, all 01:27:00.160 --> 01:27:03.277 that kind of stuff. It's more accurate. 01:27:03.360 --> 01:27:05.037 >> Yeah. And it's and I mean mostly it's 01:27:05.120 --> 01:27:08.637 it's scalable. Like I mean that we we 01:27:08.720 --> 01:27:10.397 the number of satellites in space has 01:27:10.480 --> 01:27:13.037 like exploded. So we have so much more 01:27:13.120 --> 01:27:16.397 sensing um today. Like way more imagery, 01:27:16.480 --> 01:27:20.157 way more sensing today than it's even like feasible for humans to to work 01:27:20.240 --> 01:27:24.797 their way through. >> Wow. Um so that was yeah that was like 01:27:24.880 --> 01:27:27.837 the first problem. >> How do you fuel it? 01:27:27.920 --> 01:27:31.597 >> You well you have to build so there's 01:27:31.680 --> 01:27:36.477 there's two parts. So first you have to build effectively like a data foundry. 01:27:36.560 --> 01:27:38.077 You have to build a a mechanism by which 01:27:38.160 --> 01:27:40.317 you're able to generate lots and lots of 01:27:40.400 --> 01:27:44.397 data to fuel these algorithms. um a lot 01:27:44.480 --> 01:27:48.317 of it synthetically so using the algorithms themselves to generate the 01:27:48.400 --> 01:27:53.837 data but then a lot of it you still need humans to validate and verify. So one of 01:27:53.920 --> 01:27:55.197 the things we did actually for this 01:27:55.280 --> 01:27:58.317 whole project is um we created a 01:27:58.400 --> 01:28:02.397 facility in St. Louis, Missouri, next to 01:28:02.480 --> 01:28:04.477 uh NGA, the National Geospatial 01:28:04.560 --> 01:28:07.917 Intelligence Agency. And we produced a 01:28:08.000 --> 01:28:12.157 center for AI data processing where we 01:28:12.240 --> 01:28:15.277 uh hired up uh imagery analysts to be 01:28:15.360 --> 01:28:17.757 able to validate the outputs coming out 01:28:17.840 --> 01:28:22.637 of the AI systems to ensure that um we were getting the correct, you know, 01:28:22.720 --> 01:28:27.197 we're getting accurate and high integrity data to feed back into the AI systems. 01:28:27.280 --> 01:28:32.557 >> Wow. Wow. Damn. 01:28:32.640 --> 01:28:36.397 Where do we go from here? >> Yeah. So then, so we were doing, so we 01:28:36.480 --> 01:28:38.637 were doing um lots of stuff around 01:28:38.720 --> 01:28:43.037 imagery and computer vision and then um 01:28:43.120 --> 01:28:44.957 uh and then we started working with the 01:28:45.040 --> 01:28:50.317 DoD on, you know, more ambitious and larger scale AI projects. So one of the 01:28:50.400 --> 01:28:51.677 things we're working with them now is 01:28:51.760 --> 01:28:53.917 this program called Thunderforge which 01:28:54.000 --> 01:28:57.997 is using AI for um military planning and 01:28:58.080 --> 01:29:02.237 operational planning. So um more broadly 01:29:02.320 --> 01:29:04.237 so so the basic idea here is can you use 01:29:04.320 --> 01:29:08.317 AI to effectively like automate major 01:29:08.400 --> 01:29:13.037 parts of the military planning process so that you're able to plan within hours 01:29:13.120 --> 01:29:17.997 versus taking many days. >> This sounds like Palunteer. It's um yeah 01:29:18.080 --> 01:29:20.957 they target different parts of the problem and we target different parts of 01:29:21.040 --> 01:29:22.157 the problem and ultimately we work 01:29:22.240 --> 01:29:24.717 together pretty well but the this is 01:29:24.800 --> 01:29:28.157 part of a broader concept that we have 01:29:28.240 --> 01:29:30.237 around agent what we call agentic 01:29:30.320 --> 01:29:33.997 warfare. So the use of AI and AI agents 01:29:34.080 --> 01:29:37.597 in warfare and the basic idea is can you 01:29:37.680 --> 01:29:40.957 go from these current processes where 01:29:41.040 --> 01:29:46.157 humans are the loop to humans being on the loop. And so can you go from you 01:29:46.240 --> 01:29:49.517 know situations where you know these workflows have to go from a person has 01:29:49.600 --> 01:29:51.917 to do a bunch of work then pass the next person they have to do a bunch of work 01:29:52.000 --> 01:29:55.037 pass the next person to the AI agents 01:29:55.120 --> 01:29:57.037 are just doing a lot of that work and 01:29:57.120 --> 01:29:58.957 humans are just checking and verifying 01:29:59.040 --> 01:30:02.397 along the way. Um and it's it's a big 01:30:02.480 --> 01:30:05.117 change. So going from you know you know 01:30:05.200 --> 01:30:07.597 if you compare both set setups side by 01:30:07.680 --> 01:30:10.157 side here you have individuals humans 01:30:10.240 --> 01:30:14.877 with decades of single domain experience who are doing each step step of this 01:30:14.960 --> 01:30:19.277 process. And then if you have the AI agents doing it, ideally you have AI 01:30:19.360 --> 01:30:21.597 agents who have, you know, thousands of 01:30:21.680 --> 01:30:24.397 years of of knowledge, all domain 01:30:24.480 --> 01:30:28.477 knowledge and are are, you know, a 01:30:28.560 --> 01:30:30.957 thousand times faster at under at doing 01:30:31.040 --> 01:30:33.597 the actual tasks. And so it's all about 01:30:33.680 --> 01:30:36.397 taking and this exists at many many 01:30:36.480 --> 01:30:41.037 different levels. So you know there's you can think about this for the sensing 01:30:41.120 --> 01:30:44.717 and intel portion that we're talking about before. So you know can you 01:30:44.800 --> 01:30:48.637 accelerate the intelligence gathering you know the process by which we take 01:30:48.720 --> 01:30:52.797 all the sensor data and turn that into insight. You can think about it for the 01:30:52.880 --> 01:30:54.797 operational planning process like how 01:30:54.880 --> 01:30:57.437 can you accelerate that uh that entire 01:30:57.520 --> 01:31:02.077 flow. You can think about it in terms of um you know on the tactical side, how do 01:31:02.160 --> 01:31:03.837 you accelerate tactical decision- 01:31:03.920 --> 01:31:07.917 making? Um so it you bleeds into every 01:31:08.000 --> 01:31:09.757 sort of like level of warfare, every or 01:31:09.840 --> 01:31:12.397 every component, but at its core, how do 01:31:12.480 --> 01:31:18.557 you use AI agents to be faster, more adaptive, and have humans just check 01:31:18.640 --> 01:31:21.117 their work? >> So when you're talking about it helps 01:31:21.200 --> 01:31:23.357 with mission planning, especially in a 01:31:23.440 --> 01:31:24.957 tactical environment, cuz that's where I 01:31:25.040 --> 01:31:29.037 come from. I mean what what is it could 01:31:29.120 --> 01:31:31.357 be any example but g can you give me an 01:31:31.440 --> 01:31:36.317 example of how it speeds up the mission planning process in a tactical 01:31:36.400 --> 01:31:40.877 environment. >> Yeah. So, so let's say that. So, this 01:31:40.960 --> 01:31:43.197 thing that we have, by the way, you know, we're working on it with 01:31:43.280 --> 01:31:46.157 Indopaccom and Yukcom right now. And and 01:31:46.240 --> 01:31:49.677 um we'll deploy more broadly, but um 01:31:49.760 --> 01:31:53.917 let's say that there's um a uh what's a 01:31:54.000 --> 01:31:57.997 good what's a good example? Um let's say there's some kind of alert that pops up, 01:31:58.080 --> 01:32:02.397 like there's something that um we didn't expect that we need to figure out how 01:32:02.480 --> 01:32:06.717 we're going to respond to. Um >> like like what kind of an alert? So I 01:32:06.800 --> 01:32:11.757 mean let's say there was like uh you know um there's like I mean you can 01:32:11.840 --> 01:32:14.477 imagine at different levels but let's say there's like a ship that popped up 01:32:14.560 --> 01:32:17.277 that we didn't expect >> as a simple example. 01:32:17.360 --> 01:32:20.877 >> So then that alert flows into a bunch of 01:32:20.960 --> 01:32:23.037 AI systems that are going to the first 01:32:23.120 --> 01:32:26.717 step is sensing. So what like let's look 01:32:26.800 --> 01:32:29.437 through all of our sensing capabilities 01:32:29.520 --> 01:32:32.717 and um let's like go reanalyze all of 01:32:32.800 --> 01:32:35.117 the data that we have and figure out how 01:32:35.200 --> 01:32:40.237 much do we know about that ship right so now a person would like an analyst would 01:32:40.320 --> 01:32:43.357 go through and like do all this you know all the ped and all this stuff to to be 01:32:43.440 --> 01:32:46.797 able to undergo this work but ideally you have AI agents that are just going 01:32:46.880 --> 01:32:51.197 they can look through all the historical sensor data they can figure out um oh 01:32:51.280 --> 01:32:53.757 actually there's like kind of a thing that showed up on this radar and there's 01:32:53.840 --> 01:32:56.237 kind thing that showed up on this satellite imagery and we can kind of 01:32:56.320 --> 01:33:00.957 like sketch together this like you know the trajectory of this of this ship. 01:33:01.040 --> 01:33:03.517 Okay. So you go through that process you try to understand what's going on and 01:33:03.600 --> 01:33:06.797 then you and then you go through and and 01:33:06.880 --> 01:33:08.957 figure out okay what are the what are 01:33:09.040 --> 01:33:11.997 the possible um courses of actions. So 01:33:12.080 --> 01:33:14.157 once you have situational awareness, 01:33:14.240 --> 01:33:16.317 then what are the courses of actions 01:33:16.400 --> 01:33:19.837 against this particular scenario and you 01:33:19.920 --> 01:33:22.957 can have an AI agent honestly just 01:33:23.040 --> 01:33:27.357 propose courses of actions. Um like hey in this scenario given this ship is is 01:33:27.440 --> 01:33:30.877 coming here you know we could fire at it. We could just wait to see what 01:33:30.960 --> 01:33:33.597 happens. We could reposition so that 01:33:33.680 --> 01:33:38.717 we're you know we're able to to um you know handle the threat better. you know, 01:33:38.800 --> 01:33:41.357 all sorts of we could we could reposition some satellites so we have 01:33:41.440 --> 01:33:42.557 greater sensing. You know, there's all 01:33:42.640 --> 01:33:44.717 sorts of different courses of actions um 01:33:44.800 --> 01:33:48.877 that we could take. And then uh once the 01:33:48.960 --> 01:33:51.597 AI produces those course of actions, 01:33:51.680 --> 01:33:56.557 it'll run each of those different course of actions through a simulator. So it'll 01:33:56.640 --> 01:34:00.317 then run uh >> it war games at real time. 01:34:00.400 --> 01:34:01.917 >> Exactly. It'll war game at real time. 01:34:02.000 --> 01:34:04.877 And so then it'll run through a simulator and say, "Okay, what's going 01:34:04.960 --> 01:34:09.837 to happen if we fire at it?" like, you know, this is what we know about red 01:34:09.920 --> 01:34:13.757 forces. This is what we know about blue forces right now. Um, if we fire at it, 01:34:13.840 --> 01:34:18.557 this is like, you know, this is the war game of how that plays out. If we um 01:34:18.640 --> 01:34:22.237 just increase our sensing, like these are the things that that the red forces 01:34:22.320 --> 01:34:26.877 could do to [__] us up. And like that's the risk that we take on. And um and 01:34:26.960 --> 01:34:31.277 then the benefit is because all of this is automatic, you can run it these war 01:34:31.360 --> 01:34:32.797 games and these simulations a million 01:34:32.880 --> 01:34:36.717 times. So it's not just like one, you 01:34:36.800 --> 01:34:41.437 know, military planners just like trying to like war game and plan it out like 01:34:41.520 --> 01:34:45.677 you know in human time. It's like you could run a million simulations cuz you 01:34:45.760 --> 01:34:47.677 don't have perfect information. You don't have perfect knowledge. So you 01:34:47.760 --> 01:34:50.557 need to kind of figure out based on the 01:34:50.640 --> 01:34:54.797 uncertainties of the situation, what are all the potential outcomes that that pop 01:34:54.880 --> 01:34:58.317 out of that? >> Wow. And then so you run like a million 01:34:58.400 --> 01:35:00.397 different simulations of each of these 01:35:00.480 --> 01:35:02.637 different courses of action. And then 01:35:02.720 --> 01:35:06.797 you can give a commander direct like you 01:35:06.880 --> 01:35:11.037 give them this whole like brief and presentation which is basically these 01:35:11.120 --> 01:35:15.357 are the courses of actions we considered. This is the this these are 01:35:15.440 --> 01:35:17.197 the likely outcomes in those courses of 01:35:17.280 --> 01:35:21.037 action. We can show you the simulated 01:35:21.120 --> 01:35:25.037 like outcome in each one of these scenarios. So we can like show you what 01:35:25.120 --> 01:35:28.317 it would look like in every one of those scenarios if it happened like 01:35:28.400 --> 01:35:31.117 representative um simulations and then 01:35:31.200 --> 01:35:35.997 the commander makes a call. >> Wow. So it's this is what it is. This is 01:35:36.080 --> 01:35:39.997 what it's doing. These are the possible courses of action. These are the 01:35:40.080 --> 01:35:43.517 consequences of each action. This is the percentage. 01:35:43.600 --> 01:35:47.677 >> Yeah. Exactly. >> And and it spits that out in what? A matter of seconds. 01:35:47.760 --> 01:35:51.277 >> Yeah. Now it takes a you know probably takes even now it probably takes a few 01:35:51.360 --> 01:35:53.917 hours cuz you know these models are a lot slower than they will be in the 01:35:54.000 --> 01:35:58.797 future but yeah I mean compare that to I mean depending on the situation like 01:35:58.880 --> 01:36:02.797 that could take you know that could take days for humans to do today like it's 01:36:02.880 --> 01:36:07.917 and and it's not from lack of will or effort or or capability. It's just 01:36:08.000 --> 01:36:11.597 >> it's a really complicated situation. If a ship pops up out of nowhere like 01:36:11.680 --> 01:36:12.877 >> there's a lot of stuff you have to 01:36:12.960 --> 01:36:16.157 consider. Um, and so, uh, that's really 01:36:16.240 --> 01:36:18.877 the the the step change here is just 01:36:18.960 --> 01:36:22.797 like a, uh, like dramatically 01:36:22.880 --> 01:36:26.557 accelerating situational awareness, dramatically accelerating like an 01:36:26.640 --> 01:36:28.877 understanding of what the different course of actions are, what could 01:36:28.960 --> 01:36:32.957 happen, what are the consequences, um, and surfacing that to commander. 01:36:33.040 --> 01:36:35.357 >> Does it make a recommendation? 01:36:35.440 --> 01:36:37.837 >> Um, this is kind of an interesting 01:36:37.920 --> 01:36:40.157 thing. We 01:36:40.240 --> 01:36:42.557 we go back and forth if we want to make 01:36:42.640 --> 01:36:45.517 a recommendation because ultimately like 01:36:45.600 --> 01:36:49.837 we don't want um to just be like you 01:36:49.920 --> 01:36:53.037 know we don't want to let commanders kind of like sleepwalk if that makes 01:36:53.120 --> 01:36:55.197 sense. We want them to like, you know, 01:36:55.280 --> 01:36:58.637 our military commanders are the best 01:36:58.720 --> 01:37:02.877 humans in the world like considering all of the potential consequences of these 01:37:02.960 --> 01:37:04.717 different course of action and also 01:37:04.800 --> 01:37:09.917 considering you know um uh and and 01:37:10.000 --> 01:37:13.197 ultimately making a call based on those potential consequences. So I think we 01:37:13.280 --> 01:37:16.077 want to ensure that 01:37:16.160 --> 01:37:20.317 commanders are still exercising their judgment in these decisions versus just, 01:37:20.400 --> 01:37:26.760 you know, making it easy for them to just say, "Oh, go with what the AI says." 01:37:27.200 --> 01:37:29.997 >> Interesting. Wow. 01:37:30.080 --> 01:37:32.317 But this but then okay, think about what 01:37:32.400 --> 01:37:35.277 happens next. So um and this is where 01:37:35.360 --> 01:37:38.157 stuff gets really freaky. So, 01:37:38.240 --> 01:37:42.397 let's say that um obviously in a world where just the blue force, just the 01:37:42.480 --> 01:37:45.277 United States has this capability, that's great. You know, we're going to 01:37:45.360 --> 01:37:46.877 we're going to be running circles around 01:37:46.960 --> 01:37:49.757 everyone else. Um but then what happens 01:37:49.840 --> 01:37:54.557 if the red force, you know, China, Russia, whomever also has the 01:37:54.640 --> 01:37:56.717 capability? Then you're in this 01:37:56.800 --> 01:38:02.077 situation where I've war gamed out the whole situation. 01:38:02.160 --> 01:38:03.597 you know, they've instantaneously 01:38:03.680 --> 01:38:06.237 wargamed out the whole situation and 01:38:06.320 --> 01:38:09.677 then it's like then then it I think I 01:38:09.760 --> 01:38:13.037 honestly think so then it's like we know 01:38:13.120 --> 01:38:15.837 and you know like blue forces, red 01:38:15.920 --> 01:38:19.117 forces, we both know that we both have 01:38:19.200 --> 01:38:23.437 like you know this perfectly war game scenarios. Which avenue do you pick? And 01:38:23.520 --> 01:38:28.077 then it becomes this really complicated almost like psychological 01:38:28.160 --> 01:38:32.637 you know kind of kind of situation which is like then it like all comes down to 01:38:32.720 --> 01:38:36.237 how good our intel is. So how good is our intel about that commander? How good 01:38:36.320 --> 01:38:40.397 is our intel about what their collection capabilities are? How good is our intel 01:38:40.480 --> 01:38:45.357 about you know what they likely know about us and vice versa. Um and it gets 01:38:45.440 --> 01:38:46.877 pretty 01:38:46.960 --> 01:38:49.997 so this is actually 01:38:50.080 --> 01:38:53.037 let's just 01:38:53.120 --> 01:38:56.797 so let's say China Russia 01:38:56.880 --> 01:38:59.037 our enemies have this capability we have 01:38:59.120 --> 01:39:05.640 this capability then it then it kind of becomes 01:39:07.840 --> 01:39:09.837 it's like the same process that we deal 01:39:09.920 --> 01:39:12.077 with now who has the better intel right 01:39:12.160 --> 01:39:15.677 it's just develop flipping and and and 01:39:15.760 --> 01:39:20.397 you're going to a course of action quicker and the enem is doing the exact 01:39:20.480 --> 01:39:24.477 same thing quicker. So it's essentially it's the exact same thing that we're 01:39:24.560 --> 01:39:27.197 doing now 01:39:27.280 --> 01:39:30.557 but faster and so if we develop it first 01:39:30.640 --> 01:39:32.797 then 01:39:32.880 --> 01:39:36.637 we achieve basically global domination. 01:39:36.720 --> 01:39:38.237 Am I correct here? 01:39:38.320 --> 01:39:40.797 >> Yeah. I think and I think timing really 01:39:40.880 --> 01:39:43.677 matters here because if we get this 01:39:43.760 --> 01:39:47.917 capability and this will go for I mean there's like 01:39:48.000 --> 01:39:51.117 there's way more there there's way more AI we'll be able to do but let's say we 01:39:51.200 --> 01:39:54.397 get this capability you know a year 01:39:54.480 --> 01:39:59.517 ahead of adversaries then you're then like we're just going 01:39:59.600 --> 01:40:00.877 to be able to respond so much faster. 01:40:00.960 --> 01:40:04.877 the the analogy I often use is like imagine we were playing chess, but for 01:40:04.960 --> 01:40:07.197 every one move you take, I can take 10 01:40:07.280 --> 01:40:09.437 moves. Like I'm just going to win. 01:40:09.520 --> 01:40:12.637 >> Um uh and that's what that's the 01:40:12.720 --> 01:40:15.757 asymmetric advantage that that comes out 01:40:15.840 --> 01:40:19.117 of this of this capability. Um and then 01:40:19.200 --> 01:40:23.597 once it but then once it equalizes then 01:40:23.680 --> 01:40:27.117 then it's like this very you know it's like to your point becomes this like 01:40:27.200 --> 01:40:29.597 adversarial intelbased you know 01:40:29.680 --> 01:40:32.537 capability based kind of conflict. How 01:40:32.620 --> 01:40:37.720 [clears throat] how do we 01:40:40.480 --> 01:40:44.397 I mean how do we combat our adversaries 01:40:44.480 --> 01:40:49.517 from having this type of intel from having this type of AI system? 01:40:49.600 --> 01:40:51.357 So 01:40:51.440 --> 01:40:54.557 I think then then 01:40:54.640 --> 01:40:57.677 I mean China's demonstrated with 01:40:57.760 --> 01:41:01.117 deepseek uh and you know models have 01:41:01.200 --> 01:41:02.797 come out since then they're going to be 01:41:02.880 --> 01:41:06.637 very competitive on AI and in 01:41:06.720 --> 01:41:10.877 uh I think in 2024 so last year there 01:41:10.960 --> 01:41:16.397 were something like 80 contracts between 01:41:16.480 --> 01:41:18.637 uh large language model AI companies in 01:41:18.720 --> 01:41:20.957 China and the People Liberations Army, 01:41:21.040 --> 01:41:25.837 the PLA, um that number is not 80 in the United 01:41:25.920 --> 01:41:27.037 States. Like the United States is like 01:41:27.120 --> 01:41:30.557 way way less than 80. So they're very 01:41:30.640 --> 01:41:33.357 clearly accelerating the integration of 01:41:33.440 --> 01:41:36.397 AI into their um national security and 01:41:36.480 --> 01:41:41.037 into their military apparatus very quickly. I don't think at this point 01:41:41.120 --> 01:41:43.277 realistically we can stop them from 01:41:43.360 --> 01:41:46.637 having this this capability I described. 01:41:46.720 --> 01:41:49.277 So then you go to the next layer down. 01:41:49.360 --> 01:41:51.037 So 01:41:51.120 --> 01:41:55.997 uh Intel. So um well the next layer down 01:41:56.080 --> 01:41:58.877 the next two things that you look at is 01:41:58.960 --> 01:42:01.757 okay how does AI impact Intel and how 01:42:01.840 --> 01:42:04.317 does AI how can we what is the 01:42:04.400 --> 01:42:07.437 adversarial AI dynamic? Like can we use 01:42:07.520 --> 01:42:09.517 our AIs to sabotage their AIs? Can they 01:42:09.600 --> 01:42:12.317 use their AIS to sabotage sabotage ours? 01:42:12.400 --> 01:42:15.597 Um and it's like AI on AI warfare 01:42:15.680 --> 01:42:18.477 effectively. Um then when you look at 01:42:18.560 --> 01:42:20.637 that scenario, okay, so the so let's dig 01:42:20.720 --> 01:42:23.197 into that. Um 01:42:23.280 --> 01:42:26.557 the first level analysis here is kind of 01:42:26.640 --> 01:42:27.837 what we were talking about before, which 01:42:27.920 --> 01:42:30.717 is that probably just boils down to how 01:42:30.800 --> 01:42:35.197 many copies of these AI systems do I have running versus how many copies do 01:42:35.280 --> 01:42:36.557 you have running. So it turns into a 01:42:36.640 --> 01:42:40.957 numbers game. If I have 10,000 AI copies 01:42:41.040 --> 01:42:45.197 running and you only have 100 AI copies running, then I'm going to run circle. 01:42:45.280 --> 01:42:49.917 I'm still going to run circles around you. And that boils down to who has 01:42:50.000 --> 01:42:51.357 >> So, 01:42:51.440 --> 01:42:55.277 >> so let's say um so let's say you have 01:42:55.360 --> 01:42:57.997 you have 100 AIs. I have I have 10,000 01:42:58.080 --> 01:43:01.917 AIS. Um I will I will take half of my 01:43:02.000 --> 01:43:05.037 AIs. I will take 5,000 of my AIS and 01:43:05.120 --> 01:43:08.637 just focus them on hacking your AIS. So, 01:43:08.720 --> 01:43:10.237 I'm they're all going to be looking for 01:43:10.320 --> 01:43:14.637 vulnerabilities in your um in your uh in 01:43:14.720 --> 01:43:18.637 your information architecture, in your data centers. I'm going to look for 01:43:18.720 --> 01:43:21.357 vulner, you know, I'm just like purely 01:43:21.440 --> 01:43:24.317 focused on cyber hacking of your 100 AIS 01:43:24.400 --> 01:43:27.117 and then my other 5,000 copies are going 01:43:27.200 --> 01:43:28.717 to do the military planning process for 01:43:28.800 --> 01:43:32.957 myself. Um then then look at think about 01:43:33.040 --> 01:43:34.717 the adversary. I have this choice. I 01:43:34.800 --> 01:43:38.077 have a 100 AIS. If I have them all focus 01:43:38.160 --> 01:43:41.037 on doing um the military planning 01:43:41.120 --> 01:43:45.117 process, I'm going to get hacked cuz I'm not doing any cyber defense. And then 01:43:45.200 --> 01:43:48.477 even if I have all of them focus on cyber defense, even those numbers are 01:43:48.560 --> 01:43:50.797 bad. It's like 100 AIs versus 5,000 AIs 01:43:50.880 --> 01:43:53.037 from you. And so I probably still get 01:43:53.120 --> 01:43:56.317 hacked. So the numbers end up mattering 01:43:56.400 --> 01:44:00.557 a lot. Um uh if even if they had even if 01:44:00.640 --> 01:44:04.157 the other adversary, let's say it's only a 2x advantage. I have 10,000 copies 01:44:04.240 --> 01:44:07.677 running and the adversary is 5,000 copies running. I can do the same thing. 01:44:07.760 --> 01:44:12.077 5,000 of my copies are just focused on hacking your AI so that your AI is 01:44:12.160 --> 01:44:17.117 incapacitated or has incorrect information or um or is poisoned in some 01:44:17.200 --> 01:44:22.317 way like basically is incapable incapacity for some reason. Um and the 01:44:22.400 --> 01:44:25.517 other half of my eyes are focused on the military planning process. Again, the 01:44:25.600 --> 01:44:28.797 adversary is screwed cuz to properly 01:44:28.880 --> 01:44:33.437 deal with a cyber attack, I need probably all 5,000 copies to be focused 01:44:33.520 --> 01:44:37.277 on cyber defense and then I have no capacity left to do 01:44:37.360 --> 01:44:38.717 the military planning. 01:44:38.800 --> 01:44:42.797 >> Wow. So it really turns into this like 01:44:42.880 --> 01:44:44.797 uh very 01:44:44.880 --> 01:44:47.117 like just in the same way that you would 01:44:47.200 --> 01:44:50.077 you would command your forces um today 01:44:50.160 --> 01:44:51.997 like all of your your you know your 01:44:52.080 --> 01:44:54.797 various your forces across all domains 01:44:54.880 --> 01:44:57.277 to like try to pins or outmaneuver the 01:44:57.360 --> 01:45:01.917 enemy. You'll do the same kind of planning for your like AI army so to 01:45:02.000 --> 01:45:04.397 speak or your AI >> allocation of assets. 01:45:04.480 --> 01:45:05.757 >> Yeah. Your allocation of assets. 01:45:05.840 --> 01:45:08.397 Exactly. And a lot of it will be okay, 01:45:08.480 --> 01:45:12.157 how many am I dedicating towards um uh 01:45:12.240 --> 01:45:15.277 hacking and sabotaging the opponent? How 01:45:15.360 --> 01:45:17.197 many am I dedicating towards my own 01:45:17.280 --> 01:45:20.637 military planning and wargaming process? 01:45:20.720 --> 01:45:23.037 Um the other thing is how many you 01:45:23.120 --> 01:45:27.837 allocate towards um uh towards you know 01:45:27.920 --> 01:45:29.997 the the other key component here is 01:45:30.080 --> 01:45:33.037 drones and uh how many you're allocating 01:45:33.120 --> 01:45:36.157 towards doing the like very tactical 01:45:36.240 --> 01:45:38.157 mission level autonomy to accomplish you 01:45:38.240 --> 01:45:41.037 know mission level objectives. Um, but 01:45:41.120 --> 01:45:44.237 it'll be it'll be like I think it really 01:45:44.320 --> 01:45:48.317 boils down to ultimately who has more resources and then what are those 01:45:48.400 --> 01:45:50.797 resources? That's going to be about 01:45:50.880 --> 01:45:53.917 large scale data centers. So who has 01:45:54.000 --> 01:45:56.717 bigger data centers and more power to 01:45:56.800 --> 01:46:00.037 run all these AI agents? 01:46:00.120 --> 01:46:04.077 [clears throat] And who who makes the determination of 01:46:04.160 --> 01:46:08.317 how many AIs we're going to put in tactical environment? How many AIs are 01:46:08.400 --> 01:46:12.797 going to go after cyber security trying to hack into the 01:46:12.880 --> 01:46:15.117 other AIs? Is that a human or is that 01:46:15.200 --> 01:46:20.237 another layer of AI that that that spits out 01:46:20.320 --> 01:46:24.877 exactly what you just said? This is what we this is our situation. Here's the 01:46:24.960 --> 01:46:27.117 courses of action. 01:46:27.200 --> 01:46:30.077 Here's the here's the consequences of 01:46:30.160 --> 01:46:35.277 what happened. So is it just AI after AI after AI that's doing all of this all 01:46:35.360 --> 01:46:38.557 these simulations? >> Yeah, then yeah. No, you're exactly 01:46:38.640 --> 01:46:42.157 right. Then I Yeah, exactly. You have another AI that's planning out and 01:46:42.240 --> 01:46:44.397 mapping out, you know, how should I 01:46:44.480 --> 01:46:48.237 allocate my AI resources to properly 01:46:48.320 --> 01:46:51.757 deal with the adversary given what I know about the adversary and then the so 01:46:51.840 --> 01:46:56.397 then what are the ways in which you know what are so then what are the key 01:46:56.480 --> 01:46:58.317 dimensions that would give you an edge 01:46:58.400 --> 01:47:01.917 versus your adversary? Well, it's if a 01:47:02.000 --> 01:47:04.957 your AI is different somehow. So, it 01:47:05.040 --> 01:47:07.597 actually is like hard for your adversary 01:47:07.680 --> 01:47:12.557 to know exactly how you're how you would act like basically strategic surprise in 01:47:12.640 --> 01:47:16.237 some form in the form of like a different thinking process or a 01:47:16.320 --> 01:47:18.397 different sort of like way of reasoning 01:47:18.480 --> 01:47:20.957 of the AI systems. Um, and then the 01:47:21.040 --> 01:47:24.477 other one is like uh ambiguity of how 01:47:24.560 --> 01:47:26.637 many how much what your resources 01:47:26.720 --> 01:47:29.917 actually are. Like if somehow I can make 01:47:30.000 --> 01:47:34.477 the adversary think that I have way fewer resources than I actually do or 01:47:34.560 --> 01:47:36.717 way more resources than I actually do, 01:47:36.800 --> 01:47:41.277 that'll be a critical element of of um 01:47:41.360 --> 01:47:45.037 yeah, of strategic surprise in those kinds of situations as well. 01:47:45.120 --> 01:47:48.237 >> Wow. Would an AI be able to be able to 01:47:48.320 --> 01:47:53.037 would would AI be able to 01:47:53.120 --> 01:47:56.797 alert if it if it [clears throat] will it know 01:47:56.880 --> 01:47:58.957 it's been hacked? 01:47:59.040 --> 01:48:01.517 So yeah, this is this is a great 01:48:01.600 --> 01:48:06.440 question that you know um 01:48:07.200 --> 01:48:14.200 right now probably yes. Uh but the 01:48:14.560 --> 01:48:16.637 you it's definitely possible in the 01:48:16.720 --> 01:48:19.997 future that you will be able to 01:48:20.080 --> 01:48:22.877 effectively hack into a system or 01:48:22.960 --> 01:48:27.517 somehow poison an AI system and uh have 01:48:27.600 --> 01:48:30.477 that activity be relatively untraceable 01:48:30.560 --> 01:48:34.477 because you would basically um you would 01:48:34.560 --> 01:48:39.437 you would hack into that AI system. So there's two ways you would do it. One is 01:48:39.520 --> 01:48:41.117 you poison the data that goes into that 01:48:41.200 --> 01:48:44.797 AI. So I'm not hacking into the AI 01:48:44.880 --> 01:48:46.957 itself. I'm just poisoning all the data 01:48:47.040 --> 01:48:50.397 that's feeding into that AI such that at 01:48:50.480 --> 01:48:53.517 any moment in the future I like I can 01:48:53.600 --> 01:48:56.557 activate that AI and basically hack it 01:48:56.640 --> 01:48:59.437 without any sort of active intrusion. 01:48:59.520 --> 01:49:03.277 But I can just do it because I've poisoned I've like poisoned the AI that 01:49:03.360 --> 01:49:05.277 go the data that goes into the AI such 01:49:05.360 --> 01:49:09.197 that if I like you know 01:49:09.280 --> 01:49:12.317 >> say it alters the decision-m process. 01:49:12.400 --> 01:49:16.397 >> Yeah. Exactly. >> But the but the the end decision maker 01:49:16.480 --> 01:49:19.037 which would be a human would not realize that. 01:49:19.120 --> 01:49:23.037 >> Yeah. Exactly. >> Okay. >> So so data poisoning is going to is but 01:49:23.120 --> 01:49:25.357 this is what's so terrifying about deepseek. One of the reasons why 01:49:25.440 --> 01:49:29.597 deepseek is really scary is um 01:49:29.680 --> 01:49:31.677 uh you know China chose to open source 01:49:31.760 --> 01:49:35.517 the model right so there's a lot of 01:49:35.600 --> 01:49:37.197 corporates large scale corporates in the 01:49:37.280 --> 01:49:40.717 United States that have chosen to use 01:49:40.800 --> 01:49:42.797 deepseek because they're like oh it's a 01:49:42.880 --> 01:49:45.197 good model and it's a good AI and it's 01:49:45.280 --> 01:49:49.357 free why not use it um but deepseek 01:49:49.440 --> 01:49:53.117 itself as a model could already be compromised could already be poisoned in 01:49:53.200 --> 01:49:55.597 some way such that you you know, there 01:49:55.680 --> 01:49:58.877 are characteristics or behavior or ways 01:49:58.960 --> 01:50:02.397 to activate deepseek that 01:50:02.480 --> 01:50:06.877 the the CCP and the PLA know about that 01:50:06.960 --> 01:50:10.957 um that we don't. Uh so so that's why 01:50:11.040 --> 01:50:15.117 deepseek is scary and why so so the first area is just data poisoning. So 01:50:15.200 --> 01:50:17.437 basically, can you poison the data that 01:50:17.520 --> 01:50:21.037 we're using to train the AIS such that 01:50:21.120 --> 01:50:23.197 to your point, I've altered the behavior 01:50:23.280 --> 01:50:28.237 of your AIS in a way that you don't know about and that's going to affect that's 01:50:28.320 --> 01:50:29.517 going to have cascading effects across 01:50:29.600 --> 01:50:31.597 your whole military operation. That's 01:50:31.680 --> 01:50:36.557 one. And then the second one is um 01:50:36.640 --> 01:50:39.357 uh is basically 01:50:39.440 --> 01:50:43.757 uh you know if if you're able to do the whole 01:50:43.840 --> 01:50:45.997 operation quickly enough you basically 01:50:46.080 --> 01:50:50.717 hack in and you uh kind of as we were 01:50:50.800 --> 01:50:54.877 talking about before you would like destroy the traces. You destroy any sort 01:50:54.960 --> 01:50:59.037 of trace that like you had hacked in and you have an agent that like hacked in 01:50:59.120 --> 01:51:00.797 like removed that trace and the evidence 01:51:00.880 --> 01:51:04.477 of you hacking in. um uh before anybody 01:51:04.560 --> 01:51:06.397 before it was alerted or notified. 01:51:06.480 --> 01:51:09.117 That's maybe a bit more extreme, but definitely the data poisoning stuff is 01:51:09.200 --> 01:51:11.277 is more concerning in the near term. 01:51:11.360 --> 01:51:14.397 >> Damn. So, how would you how would you 01:51:14.480 --> 01:51:17.197 defeat it? I mean, it's So, if if it 01:51:17.280 --> 01:51:19.597 were to be hacked and you knew it was 01:51:19.680 --> 01:51:24.397 hacked, then AI becomes completely irrelevant. Correct. 01:51:24.480 --> 01:51:25.917 >> Well, the issue is we're still going to 01:51:26.000 --> 01:51:28.797 rely on it for lots of things. So, um, 01:51:28.880 --> 01:51:31.757 >> it would it would have to come down to 01:51:31.840 --> 01:51:34.237 the human mind again and you would have 01:51:34.320 --> 01:51:37.037 to you would have to let's say it's a 01:51:37.120 --> 01:51:41.400 ship. You would have to 01:51:41.520 --> 01:51:44.877 know everything that you've done in the history so that it it doesn't detect 01:51:44.960 --> 01:51:47.517 what tactic you're going to use and do 01:51:47.600 --> 01:51:49.597 something 01:51:49.680 --> 01:51:51.917 just something that's never been seen 01:51:52.000 --> 01:51:58.237 before in order to confuse the adversar's AI. Correct. 01:51:58.320 --> 01:52:00.397 >> Yeah. >> So, you'd have to make a drastic change 01:52:00.480 --> 01:52:03.677 that you don't know know if it's actually going to work so that the AI 01:52:03.760 --> 01:52:08.237 doesn't detect, oh [__] we've seen this before. This is what it's about to do. 01:52:08.320 --> 01:52:10.557 >> Yeah. Yeah. So, so to your point, yeah, 01:52:10.640 --> 01:52:12.237 strategic surprise becomes the name of 01:52:12.320 --> 01:52:16.877 the game very quickly. Um, and and how 01:52:16.960 --> 01:52:20.797 do you create an operation such that you 01:52:20.880 --> 01:52:24.397 maximize the amount of strategic surprise against an adversarial AI? 01:52:24.480 --> 01:52:27.037 That's one. And then honestly, the 01:52:27.120 --> 01:52:31.437 second thing that's that's really critical is a lot of this will just 01:52:31.520 --> 01:52:36.157 plain up boil down to like straight up boil down to how 01:52:36.240 --> 01:52:37.997 many copies you have running and how 01:52:38.080 --> 01:52:41.437 large your data centers are and um how 01:52:41.520 --> 01:52:44.317 much industrial capacity you have to run 01:52:44.400 --> 01:52:47.677 these AIs both centrally and at the edge 01:52:47.760 --> 01:52:50.077 in all the war in all the theaters and 01:52:50.160 --> 01:52:54.157 all the the um in every in every 01:52:54.240 --> 01:52:56.717 environment. um 01:52:56.800 --> 01:52:58.877 >> how fast will it learn new technology? 01:52:58.960 --> 01:53:02.397 So, let's just take for example Seronic. 01:53:02.480 --> 01:53:07.357 They're making autonomous surface warfare vehicles or Palmer Lucky, you 01:53:07.440 --> 01:53:08.877 know, he's doing the autonomous 01:53:08.960 --> 01:53:11.037 submarines and [sighs] 01:53:11.120 --> 01:53:15.197 and [clears throat] so when when 01:53:15.280 --> 01:53:17.597 am I trying to say here? So, let's say 01:53:17.680 --> 01:53:20.717 we're at war with China. China has all 01:53:20.800 --> 01:53:23.437 the data, all the history back from 01:53:23.520 --> 01:53:25.757 whatever World War II on different 01:53:25.840 --> 01:53:28.237 capabilities that we have. And what 01:53:28.320 --> 01:53:31.597 happens when a new when something new is 01:53:31.680 --> 01:53:34.077 introduced onto the battle space like 01:53:34.160 --> 01:53:37.037 Seronics autonomous vehicles or Eperus 01:53:37.120 --> 01:53:38.637 or 01:53:38.720 --> 01:53:41.917 uh or Palmer's rockets or his 01:53:42.000 --> 01:53:45.437 submarines? How how would the 01:53:45.520 --> 01:53:50.077 how would the AI get the data set 01:53:50.160 --> 01:53:53.757 to make a decision or or not make decisions but come up with what you're 01:53:53.840 --> 01:53:56.317 talking about courses of actions 01:53:56.400 --> 01:54:00.077 consequences what it's about to do p you 01:54:00.160 --> 01:54:03.997 know probability of what's going to happen how how fast will it be able to 01:54:04.080 --> 01:54:05.917 learn 01:54:06.000 --> 01:54:07.757 when something new is introduced onto 01:54:07.840 --> 01:54:11.277 the battle space? Yeah, this is this is 01:54:11.360 --> 01:54:13.997 a great question. In general, 01:54:14.080 --> 01:54:17.277 so the so like the first time it sees a 01:54:17.360 --> 01:54:20.957 a totally new, let's say a USV or UUV or 01:54:21.040 --> 01:54:26.157 whatever it might be that that it's never seen before, um it won't be a, you 01:54:26.240 --> 01:54:29.117 know, it won't be able to predict what's going to happen like cuz, you know, it 01:54:29.200 --> 01:54:32.797 won't know how fast it's going to go. It won't know 01:54:32.880 --> 01:54:35.997 what what, you know, what um uh what 01:54:36.080 --> 01:54:38.637 munitions it has. It won't know what its 01:54:38.720 --> 01:54:43.197 range is. It won't it won't know all the key uh the key facts. Unless, by the 01:54:43.280 --> 01:54:45.437 way, they have really good intel and they already know all those things 01:54:45.520 --> 01:54:49.117 because they've hacked us. But let's assume they don't know. So the first few 01:54:49.200 --> 01:54:51.037 conflicts, it's not really going to be 01:54:51.120 --> 01:54:53.597 able to to figure out what's happening. 01:54:53.680 --> 01:54:57.197 And that that's a that's a key component 01:54:57.280 --> 01:54:58.797 of strategic surprise is always having 01:54:58.880 --> 01:55:02.877 new platforms that won't be sort of 01:55:02.960 --> 01:55:04.957 simulatable, let's say, by enemy 01:55:05.040 --> 01:55:08.157 wargaming tech. Um, so that's that's 01:55:08.240 --> 01:55:11.597 definitely part of it. Um, but at a 01:55:11.680 --> 01:55:14.797 certain point it's going to know what 01:55:14.880 --> 01:55:17.757 the hardware are capable of and it's 01:55:17.840 --> 01:55:19.677 going to be able to run the simulations 01:55:19.760 --> 01:55:23.277 to to understand how that changes the 01:55:23.360 --> 01:55:25.677 calculus. Um because ultimately right 01:55:25.760 --> 01:55:28.717 what's going to happen is 01:55:28.800 --> 01:55:33.037 and some of this stuff like you know this is this is like 01:55:33.120 --> 01:55:35.517 you know some of the stuff is dissonant because obviously if you look at what 01:55:35.600 --> 01:55:38.557 happens today in the military it looks nothing like this but let's play the 01:55:38.640 --> 01:55:41.837 play the tape forward and like see what happens in the future. Ultimately, 01:55:41.920 --> 01:55:42.957 you're going to run large scale 01:55:43.040 --> 01:55:45.837 simulations and it's going to figure 01:55:45.920 --> 01:55:51.757 out, hey, this new, you know, uh, uh, 01:55:51.840 --> 01:55:55.517 unmanned surface vehicle has this much range. It can go this quickly. It can 01:55:55.600 --> 01:55:56.957 maneuver in this way. It has this kind 01:55:57.040 --> 01:55:59.917 of munitions. Um, it has this kind of 01:56:00.000 --> 01:56:05.117 connectivity. Uh, it is vulnerable to these kinds of, you know, EW attacks, 01:56:05.200 --> 01:56:08.957 whatever they may be. Um, it can be jammed in these ways. And those will all 01:56:09.040 --> 01:56:11.917 just be parameters for the simulation um 01:56:12.000 --> 01:56:17.677 to run. So I think >> but initially it would have no 01:56:17.760 --> 01:56:20.317 recommendations. >> Initially you'd have strategic surprise. 01:56:20.400 --> 01:56:22.477 >> So opsec when it comes to weapons 01:56:22.560 --> 01:56:25.437 capabilities is still just paramount and 01:56:25.520 --> 01:56:28.637 it will I mean will it always come back 01:56:28.720 --> 01:56:31.757 to the human mind? 01:56:31.840 --> 01:56:35.037 >> Uh yeah I believe so. I believe that you 01:56:35.120 --> 01:56:37.197 know we have this concept that we talk 01:56:37.280 --> 01:56:40.237 about a lot which is human sovereignty. 01:56:40.320 --> 01:56:43.677 So um AI systems are going to get way 01:56:43.760 --> 01:56:46.157 better but how do we ensure that humans 01:56:46.240 --> 01:56:47.917 remain sovereign? How do we m ensure 01:56:48.000 --> 01:56:52.157 that humans maintain real control over 01:56:52.240 --> 01:56:55.437 what matters? So maintain control over our political systems, maintain control 01:56:55.520 --> 01:56:59.997 over our militaries, maintain control over our economic systems, you know uh 01:57:00.080 --> 01:57:02.397 our major industries, all that kind of 01:57:02.480 --> 01:57:05.917 stuff. And so um and I believe it's 01:57:06.000 --> 01:57:10.397 pretty paramount in the military. You're ne you are not going to want to take 01:57:10.480 --> 01:57:12.637 certainly just as like a as like a 01:57:12.720 --> 01:57:17.677 simplistic thing. We're not going to give AI the capabilities to unilaterally 01:57:17.760 --> 01:57:19.357 fire nuclear weapons. >> Mhm. 01:57:19.440 --> 01:57:22.077 >> Like we're never going to do that. Um 01:57:22.160 --> 01:57:25.357 and so ultimately so much of what is 01:57:25.440 --> 01:57:28.877 going to become really critical is the 01:57:28.960 --> 01:57:31.037 aggregation of information, simulations, 01:57:31.120 --> 01:57:35.437 wargaming, um planning to humans 01:57:35.520 --> 01:57:38.637 to ultimately make the proper decisions. 01:57:38.720 --> 01:57:40.637 And by the way, so much of this will 01:57:40.720 --> 01:57:44.397 will start bleeding into the diplomatic 01:57:44.480 --> 01:57:46.477 like diplomacy, diplomatic decisions 01:57:46.560 --> 01:57:48.877 that need to be made. Um, it'll bleed 01:57:48.960 --> 01:57:54.077 into like uh into economic warfare. Like 01:57:54.160 --> 01:57:55.437 it'll bleed into 01:57:55.520 --> 01:57:59.197 >> I mean this this goes all the way into 01:57:59.280 --> 01:58:01.197 I could see this going all the way into 01:58:01.280 --> 01:58:04.877 relationship building with with uh in 01:58:04.960 --> 01:58:07.597 between nations. Should we, you know, 01:58:07.680 --> 01:58:09.997 what are the what are the outcomes if we 01:58:10.080 --> 01:58:13.437 become allies with 01:58:13.520 --> 01:58:17.597 Russia? Yeah. >> You know what what what are the courses 01:58:17.680 --> 01:58:20.157 of action? What are the consequences? I 01:58:20.240 --> 01:58:22.717 mean, is it does it so it bleeds into 01:58:22.800 --> 01:58:25.757 everything? Politics, 01:58:25.840 --> 01:58:30.557 allies, adversaries, warfare, economics, 01:58:30.640 --> 01:58:33.517 all of it. >> Yeah. Totally. Cuz if you ultimately 01:58:33.600 --> 01:58:35.117 boil it down, what is the capability? 01:58:35.200 --> 01:58:40.077 The capability is sensing and situational awareness. So, 01:58:40.160 --> 01:58:41.837 I'm I'm going to know I'm going to be 01:58:41.920 --> 01:58:43.277 able to go through troves and troves of 01:58:43.360 --> 01:58:47.197 data. OSENT other forms of um uh like 01:58:47.280 --> 01:58:51.677 open source intel uh different kinds of of various intel feeds that I have and 01:58:51.760 --> 01:58:55.677 know what is the current status what's going on what is the what is the current 01:58:55.760 --> 01:58:58.397 situation it'll be able to aggregate all 01:58:58.480 --> 01:59:02.717 that data in to provide a a comprehensive view as to what those 01:59:02.800 --> 01:59:04.477 behaviors are and then it'll give you 01:59:04.560 --> 01:59:08.317 the ability to predict um and it'll give 01:59:08.400 --> 01:59:12.077 you the ability to effectively play 01:59:12.160 --> 01:59:15.917 forward you know every potential action you could take, what would happen in 01:59:16.000 --> 01:59:18.317 those scenarios with some probabilistic 01:59:18.400 --> 01:59:21.517 uh view, some some probabilities and 01:59:21.600 --> 01:59:26.477 then yeah, you're going to use that for every major decision like the the 01:59:26.560 --> 01:59:27.917 military and the government should use 01:59:28.000 --> 01:59:30.877 this for every major decision we make. 01:59:30.960 --> 01:59:33.837 We should do it for trade policies. We 01:59:33.920 --> 01:59:36.237 should do it for diplomatic relations. 01:59:36.320 --> 01:59:39.597 We should do it for um uh we should do 01:59:39.680 --> 01:59:43.677 it you know we're looking outwards but honestly we should also do it for like 01:59:43.760 --> 01:59:45.997 internal policies like you know what are 01:59:46.080 --> 01:59:48.397 our healthcare policies what are our um 01:59:48.480 --> 01:59:51.357 you know uh all that kind of stuff too 01:59:51.440 --> 01:59:55.357 but um so it will this capability of 01:59:55.440 --> 01:59:59.117 sort of um effectively 01:59:59.200 --> 02:00:01.837 all domain sensing plus planning um is 02:00:01.920 --> 02:00:04.717 going to be paramount. 02:00:04.800 --> 02:00:07.197 Do you 02:00:07.280 --> 02:00:10.077 and I have so many questions. 02:00:10.160 --> 02:00:13.277 Do you see a world where 02:00:13.360 --> 02:00:17.917 AI becomes so powerful 02:00:18.000 --> 02:00:22.600 throughout the world that it becomes obsolete 02:00:22.720 --> 02:00:25.517 and we're right back to where we we are 02:00:25.600 --> 02:00:28.717 we were I don't know 10 years ago, 20 02:00:28.800 --> 02:00:33.480 years ago where it's all human decisionm. 02:00:34.320 --> 02:00:37.357 Well, >> will it outdo itself? 02:00:37.440 --> 02:00:39.757 >> Um, 02:00:39.840 --> 02:00:44.797 a few thoughts here. I think so, 02:00:44.880 --> 02:00:49.037 one of the thing, so I think the first stage of what's going to happen is like 02:00:49.120 --> 02:00:53.117 kind of what I'm saying like human is the loop to human on the loop. We're 02:00:53.200 --> 02:00:55.917 going to right now humans do a lot of 02:00:56.000 --> 02:01:01.757 just like like brute force manpower 02:01:01.840 --> 02:01:05.837 work in all sorts of different pl you know in the economy and in warfare etc. 02:01:05.920 --> 02:01:10.557 >> Um that'll that's that's like the first level of of of major automation that's 02:01:10.640 --> 02:01:12.237 going to that's going to take place. So 02:01:12.320 --> 02:01:16.797 then it's like about, you know, your 02:01:16.880 --> 02:01:19.667 um strategic decision- making 02:01:19.750 --> 02:01:23.277 [clears throat] and your ability to um 02:01:23.360 --> 02:01:27.837 and your ability to make high judgment decisions um that considered long-term, 02:01:27.920 --> 02:01:29.117 short-term, medium-term, all that kind 02:01:29.200 --> 02:01:31.677 of stuff. Um 02:01:31.760 --> 02:01:36.077 at a certain point of of well 02:01:36.160 --> 02:01:37.917 at a certain as the AI continues to 02:01:38.000 --> 02:01:39.357 improve and improve and improve and 02:01:39.440 --> 02:01:43.357 improve um it will operate at a pace 02:01:43.440 --> 02:01:45.917 that is very very difficult for humans 02:01:46.000 --> 02:01:51.117 to keep up with. Um and in you know this 02:01:51.200 --> 02:01:53.277 will start happening in R&D first in re 02:01:53.360 --> 02:01:58.157 research and development like AI will be able to start doing lots of scientific 02:01:58.240 --> 02:02:01.997 research lots of R&D into new weapon systems lots of R&D into new mil you 02:02:02.080 --> 02:02:04.637 know military platforms etc um much 02:02:04.720 --> 02:02:09.437 faster than than humans are you know would be able to do and then humans will 02:02:09.520 --> 02:02:11.037 just check over their work and and 02:02:11.120 --> 02:02:13.597 decide and so it's going to sort of race 02:02:13.680 --> 02:02:18.077 faster and faster and faster and Um and 02:02:18.160 --> 02:02:22.077 the so then what happens I think it'll what it'll do is it'll create 02:02:22.160 --> 02:02:24.957 dramatically more weight on the few 02:02:25.040 --> 02:02:27.837 decisions that humans make. So any 02:02:27.920 --> 02:02:30.397 decision that like all the way to the 02:02:30.480 --> 02:02:34.237 extreme right is um you know the 02:02:34.320 --> 02:02:37.757 president or or you know whomever making 02:02:37.840 --> 02:02:41.997 decisions about do I let my AI collaborate with another country's AI 02:02:42.080 --> 02:02:46.957 like that will be like a decision of just like dramatic consequence. uh much 02:02:47.040 --> 02:02:48.397 higher consequence than like similar 02:02:48.480 --> 02:02:51.677 decisions today. So I think it almost to 02:02:51.760 --> 02:02:55.437 your point it like it will as it 02:02:55.520 --> 02:02:57.917 accelerates we'll end up at a place 02:02:58.000 --> 02:03:00.157 where you're right it all boils down to 02:03:00.240 --> 02:03:03.277 human decision-m but those decisions 02:03:03.360 --> 02:03:05.197 will carry 02:03:05.280 --> 02:03:07.757 like a thousand times more consequence. 02:03:07.840 --> 02:03:11.677 >> Mhm. Mhm. [cough] How do you decide who you're going to 02:03:11.760 --> 02:03:14.637 work with? I mean it's an international 02:03:14.720 --> 02:03:18.877 company. Yeah. Um, so we've had 02:03:18.960 --> 02:03:20.957 >> Who all are you working with? 02:03:21.040 --> 02:03:25.197 >> Uh, well, so first thing is we're pretty 02:03:25.280 --> 02:03:29.277 we're pretty picky about who we work with ultimately just because we have um 02:03:29.360 --> 02:03:32.717 we only have so many resources and building these systems and building 02:03:32.800 --> 02:03:35.437 these data sets like is pretty involved 02:03:35.520 --> 02:03:38.957 um as as kind of we've discussed. So you 02:03:39.040 --> 02:03:44.237 know our aim generally is how do you work with the best in every industry? 02:03:44.320 --> 02:03:46.797 you know, how do you how do you work with, you know, like kind of was 02:03:46.880 --> 02:03:49.437 mentioning, the number one bank, the number one pharma, the number one telco, 02:03:49.520 --> 02:03:53.357 number one military, etc. Um, the only 02:03:53.440 --> 02:03:55.837 addition to this that I would say we 02:03:55.920 --> 02:04:00.957 viewed as as as important is how are we 02:04:01.040 --> 02:04:04.157 um as you as we play the tape forward 02:04:04.240 --> 02:04:07.677 and everything we're just discussing. 02:04:07.760 --> 02:04:11.117 It's really important that 02:04:11.200 --> 02:04:15.437 um as much of the world runs on an 02:04:15.520 --> 02:04:19.117 American AI stack uh versus a CCP AI 02:04:19.200 --> 02:04:21.037 stack that becomes really really 02:04:21.120 --> 02:04:25.277 important. Um and uh it matters not only 02:04:25.360 --> 02:04:29.437 for ideology and you know kind of as we were 02:04:29.520 --> 02:04:31.917 talking about before like propaganda and control and all that kind of stuff but 02:04:32.000 --> 02:04:34.557 it also really matters just for like you 02:04:34.640 --> 02:04:38.477 know at a pure operational level like 02:04:38.560 --> 02:04:40.957 we're going to want to be able to have 02:04:41.040 --> 02:04:45.117 as extended of of of AI capabilities as 02:04:45.200 --> 02:04:48.520 possible. So, 02:04:49.120 --> 02:04:54.040 okay. So, the way I understand this is 02:04:56.240 --> 02:04:58.957 you're working with X country. We'll 02:04:59.040 --> 02:05:03.997 just say, we'll just say country X. 02:05:04.080 --> 02:05:08.397 You give country X the AI model to 02:05:08.480 --> 02:05:10.477 utilize for whatever they're doing. 02:05:10.560 --> 02:05:15.757 Let's just say warfare. we own, but they they have to tap into a 02:05:15.840 --> 02:05:17.357 US-based 02:05:17.440 --> 02:05:20.957 data center. Am I correct here? And so, 02:05:21.040 --> 02:05:23.757 as long as we control the data center 02:05:23.840 --> 02:05:26.397 that's feeding that AI model, we 02:05:26.480 --> 02:05:30.237 essentially own it and that and country 02:05:30.320 --> 02:05:34.077 X just has to trust that 02:05:34.160 --> 02:05:36.637 scale AI has their best interest. 02:05:36.720 --> 02:05:38.317 >> Yeah. That's like next level. 02:05:38.400 --> 02:05:40.717 >> And if they change, if they change, 02:05:40.800 --> 02:05:45.037 let's say country X now forms an 02:05:45.120 --> 02:05:48.797 alliance with China, they decide 02:05:48.880 --> 02:05:50.397 they don't want to be a part of America, 02:05:50.480 --> 02:05:53.837 then we just yank the AI or the not the 02:05:53.920 --> 02:05:57.037 AI, the the the the data that feeds that 02:05:57.120 --> 02:06:01.117 AI or manipulate that data to where it's 02:06:01.200 --> 02:06:05.517 essentially been hacked. Am >> I correct? And that's how we keep 02:06:05.600 --> 02:06:07.277 ourselves safe. 02:06:07.360 --> 02:06:10.397 >> Uh yes. And then with the addition like 02:06:10.480 --> 02:06:14.637 I think the way that um at least we think about it today and I think a lot 02:06:14.720 --> 02:06:18.877 of people think about it today is like it's okay for the data center to be 02:06:18.960 --> 02:06:23.757 located elsewhere located in the country. Um as long as it's US owned and 02:06:23.840 --> 02:06:28.717 operated um because then we still have control in 02:06:28.800 --> 02:06:33.277 you know any sort of scenario that happens. And um the only other thing I 02:06:33.360 --> 02:06:35.437 would say is we're much more focused 02:06:35.520 --> 02:06:39.677 initially on just um low stakes uses of 02:06:39.760 --> 02:06:43.117 AI. So can you use AI to help 02:06:43.200 --> 02:06:47.117 uh the education uh industry in one of these countries or can you use it to 02:06:47.200 --> 02:06:49.037 help the healthcare industry or can you 02:06:49.120 --> 02:06:53.357 use it to aid in um in like you know 02:06:53.440 --> 02:06:57.117 permitting processes or you know low I think low stakes use cases matter a lot 02:06:57.200 --> 02:06:59.997 more initially. Um but I really do think 02:07:00.080 --> 02:07:04.637 like uh you know we have this concept of geopolitical swing states. There are 02:07:04.720 --> 02:07:05.997 there are a number of countries right 02:07:06.080 --> 02:07:08.797 now in the world where whether they side 02:07:08.880 --> 02:07:13.117 with the US or China over time is going 02:07:13.200 --> 02:07:17.357 to have immense consequences for 02:07:17.440 --> 02:07:21.677 certainly what a what a potential conflict scenario looks like but also 02:07:21.760 --> 02:07:23.677 even what like the long-term cold war 02:07:23.760 --> 02:07:26.317 scenario looks like. like what happens 02:07:26.400 --> 02:07:29.597 um over time in this as you know our 02:07:29.680 --> 02:07:32.877 countries are interacting. So um so I 02:07:32.960 --> 02:07:35.597 view I view AI as like one of these key 02:07:35.680 --> 02:07:39.357 elements of diplomacy and long-term sort 02:07:39.440 --> 02:07:42.317 of uh like 02:07:42.400 --> 02:07:43.917 long-term 02:07:44.000 --> 02:07:47.117 um strategic impact in the in the 02:07:47.200 --> 02:07:49.357 international war game. [sighs] 02:07:49.440 --> 02:07:52.397 >> How would AI be implemented into our 02:07:52.480 --> 02:07:55.437 government? I mean, I can't remember 02:07:55.520 --> 02:07:57.917 exactly what you said, implemented to 02:07:58.000 --> 02:08:00.637 run, you know, our political sphere. 02:08:00.720 --> 02:08:04.397 What does that look like? >> Yeah. So, um, uh, 02:08:04.480 --> 02:08:07.437 >> because so much of that is people's 02:08:07.520 --> 02:08:10.477 values and and what people believe in 02:08:10.560 --> 02:08:14.797 and stand for and, you know, I mean, it 02:08:14.880 --> 02:08:18.397 like today for example, I mean, country 02:08:18.480 --> 02:08:20.157 is probably more polarized than it's 02:08:20.240 --> 02:08:23.517 ever been. And so, how do you how do you 02:08:23.600 --> 02:08:26.877 get an AI model to run government when 02:08:26.960 --> 02:08:29.357 it is this polarized and there's so many 02:08:29.440 --> 02:08:32.477 different ideologies and part of the 02:08:32.560 --> 02:08:33.917 country is way over here, the other 02:08:34.000 --> 02:08:36.797 part's way over here. How how would an 02:08:36.880 --> 02:08:40.237 AI model 02:08:40.320 --> 02:08:41.677 run that? 02:08:41.760 --> 02:08:47.197 >> Yeah. So, um the we have this concept of 02:08:47.280 --> 02:08:49.117 kind of like agentic warfare, agentic 02:08:49.200 --> 02:08:52.637 government. So can you can just like the 02:08:52.720 --> 02:08:54.317 same thing can you take these very 02:08:54.400 --> 02:08:57.437 inefficient processes in government and 02:08:57.520 --> 02:09:01.917 start replacing those with AI related functions so that you're you're just 02:09:02.000 --> 02:09:04.397 you're just improving efficiency and improving outcomes. 02:09:04.480 --> 02:09:07.517 >> Give me give me uh a specific example. 02:09:07.600 --> 02:09:09.917 >> Yeah. So what like one super simple one 02:09:10.000 --> 02:09:12.877 right now I think the the average time 02:09:12.960 --> 02:09:18.797 it takes for a veteran to see a doctor in in the VA is something like 22 days. 02:09:18.880 --> 02:09:21.517 Mhm. >> Um it's way too long and part of that is 02:09:21.600 --> 02:09:24.957 because of a host of antiquated 02:09:25.040 --> 02:09:30.637 processes and workflows and you know um just in general that system's not 02:09:30.720 --> 02:09:33.117 working. I think we can all look at that and say that's not a that's not a 02:09:33.200 --> 02:09:36.317 functional system. And so um can you use 02:09:36.400 --> 02:09:40.237 AI to you know AI agents to automate 02:09:40.320 --> 02:09:43.597 some parts of that process automatically get whatever approvals need to be gotten 02:09:43.680 --> 02:09:47.117 get whatever information needs to be gotten such that that that 22 days 02:09:47.200 --> 02:09:49.437 becomes a day or two or something like 02:09:49.520 --> 02:09:51.837 that. Um that I think is like a 02:09:51.920 --> 02:09:55.117 no-brainer just pure win for government 02:09:55.200 --> 02:09:58.877 efficiency overall. Um another one that 02:09:58.960 --> 02:10:04.077 uh other ones that are like big are like you know permitting processes. So if I 02:10:04.160 --> 02:10:07.597 want to build a new data center somewhere or even I just want to like 02:10:07.680 --> 02:10:11.517 remodel my home the you know permitting processes depending where you are could 02:10:11.600 --> 02:10:15.277 take literally take years uh for all of that to go down. And part of that is 02:10:15.360 --> 02:10:18.637 like there's so many different approvals that need to happen. There's so many 02:10:18.720 --> 02:10:21.357 like there's all these like different workflows and things that need to like 02:10:21.440 --> 02:10:25.677 happen. What if instead we just codified what are the rules of the system and had 02:10:25.760 --> 02:10:29.517 an AI agent just go automatically go through that permitting process so that 02:10:29.600 --> 02:10:31.437 you could get that permit or or get the 02:10:31.520 --> 02:10:34.717 permit denied within like a day, right? 02:10:34.800 --> 02:10:38.557 Um so and just that times a million like 02:10:38.640 --> 02:10:42.717 like the uh like um like one of the 02:10:42.800 --> 02:10:44.397 things from from Doge that they found 02:10:44.480 --> 02:10:47.037 right is that you know the uh the 02:10:47.120 --> 02:10:49.837 retirements are stored in the mine Iron 02:10:49.920 --> 02:10:52.717 Mountain mine uh literally a literal 02:10:52.800 --> 02:10:56.557 like iron mine um are like the paper 02:10:56.640 --> 02:10:57.997 copies of the retirements for all the 02:10:58.080 --> 02:11:03.597 federal employees. like can we just take that which is two generations behind um 02:11:03.680 --> 02:11:07.597 in terms of tech like it's like literally pen and paper um and then use 02:11:07.680 --> 02:11:11.437 AI to go from two generations behind to two generations forward like can we just 02:11:11.520 --> 02:11:13.197 automate as much of those processes as 02:11:13.280 --> 02:11:17.277 possible so um so I see as just like you 02:11:17.360 --> 02:11:19.517 know all over the there's so much low 02:11:19.600 --> 02:11:22.557 hanging fruit in terms of just making um 02:11:22.640 --> 02:11:25.197 current government services and 02:11:25.280 --> 02:11:27.117 government processes way more efficient 02:11:27.200 --> 02:11:30.477 I think that I I haven't met anybody who 02:11:30.560 --> 02:11:32.877 doesn't think this is the case. 02:11:32.960 --> 02:11:35.117 >> So, um that's just that's just all the 02:11:35.200 --> 02:11:41.677 level one stuff. Um I think uh the 02:11:41.760 --> 02:11:44.637 Yeah, that's just all the level one stuff in improving how our government operates. 02:11:44.720 --> 02:11:49.400 >> Would it would it eventually replace politicians? 02:11:50.240 --> 02:11:51.837 >> That's a good question. I think 02:11:51.920 --> 02:11:55.277 ultimately like we 02:11:55.360 --> 02:11:59.357 um so 02:11:59.440 --> 02:12:02.877 first off just like um uh taking a step 02:12:02.960 --> 02:12:05.997 back it's definitely the case that 02:12:06.080 --> 02:12:10.237 policy the speed of policym and the speed of legislation and the speed at 02:12:10.320 --> 02:12:13.277 which the government reacts to new technologies like that's going to have 02:12:13.360 --> 02:12:17.117 to speed up. Um, you know, we uh I've 02:12:17.200 --> 02:12:19.837 spent a lot of time in DC trying to make 02:12:19.920 --> 02:12:22.877 sure that, you know, as a country we get 02:12:22.960 --> 02:12:25.997 the right kind of AI legislation and the 02:12:26.080 --> 02:12:27.997 right kind of AI regulation to ensure 02:12:28.080 --> 02:12:31.837 that this all goes well for us. Um, 02:12:31.920 --> 02:12:36.637 it's been years of trying to get that done. You know, we still haven't really 02:12:36.720 --> 02:12:39.837 figured that out as a country. What is what is the right AI regulatory 02:12:39.920 --> 02:12:43.357 framework? Like that's still it's still undecided. I mean, how do you even 02:12:43.440 --> 02:12:45.757 describe this stuff to the dinosaurs 02:12:45.840 --> 02:12:49.597 that are still sitting in DC? I mean, 02:12:49.680 --> 02:12:52.557 we've got people stroking out on camera. 02:12:52.640 --> 02:12:56.797 We've got people literally dying in office. I mean, we got people up there 02:12:56.880 --> 02:12:58.157 that probably can't even figure out how 02:12:58.240 --> 02:13:01.037 to open a [__] email. And then you 02:13:01.120 --> 02:13:06.600 come in, 28 years old, built scale AI. 02:13:08.000 --> 02:13:10.957 I I I mean I just I mean just going all 02:13:11.040 --> 02:13:13.437 the way back to when you know 02:13:13.520 --> 02:13:18.317 Zuckerberg's sitting there you know talking to Congress. It's it's I mean 02:13:18.400 --> 02:13:22.397 and I don't agree with everything he did and whatever. It doesn't matter. But I 02:13:22.480 --> 02:13:24.957 look at that and I'm like, 02:13:25.040 --> 02:13:27.917 you guys have been sitting in DC, 02:13:28.000 --> 02:13:29.437 probably don't even know how to open 02:13:29.520 --> 02:13:32.237 your own email 02:13:32.320 --> 02:13:36.237 and you're talking to a tech genius 02:13:36.320 --> 02:13:38.717 who's trying to dub this down and make 02:13:38.800 --> 02:13:42.717 you understand. I mean, I get one day 02:13:42.800 --> 02:13:46.397 with you, you know what I mean? And to try to wrap my head around this and they 02:13:46.480 --> 02:13:49.917 have 50 million other things they're dealing with. They're not up to speed on 02:13:50.000 --> 02:13:55.037 tech. I mean, how how do you even begin to >> Yeah. 02:13:55.120 --> 02:13:55.917 >> tap in? 02:13:56.000 --> 02:13:59.517 >> I mean, I think a lot of it I think um 02:13:59.600 --> 02:14:03.117 the first thing and I think this is like 02:14:03.200 --> 02:14:08.157 a lot of people in the know understand this like a lot of the minute decisions 02:14:08.240 --> 02:14:10.717 really end up being made by staffers, right? Mhm. 02:14:10.800 --> 02:14:14.477 >> And um and 02:14:14.560 --> 02:14:18.477 I think like generally speaking like staffer you have to be extremely 02:14:18.560 --> 02:14:23.037 competent as a staffer um no matter what like there's just it's a very chaotic 02:14:23.120 --> 02:14:26.077 job. There's a lot that's there's a lot that's going on and they have to make 02:14:26.160 --> 02:14:29.117 very fast decisions. Um 02:14:29.200 --> 02:14:31.037 uh the other thing is I think I think 02:14:31.120 --> 02:14:36.157 analogies are are pretty helpful. Like I think you know everybody alive today has 02:14:36.240 --> 02:14:37.917 seen the pace of technology progress 02:14:38.000 --> 02:14:40.157 just increase and increase and increase 02:14:40.240 --> 02:14:43.277 and increase. Like I think that you know 02:14:43.360 --> 02:14:45.037 you'd be hardressed to find anyone who 02:14:45.120 --> 02:14:49.037 doesn't believe that AI will be this 02:14:49.120 --> 02:14:52.237 worldchanging technology. Now exactly how it'll change the world I think 02:14:52.320 --> 02:14:54.317 that's where it gets fuzzier but um but 02:14:54.400 --> 02:14:58.477 it will be worldchanging technology. Um, 02:14:58.560 --> 02:15:02.797 but I but the issue is like I mean the 02:15:02.880 --> 02:15:06.157 the political system just doesn't respond very quickly, right? And and 02:15:06.240 --> 02:15:08.877 that's that's gonna that's going to be 02:15:08.960 --> 02:15:13.357 very harmful. I mean we need to be able to respond very quickly to these new 02:15:13.440 --> 02:15:17.677 technologies. >> Um, and so 02:15:17.760 --> 02:15:19.757 uh I and I think they'll become more and 02:15:19.840 --> 02:15:24.797 more obvious. Like I think I think as AI and other technologies accelerate, it'll 02:15:24.880 --> 02:15:26.557 be very obvious that like the world will 02:15:26.640 --> 02:15:31.917 just change so quickly and frankly I think voters are going to demand faster 02:15:32.000 --> 02:15:34.717 action. >> Um and I so I think I think our 02:15:34.800 --> 02:15:38.877 government is set up to um to accelerate 02:15:38.960 --> 02:15:43.357 but um >> but that's that's what needs to happen. 02:15:43.440 --> 02:15:45.357 >> How do we power all this? I mean that's 02:15:45.440 --> 02:15:49.197 that's a big discussion you know and 02:15:49.280 --> 02:15:52.477 everybody is seems so apprehensive to go 02:15:52.560 --> 02:15:55.037 nuclear 02:15:55.120 --> 02:15:57.917 the grid is 02:15:58.000 --> 02:16:01.117 extremely outdated. I mean we just saw 02:16:01.200 --> 02:16:06.077 the light flickers here about I don't know 30 minutes ago. Power outages 02:16:06.160 --> 02:16:07.517 happening all the time. There was just a 02:16:07.600 --> 02:16:09.677 big one all of Spain. 02:16:09.760 --> 02:16:14.407 >> Uh Portugal, Italy. I mean, 02:16:14.490 --> 02:16:17.037 [clears throat] it's happening all the time in the US. 02:16:17.120 --> 02:16:21.117 Power outages. How are we going to be able to power all 02:16:21.200 --> 02:16:24.397 this stuff? I What would you like to see happen? 02:16:24.480 --> 02:16:26.957 >> Yeah. I mean, first of all, if you look 02:16:27.040 --> 02:16:30.317 at if you take a graph of 02:16:30.400 --> 02:16:32.957 Chinese total China's total power 02:16:33.040 --> 02:16:36.557 capacity over the past 20 years versus 02:16:36.640 --> 02:16:38.957 US total power capacity over the past 20 02:16:39.040 --> 02:16:42.637 years, the China graph is like straight 02:16:42.720 --> 02:16:46.237 up and to the right. They're just adding crazy amounts of power. They've doubled 02:16:46.320 --> 02:16:48.797 it in the last decade, I think. Double >> doubled 02:16:48.880 --> 02:16:50.077 >> doubled their power capacity in the last 02:16:50.160 --> 02:16:54.877 decade. Um, and uh, the United States is 02:16:54.960 --> 02:16:58.237 basically flat. Um, it's grown like a 02:16:58.320 --> 02:17:01.517 little bit. Uh, and so 02:17:01.600 --> 02:17:03.837 we're like that. That's what's happening 02:17:03.920 --> 02:17:05.517 right now. Right now, China's doubling 02:17:05.600 --> 02:17:09.357 every decade or so. Um, US is is is 02:17:09.440 --> 02:17:12.957 basically flat. And we're looking at, 02:17:13.040 --> 02:17:16.397 you know, the for to just power the data 02:17:16.480 --> 02:17:19.117 centers that that today AI companies 02:17:19.200 --> 02:17:20.797 know they want to build. we're going to 02:17:20.880 --> 02:17:23.597 need something like a doubling of our 02:17:23.680 --> 02:17:26.237 energy capacity and that needs to happen 02:17:26.320 --> 02:17:28.557 very very quickly like almost you know 02:17:28.640 --> 02:17:30.637 that has to happen almost immediately. 02:17:30.720 --> 02:17:34.157 Um and so you have to believe that our 02:17:34.240 --> 02:17:35.837 graph is going to go from totally flat 02:17:35.920 --> 02:17:38.877 to vertical faster vertical than than 02:17:38.960 --> 02:17:44.077 China's energy um uh growth. And China 02:17:44.160 --> 02:17:47.917 on the in the meantime is just is growing is is growing perfectly quickly. 02:17:48.000 --> 02:17:51.757 They'll accelerate. they'll add more power to their grid. Like I think it's 02:17:51.840 --> 02:17:54.877 very hard to imagine realistic scenarios 02:17:54.960 --> 02:17:58.237 where without drastic action, the United 02:17:58.320 --> 02:18:00.557 States is able to grow its energy 02:18:00.640 --> 02:18:03.197 capacity faster than China. Um, 02:18:03.280 --> 02:18:05.917 >> now where are we on the So if China's 02:18:06.000 --> 02:18:10.317 going straight up and we're flatlined, I mean, does that mean are you saying that 02:18:10.400 --> 02:18:14.317 China has surpassed our power capabilities or are we still above them 02:18:14.400 --> 02:18:17.277 even though they're on the rise? Uh they're they're definitely above us 02:18:17.360 --> 02:18:18.317 because they have a bigger population 02:18:18.400 --> 02:18:21.517 and they have way more industrials. Uh 02:18:21.600 --> 02:18:25.357 so they have uh I'll double check. I 02:18:25.440 --> 02:18:26.797 they definitely have more power total 02:18:26.880 --> 02:18:28.957 than us. Um more power generation 02:18:29.040 --> 02:18:33.117 capabilities. And um and by the way, 02:18:33.200 --> 02:18:36.957 like it's actually not rocket science why that is. It's if you look at if you 02:18:37.040 --> 02:18:41.997 then break that down to sources of that power in China, it's because coal is 02:18:42.080 --> 02:18:45.277 like 80% of that. >> Yeah. They're all They're all coal. 02:18:45.360 --> 02:18:47.197 Yeah, correct. It's just tons of coal. 02:18:47.280 --> 02:18:52.317 And then we've actually like if you look in the US, renewables have grown a lot, 02:18:52.400 --> 02:18:55.517 but a lot of it, the reason the overall number is flat is because we're using 02:18:55.600 --> 02:18:58.637 renewables to replace coal, natural gas, 02:18:58.720 --> 02:19:01.997 like fossil fuels. Um, and so when you 02:19:02.080 --> 02:19:07.197 net it out in the US, we're flat. And then in China, it's it's straight up. So 02:19:07.280 --> 02:19:10.637 that's the first thing like we need we need drastic action. You know, the 02:19:10.720 --> 02:19:14.157 administration has the National Energy Dominance Council. Um, we've sat down 02:19:14.240 --> 02:19:18.077 with them a few times like that we we got have to we have to take drastic 02:19:18.160 --> 02:19:20.877 action to enable us to to at least start 02:19:20.960 --> 02:19:25.197 matching their speed of adding energy to the grid and ideally surpass it. That's 02:19:25.280 --> 02:19:27.837 like that's the first thing. The second 02:19:27.920 --> 02:19:30.557 thing like you're talking about is our 02:19:30.640 --> 02:19:33.037 grid is extremely antiquated and that's 02:19:33.120 --> 02:19:36.717 a major strategic risk. Um, you know, I 02:19:36.800 --> 02:19:39.117 don't know what this what the cause or 02:19:39.200 --> 02:19:42.237 the source of the the um outage across 02:19:42.320 --> 02:19:46.797 Spain was, but you know, some people think it was a foreign actor um or some 02:19:46.880 --> 02:19:48.797 kind of some kind of cyber attack of 02:19:48.880 --> 02:19:51.197 some sort. Um, 02:19:51.280 --> 02:19:53.917 I guarantee you the US energy grid is 02:19:54.000 --> 02:19:56.397 extremely susceptible to large scale 02:19:56.480 --> 02:20:00.557 cyber attacks. Um it would be uh you 02:20:00.640 --> 02:20:02.237 know and the way that you know the the 02:20:02.320 --> 02:20:03.757 sophistication of these cyber attacks 02:20:03.840 --> 02:20:07.997 sometimes is like so stupid. It's like 02:20:08.080 --> 02:20:12.077 if you find the right like uh like power 02:20:12.160 --> 02:20:15.997 plant login terminal to go into sometimes people don't change the 02:20:16.080 --> 02:20:20.397 username and password from the default which is username and password. And so 02:20:20.480 --> 02:20:22.797 you can just find like some power 02:20:22.880 --> 02:20:26.877 station in like Wyoming that still has 02:20:26.960 --> 02:20:29.837 an the username and password's username and password. You log in and you can 02:20:29.920 --> 02:20:31.997 shut down the entire uh powering the 02:20:32.080 --> 02:20:37.037 entire region. So the like the so so our 02:20:37.120 --> 02:20:38.877 grid just because of how antiquated is 02:20:38.960 --> 02:20:40.397 how decentralized it is every all of 02:20:40.480 --> 02:20:45.277 that is hyper hyper susceptible to uh to 02:20:45.360 --> 02:20:47.597 cyber attacks hyper susceptible to 02:20:47.680 --> 02:20:52.397 foreign action foreign activity and um that matters now like right now if you 02:20:52.480 --> 02:20:56.317 take the energy grid in a major city people will die so it's like it's bad 02:20:56.400 --> 02:20:59.837 now but then let's go back to what we were just talking about with AI like 02:20:59.920 --> 02:21:02.477 let's say we have large scale AI on AI 02:21:02.560 --> 02:21:05.997 warfare with China. They just take out 02:21:06.080 --> 02:21:09.917 the power grid, take out our data centers and the power fueling those data 02:21:10.000 --> 02:21:12.637 centers and then we're sitting ducks. I 02:21:12.720 --> 02:21:15.117 mean, not only that, but it's my 02:21:15.200 --> 02:21:19.837 understanding that China actually 02:21:19.920 --> 02:21:21.917 produces and manufactures a lot of the 02:21:22.000 --> 02:21:23.757 major components that go into our grid 02:21:23.840 --> 02:21:26.957 like the transformers. If we don't even, 02:21:27.040 --> 02:21:29.277 to my understanding, we don't even check 02:21:29.360 --> 02:21:32.717 those for malware, Trojan horses, [__] 02:21:32.800 --> 02:21:36.957 like that. In fact, DOE actually did an 02:21:37.040 --> 02:21:39.997 inspection on one and never and never 02:21:40.080 --> 02:21:41.837 even released the results of what they 02:21:41.920 --> 02:21:44.637 found, which probably means they found 02:21:44.720 --> 02:21:48.637 some [__] And I mean, I just I don't 02:21:48.720 --> 02:21:51.597 know 02:21:51.680 --> 02:21:54.317 how we combat that. I mean just like the 02:21:54.400 --> 02:21:56.557 like what what is where did that happen 02:21:56.640 --> 02:21:59.517 elsewhere? Like look at um salt typhoon 02:21:59.600 --> 02:22:02.557 like uh this was a recent hack that was 02:22:02.640 --> 02:22:06.317 declassified which is that Chinese 02:22:06.400 --> 02:22:08.717 malware and cyber act activity like 02:22:08.800 --> 02:22:12.957 basically had fully infiltrated our um 02:22:13.040 --> 02:22:15.197 major telecom providers. I think AT&T 02:22:15.280 --> 02:22:20.157 was like entirely uh like um entirely 02:22:20.240 --> 02:22:22.397 compromised by this hack called Salt 02:22:22.480 --> 02:22:26.717 Typhoon um from the CCP and uh and 02:22:26.800 --> 02:22:31.277 that's they did that so that they could read all the messages like all the SMS 02:22:31.360 --> 02:22:33.677 all the audio um they were able to to 02:22:33.760 --> 02:22:38.637 capture as part of that as part of an intel gathering operation. Um, but if 02:22:38.720 --> 02:22:42.557 they're able to hack into our telco, they've sure as hell, you know, they're 02:22:42.640 --> 02:22:45.757 clearly capable of hacking into our energy grid, clearly h capable of 02:22:45.840 --> 02:22:47.677 hacking to any other any of our other 02:22:47.760 --> 02:22:51.677 critical infrastructure. And um, and it 02:22:51.760 --> 02:22:56.157 just goes back to what we're talking about, like the energy grid, a if we 02:22:56.240 --> 02:23:00.237 can't produce enough power, we're hosed, and b, if the adversaries can take out 02:23:00.320 --> 02:23:03.037 our power at will, we're hosed. 02:23:03.120 --> 02:23:06.637 >> Mhm. Um, and so we have this major major 02:23:06.720 --> 02:23:09.357 vulnerability as a country on just like 02:23:09.440 --> 02:23:11.677 the cyber posture of our energy grid. I 02:23:11.760 --> 02:23:14.397 think it's like I think it's one of the 02:23:14.480 --> 02:23:18.637 the biggest like very obvious like 02:23:18.720 --> 02:23:22.797 flatout um uh like clear vulnerabilities 02:23:22.880 --> 02:23:26.477 of our overall of our entire country. Um 02:23:26.560 --> 02:23:30.477 a just like you create civil unrest. you can like take, you know, imagine you 02:23:30.560 --> 02:23:33.437 took Houston's power grid out. People 02:23:33.520 --> 02:23:38.797 would die and uh you cause like all sorts of chaos. But then if you but then 02:23:38.880 --> 02:23:43.037 you take out these data centers, you take out um military bases, you take out 02:23:43.120 --> 02:23:47.997 radar systems, um you take out, you know, you name it. You can take out 02:23:48.080 --> 02:23:52.477 >> almost any piece of homeland infrastructure and that those create 02:23:52.560 --> 02:23:56.920 huge strategic openings for your adversaries. 02:23:57.280 --> 02:23:59.597 I >> mean, what M you have to run in these 02:23:59.680 --> 02:24:01.757 circles. I mean, you're building massive 02:24:01.840 --> 02:24:06.157 data centers, correct? And so, 02:24:06.240 --> 02:24:10.317 when you go to DC and you're advocating, hey, we need more power, and you just I 02:24:10.400 --> 02:24:13.357 I didn't What's the association you met with? 02:24:13.440 --> 02:24:17.597 >> Uh the National Energy Dominance Council, I mean, what do they say? 02:24:17.680 --> 02:24:21.037 >> They totally agree. I mean, they know we have to build more power. And then it's 02:24:21.120 --> 02:24:22.957 about So, then you get to the next layer 02:24:23.040 --> 02:24:25.917 of detail. It's like, okay, how can we 02:24:26.000 --> 02:24:30.637 how do we accelerate nuclear? How do we accelerate the permitting process? Um, 02:24:30.720 --> 02:24:33.597 what are existing power generation capabilities that we turned off that we 02:24:33.680 --> 02:24:35.917 can turn back on? Um, like you go 02:24:36.000 --> 02:24:37.757 through all the natural things to do. 02:24:37.840 --> 02:24:40.637 Like it's I mean I think I think we know 02:24:40.720 --> 02:24:45.277 what to do. The question is if we can get out of our own way and if and then 02:24:45.360 --> 02:24:49.917 if our grid is so antiquated that even that vulnerability like kind of means 02:24:50.000 --> 02:24:51.837 that we can be taken out any time. 02:24:51.920 --> 02:24:53.357 >> I mean I may have made an assumption. 02:24:53.440 --> 02:24:55.277 Are you are you building data centers? 02:24:55.360 --> 02:24:58.397 We we ourselves are not building data centers. 02:24:58.480 --> 02:25:00.157 >> We partner with companies that yeah that 02:25:00.240 --> 02:25:02.557 are building you know the largest data centers in the world. 02:25:02.640 --> 02:25:07.677 >> Okay. And so I've I've also heard rumors that these major data centers are 02:25:07.760 --> 02:25:11.757 starting to just create their own power source. Is that is there any validity to 02:25:11.840 --> 02:25:14.957 that? >> Yeah. So a lot of designs these days 02:25:15.040 --> 02:25:18.717 involve can you just create a SMR a 02:25:18.800 --> 02:25:21.917 small um uh like a like a nuclear 02:25:22.000 --> 02:25:23.917 reactor per data center. Can you 02:25:24.000 --> 02:25:25.997 basically like have a nuclear reactor 02:25:26.080 --> 02:25:29.757 colllocated with the data center um to 02:25:29.840 --> 02:25:34.397 uh to power that that data center's capacity? Um which I think is a good 02:25:34.480 --> 02:25:37.837 idea. The issue is like I mean China is going to be way ahead of us on that. The 02:25:37.920 --> 02:25:39.757 largest nuclear power plant in the world 02:25:39.840 --> 02:25:45.117 is in China. So um you know we're you 02:25:45.200 --> 02:25:49.197 obviously we need to lean into nuclear that needs to happen. Obviously we need 02:25:49.280 --> 02:25:52.797 to to lean into all power generation sources. when you kind of an all the 02:25:52.880 --> 02:25:55.917 above approach to power generation. Um 02:25:56.000 --> 02:26:00.477 uh but even that doesn't get us to a posture where you're confidently 02:26:00.560 --> 02:26:01.997 exceeding China. You're just kind of 02:26:02.080 --> 02:26:05.597 catching up to where they are. And so um 02:26:05.680 --> 02:26:07.837 I mean this is a huge a huge issue. 02:26:07.920 --> 02:26:11.677 >> Yeah. Let's take a quick break. When we come back I want to I want to dive more 02:26:11.760 --> 02:26:17.400 into China's capabilities and and our capabilities. 02:26:18.320 --> 02:26:20.320 All 02:26:30.000 --> 02:26:31.437 right, Alex, we're back from the break. 02:26:31.520 --> 02:26:34.797 We're getting ready to discuss some of our capabilities versus China's 02:26:34.880 --> 02:26:36.397 capabilities. 02:26:36.480 --> 02:26:38.957 And, you know, we we we just got done 02:26:39.040 --> 02:26:42.237 kind of talking about power. 02:26:42.320 --> 02:26:46.397 Is China leading the US in any other 02:26:46.480 --> 02:26:51.037 realms when it comes to the AI race? I mean, Xi Jinping has even has said 02:26:51.120 --> 02:26:55.677 himself, you know, the the the winner of the AI race will achieve global 02:26:55.760 --> 02:26:59.517 domination. [sighs] Yeah. I think well, the first thing 02:26:59.600 --> 02:27:01.277 almost as you're mentioning to 02:27:01.360 --> 02:27:03.517 understand is China has been operating 02:27:03.600 --> 02:27:06.637 against an AI master plan since 2018. 02:27:06.720 --> 02:27:11.917 they um the CCP put out a a broad whole 02:27:12.000 --> 02:27:13.997 of government, you know, civil mil 02:27:14.080 --> 02:27:17.997 military fusion plan to win on AI. Uh 02:27:18.080 --> 02:27:19.597 like you're mentioning, Xiinping himself 02:27:19.680 --> 02:27:23.757 has been has spoken about how AI is 02:27:23.840 --> 02:27:26.477 going to define the future winners of 02:27:26.560 --> 02:27:29.277 this global competition. um in military 02:27:29.360 --> 02:27:30.797 from a military standpoint they say 02:27:30.880 --> 02:27:35.517 explicitly hey we believe that AI is a 02:27:35.600 --> 02:27:39.037 leapfrog technology which means even though our military is worse than 02:27:39.120 --> 02:27:41.197 America's military today if we 02:27:41.280 --> 02:27:44.317 overinvest in AI we we have a more AI 02:27:44.400 --> 02:27:46.317 enabled military than theirs we could 02:27:46.400 --> 02:27:50.077 leapfrog them um so there they've been 02:27:50.160 --> 02:27:53.197 super invested uh right now I think the 02:27:53.280 --> 02:27:55.117 best way to kind of paint the current 02:27:55.200 --> 02:27:58.557 situation is they are way ahead on power 02:27:58.640 --> 02:28:01.117 and power generation. They're behind on 02:28:01.200 --> 02:28:05.117 chips um but catching up on chips. Um 02:28:05.200 --> 02:28:09.437 they uh are ahead of us on data. Um 02:28:09.520 --> 02:28:12.077 China has had again since 2018 a 02:28:12.160 --> 02:28:13.677 largecale 02:28:13.760 --> 02:28:18.877 um operation to dominate on data. Um and 02:28:18.960 --> 02:28:22.557 uh today in 2023 I think there were over 02:28:22.640 --> 02:28:25.437 2 million people in China who were 02:28:25.520 --> 02:28:28.397 working as working inside data factories 02:28:28.480 --> 02:28:31.597 basically as data labelers or annotators 02:28:31.680 --> 02:28:35.757 basically creating data to fuel into those into AI systems. I think that 02:28:35.840 --> 02:28:37.677 number in the US by comparison is 02:28:37.760 --> 02:28:42.717 something like 100,000. Um uh so they're 02:28:42.800 --> 02:28:44.557 outspending [clears throat] us 12 to1 on 02:28:44.640 --> 02:28:48.077 data. um they have over seven cities, 02:28:48.160 --> 02:28:50.877 full cities in China that are dedicated 02:28:50.960 --> 02:28:54.317 data hubs um that are basically powering 02:28:54.400 --> 02:28:56.957 uh you know this this like broad 02:28:57.040 --> 02:28:59.837 approach to data dominance. Um and then 02:28:59.920 --> 02:29:03.357 on algorithms I think they have they are 02:29:03.440 --> 02:29:07.357 on par with us um because of large scale 02:29:07.440 --> 02:29:11.437 espionage. So, uh, and this is, I think, 02:29:11.520 --> 02:29:13.197 one of these open secrets in the tech 02:29:13.280 --> 02:29:16.557 industry that Chinese intelligence 02:29:16.640 --> 02:29:20.397 basically steals all of the IP and 02:29:20.480 --> 02:29:23.117 technological secrets from um from the 02:29:23.200 --> 02:29:26.317 United States. Uh, there are a bunch of 02:29:26.400 --> 02:29:29.037 very concerning reports here. So one is 02:29:29.120 --> 02:29:33.357 there was um a Google engineer who took 02:29:33.440 --> 02:29:37.597 the designs and and all the IP of how 02:29:37.680 --> 02:29:41.117 Google designed their AI chips and just 02:29:41.200 --> 02:29:43.677 took those and and moved to China and 02:29:43.760 --> 02:29:46.237 then started a company um on top using 02:29:46.320 --> 02:29:49.117 those using those designs. Um, the way 02:29:49.200 --> 02:29:51.517 he got those designs, by the way, it was 02:29:51.600 --> 02:29:55.117 this guy um Leon Leon Ding, I think. Um, 02:29:55.200 --> 02:29:58.797 the way he stole the data out of out of 02:29:58.880 --> 02:30:00.957 um Google's corporate cloud, by the way, 02:30:01.040 --> 02:30:03.837 was that he it was so stupid. He just 02:30:03.920 --> 02:30:06.557 took all the code, he copy pasted it 02:30:06.640 --> 02:30:09.437 into Apple notes into like the notes app 02:30:09.520 --> 02:30:15.197 and then um exported to a PDF and printed it and just walked walked out with it. 02:30:15.280 --> 02:30:19.917 >> That's it. >> That's it. Um, so, uh, that was this was 02:30:20.000 --> 02:30:23.437 later discovered. You know, we we found out this happened, but for months we 02:30:23.520 --> 02:30:28.077 had, you know, we had no idea that they had stolen all this critical IP. Um, 02:30:28.160 --> 02:30:30.077 Stanford University, this just came out 02:30:30.160 --> 02:30:33.197 last week. Stanford University is uh is 02:30:33.280 --> 02:30:35.997 entirely infiltrated by CCP operatives. 02:30:36.080 --> 02:30:40.477 Um, few crazy facts. So first uh uh it 02:30:40.560 --> 02:30:44.237 is by law in China any Chinese citizen 02:30:44.320 --> 02:30:49.357 must comply with Chinese with CCP intelligence gathering operations. So if 02:30:49.440 --> 02:30:53.597 you're a Chinese citizen, you're living in the United States and the um 02:30:53.680 --> 02:30:57.277 intelligence agencies in China reach out to you, you have to comply with them and 02:30:57.360 --> 02:30:59.917 so you have to give them what you're seeing, what you're what you're finding, 02:31:00.000 --> 02:31:03.597 etc. Um uh so that and there's tons of 02:31:03.680 --> 02:31:07.597 Chinese nationals, Chinese citizens in across 02:31:07.680 --> 02:31:11.277 all the major elite universities, across all the major tech companies, across all 02:31:11.360 --> 02:31:12.797 the major AI labs, like they're 02:31:12.880 --> 02:31:14.797 everywhere. The second thing that's 02:31:14.880 --> 02:31:20.157 crazy is uh you know about a sixth of um 02:31:20.240 --> 02:31:24.077 Chinese students uh so so stu like 02:31:24.160 --> 02:31:25.437 Chinese citizens who are students in 02:31:25.520 --> 02:31:28.797 America are on uh scholarships sponsored 02:31:28.880 --> 02:31:32.717 by the CCP itself and for those on these 02:31:32.800 --> 02:31:36.557 scholarships they have to report back to a handler basically what are the things 02:31:36.640 --> 02:31:40.717 they find what are the things they um they're learning otherwise their 02:31:40.800 --> 02:31:43.517 scholarships get revoked so we There's 02:31:43.600 --> 02:31:45.917 there's an incredibly large scale 02:31:46.000 --> 02:31:48.877 intelligence operation running in the 02:31:48.960 --> 02:31:51.757 against the US tech industry which is 02:31:51.840 --> 02:31:53.917 just collecting all the information and 02:31:54.000 --> 02:31:56.557 secrets and technological secrets from 02:31:56.640 --> 02:31:59.277 our greatest research institutions, our 02:31:59.360 --> 02:32:01.597 universities, our lab AI labs, our tech 02:32:01.680 --> 02:32:05.517 companies um at at massive scale. And 02:32:05.600 --> 02:32:08.797 honestly, I think this is a very underrated element of how China caught 02:32:08.880 --> 02:32:11.837 up so quickly. So, um, you know, 02:32:11.920 --> 02:32:16.477 Deepseek came out of nowhere. Everyone was so surprised at how capable their 02:32:16.560 --> 02:32:20.717 model was and how they learned all these tricks. You know, how much of that is 02:32:20.800 --> 02:32:25.357 because they came up with all them on their own or they managed to have a like 02:32:25.440 --> 02:32:28.957 exquisite high-end espionage operation to steal all of our trade secrets from 02:32:29.040 --> 02:32:33.647 the United States and then reimplement them back in China? 02:32:33.730 --> 02:32:34.637 [sighs] 02:32:34.720 --> 02:32:37.277 >> What does our espionage look like? 02:32:37.360 --> 02:32:41.357 Well, there was a uh I think nowhere 02:32:41.440 --> 02:32:43.677 close to as good. I mean, I think um so 02:32:43.760 --> 02:32:47.677 one thing that uh that that the CCP did 02:32:47.760 --> 02:32:51.997 for Deepseek, the Deepseek Lab is um 02:32:52.080 --> 02:32:55.837 after Deepseek blew up and uh and um the 02:32:55.920 --> 02:32:57.677 CEO of DeepSeek met with the Chinese 02:32:57.760 --> 02:33:00.717 premier uh they then locked up all the 02:33:00.800 --> 02:33:05.837 researchers into a um uh inside I 02:33:05.920 --> 02:33:08.637 shouldn't say locked up, but they like huddled all the researchers together and 02:33:08.720 --> 02:33:12.557 they took all their passports. So none of the AI researchers who work at 02:33:12.640 --> 02:33:17.277 deepseek are are able to leave the country at all and they can't they don't 02:33:17.360 --> 02:33:18.957 come into contact with any foreigners. 02:33:19.040 --> 02:33:22.157 So they basically lock down the entire 02:33:22.240 --> 02:33:26.477 you know research effort. Um so that it you know that makes it very very hard to 02:33:26.560 --> 02:33:31.117 to conduct any sort of espionage into the into that operation. Um and then 02:33:31.200 --> 02:33:33.197 there's that report uh this is all in 02:33:33.280 --> 02:33:36.077 the news but like you know a decade ago 02:33:36.160 --> 02:33:40.477 15 years ago um all of or many of the 02:33:40.560 --> 02:33:43.037 CIA operatives US CIA operatives in 02:33:43.120 --> 02:33:47.677 China um were all killed because they 02:33:47.760 --> 02:33:51.517 were sort of um compromised because one of the communication channels they were 02:33:51.600 --> 02:33:55.917 using was compromised by Chinese intelligence and you know the CCP was 02:33:56.000 --> 02:33:57.837 able to to effectively like round a lot 02:33:57.920 --> 02:34:00.797 of them up and kill them. So our 02:34:00.880 --> 02:34:03.117 comparable their espionage in us is like 02:34:03.200 --> 02:34:07.437 extremely deep you know ex huge risk 02:34:07.520 --> 02:34:09.917 there's incredible amounts of of you 02:34:10.000 --> 02:34:13.277 know we're deeply deeply penetrated by 02:34:13.360 --> 02:34:17.597 uh by Chinese intel um and comparatively 02:34:17.680 --> 02:34:22.477 as far as I know we have like you know much less capability and I think they've 02:34:22.560 --> 02:34:28.237 designed it such that it's very hard to infiltrate their AI efforts. Um 02:34:28.320 --> 02:34:31.837 >> jeez. >> So that's how other so they're they're you know 02:34:31.920 --> 02:34:33.757 >> they're they're ahead of us on data. 02:34:33.840 --> 02:34:36.317 They're they're able to catch up through 02:34:36.400 --> 02:34:38.797 espionage on algorithms pretty easily. 02:34:38.880 --> 02:34:42.317 Um uh they're ahead of us on power. So 02:34:42.400 --> 02:34:46.077 what are we aheaded? Well, right now we're ahead in chips. Um, and that's 02:34:46.160 --> 02:34:48.317 kind of our saving grace is that uh the 02:34:48.400 --> 02:34:51.677 Nvidia chips and the entire stack there 02:34:51.760 --> 02:34:54.477 are the pride of the world and you know 02:34:54.560 --> 02:34:56.557 we're the most advanced on these chips. 02:34:56.640 --> 02:35:00.397 Chinese chips are also catching up. Um there's like a bunch of recent reports 02:35:00.480 --> 02:35:04.557 that Huawei chips are are getting to be they're basically like one generation 02:35:04.640 --> 02:35:07.197 behind the Nvidia chips. >> So they're close. 02:35:07.280 --> 02:35:11.197 >> They're close. Um so all of this is uh 02:35:11.280 --> 02:35:13.437 is pretty concerning. There was another 02:35:13.520 --> 02:35:17.037 um report that came out of CSIS recently 02:35:17.120 --> 02:35:20.237 that there was a um a Chinese effort 02:35:20.320 --> 02:35:25.517 called it's like the next generation uh brain understanding project or 02:35:25.600 --> 02:35:27.197 something where they're basically trying 02:35:27.280 --> 02:35:33.757 to use AI to fully um understand human 02:35:33.840 --> 02:35:36.317 human personality effectively and human 02:35:36.400 --> 02:35:40.317 psych psychological behaviors. Um I 02:35:40.400 --> 02:35:44.877 imagine that's ultimately for effectively like information warfare. Um 02:35:44.960 --> 02:35:49.757 uh as we were talking about at breakfast like I mean China has largecale 02:35:49.840 --> 02:35:53.117 information operations large scale information warfare and has been has 02:35:53.200 --> 02:35:58.397 been doing that for decades and you know literally decades 02:35:58.480 --> 02:35:59.917 um going back all the way to like 02:36:00.000 --> 02:36:02.077 in-person operations in Hong Kong like 02:36:02.160 --> 02:36:07.357 they are so sophisticated all that and AI is going to enable them to just move 02:36:07.440 --> 02:36:10.717 much faster as well. How do we combat that? 02:36:10.800 --> 02:36:13.677 >> Uh well, I mean, I think we need our own 02:36:13.760 --> 02:36:18.397 information operations efforts. Like, I think that's pretty critical. Um that's 02:36:18.480 --> 02:36:20.157 that's specifically on that thread. And 02:36:20.240 --> 02:36:22.877 then I think we we need to 02:36:22.960 --> 02:36:27.837 uh we need to acknowledge that at the 02:36:27.920 --> 02:36:32.717 end of the day um you know we are a more innovative country but we have to 02:36:32.800 --> 02:36:34.557 dramatically 02:36:34.640 --> 02:36:39.837 you know get our [__] together if we want to win longterm in AI. We need to 02:36:39.920 --> 02:36:43.197 um >> we need to onore ship manufactur chip 02:36:43.280 --> 02:36:47.437 manufacturing like we need to be manufacturing huge numbers of chips. We 02:36:47.520 --> 02:36:50.557 can't be dependent on Taiwan to manufacture our high-end chips. 02:36:50.640 --> 02:36:53.117 >> Are we doing that yet? At any capacity? 02:36:53.200 --> 02:36:55.597 >> Uh extremely small capacity. Like there 02:36:55.680 --> 02:37:00.317 there are a few fabs in Arizona that can produce some chips. Um but the vast 02:37:00.400 --> 02:37:01.437 majority of the volume still comes out 02:37:01.520 --> 02:37:05.197 of Taiwan. Uh we need to tighten up 02:37:05.280 --> 02:37:08.957 security in our in our AI companies 02:37:09.040 --> 02:37:11.117 dramatically. like we need to we need to 02:37:11.200 --> 02:37:14.397 have proper counter intel on you know 02:37:14.480 --> 02:37:18.477 what is the espionage risk in within these companies. Um we need to solve the 02:37:18.560 --> 02:37:20.957 power problem that we talked about. Um 02:37:21.040 --> 02:37:25.837 we need to have uh we need to be investing into you know the cyber 02:37:25.920 --> 02:37:30.557 threats like investing into large scale cyber defense. Um we need to invest into 02:37:30.640 --> 02:37:34.557 data. we need our own programs around data dominance to ensure that you know 02:37:34.640 --> 02:37:37.197 China doesn't just run away with uh with 02:37:37.280 --> 02:37:39.197 higher quality and greater AI data sets 02:37:39.280 --> 02:37:42.157 than us. So you can go through each of 02:37:42.240 --> 02:37:47.917 the elements and build like the proper plan for the United States to win. Um 02:37:48.000 --> 02:37:48.957 but uh 02:37:49.040 --> 02:37:52.477 >> have we started any of that? Um I mean I 02:37:52.560 --> 02:37:56.957 think some things are underway but uh 02:37:57.040 --> 02:37:59.997 not I mean not enough nowhere close to 02:38:00.080 --> 02:38:02.397 enough for for to to be sure that the US 02:38:02.480 --> 02:38:07.197 will win. Definitely not. And they also have a fundamental advantage. You know, 02:38:07.280 --> 02:38:11.437 uh, one of the things that that people say a lot now is like, oh, like what we 02:38:11.520 --> 02:38:14.957 need in the United States is an AI Manhattan project where we like, you 02:38:15.040 --> 02:38:18.637 know, we collect all the brilliant minds together, we collect our resources, and 02:38:18.720 --> 02:38:21.277 we have one large um, effort in in the 02:38:21.360 --> 02:38:26.157 United States. Well, it turns out like it's actually really hard to pull that 02:38:26.240 --> 02:38:30.077 off in the United States, but China can pull it off super easily. China can just 02:38:30.160 --> 02:38:34.557 say, "Hey, all the best AI people, you now work in one company. You we're going 02:38:34.640 --> 02:38:36.637 to pull together all of your resources. 02:38:36.720 --> 02:38:38.317 You are you all are we're going to put 02:38:38.400 --> 02:38:43.677 you right next to the largest nuclear power plant in the in the world. Like 02:38:43.760 --> 02:38:47.037 we're going to build the largest data center in the world here. All the chips 02:38:47.120 --> 02:38:48.717 that China has are going to go towards 02:38:48.800 --> 02:38:51.757 building this this like large scale AI 02:38:51.840 --> 02:38:54.477 project." and they just have the ability 02:38:54.560 --> 02:38:56.957 to collect all of their resources 02:38:57.040 --> 02:39:00.317 together and throw it at at winning on 02:39:00.400 --> 02:39:02.397 the AI race. Um whereas in the United 02:39:02.480 --> 02:39:05.517 States, we have all these companies. Um 02:39:05.600 --> 02:39:08.077 and you know, the United States 02:39:08.160 --> 02:39:12.557 government as of yet like it's not going to force all these companies to combine 02:39:12.640 --> 02:39:16.317 and merge. Like that's that's like such an that today would be viewed as such an 02:39:16.400 --> 02:39:19.277 overreach of government power. Um, but 02:39:19.360 --> 02:39:24.077 because of that, we're going to have like, you know, five fragmented AI 02:39:24.160 --> 02:39:27.437 efforts. And maybe in aggregate we'll have way more chips and in aggregate 02:39:27.520 --> 02:39:31.037 we'll have more power and in aggregate we'll have, you know, more great 02:39:31.120 --> 02:39:34.157 researchers, but we're not going to be able to focus those efforts whereas 02:39:34.240 --> 02:39:37.037 China is easily going to be able to focus all their efforts. 02:39:37.120 --> 02:39:42.157 >> Wow. You had mentioned something downstairs about uh nuclear weapons. 02:39:42.240 --> 02:39:43.917 >> Yeah, >> I believe. 02:39:44.000 --> 02:39:46.957 >> Yeah. So this is this is where stuff 02:39:47.040 --> 02:39:51.357 gets um stuff gets really weird for for 02:39:51.440 --> 02:39:56.877 um national security which is uh you 02:39:56.960 --> 02:39:59.757 you could you could clearly imagine 02:39:59.840 --> 02:40:02.157 scenarios where 02:40:02.240 --> 02:40:06.797 advanced very advanced cyber AI um 02:40:06.880 --> 02:40:09.997 invalidates nuclear deterrence. What do 02:40:10.080 --> 02:40:13.037 I mean by this? right now um you know 02:40:13.120 --> 02:40:16.157 nobody fires nukes because we have MAD 02:40:16.240 --> 02:40:20.957 we have mutually assured destruction and if I do a first strike against another 02:40:21.040 --> 02:40:24.877 country they're going to be able to while that nuke is in the air do a 02:40:24.960 --> 02:40:29.197 second strike and um we'll both you know there'll be destruction on both sides 02:40:29.280 --> 02:40:31.837 it'll it'll be really bad. So because of 02:40:31.920 --> 02:40:34.557 this second strike capability 02:40:34.640 --> 02:40:39.837 um luckily we have a proper you know we have real deterrence. Well, what if 02:40:39.920 --> 02:40:41.517 instead 02:40:41.600 --> 02:40:45.437 um let's say uh let's say I'm, you know, 02:40:45.520 --> 02:40:48.317 the United States and I have the most 02:40:48.400 --> 02:40:51.917 advanced AI cyber hacking capabilities 02:40:52.000 --> 02:40:54.557 in the world. So, I can build AI agents 02:40:54.640 --> 02:40:59.677 that uh hack into um that can hack into any other country, can like turn off 02:40:59.760 --> 02:41:03.037 their energy grid, can disable their weapon systems, can disable everything. 02:41:03.120 --> 02:41:06.477 So, what do I do instead? I launch the 02:41:06.560 --> 02:41:09.277 first strike and I immediat and or like 02:41:09.360 --> 02:41:12.477 first I send in my my cyber AI agent 02:41:12.560 --> 02:41:16.557 capabilities. I send my cyber AI 02:41:16.640 --> 02:41:19.677 uh you know force effectively to disable 02:41:19.760 --> 02:41:23.037 all the weapon systems of 02:41:23.120 --> 02:41:26.077 of the of the enemy country. And because 02:41:26.160 --> 02:41:29.917 it's uh my I have like such so much AI 02:41:30.000 --> 02:41:33.197 capacity, I can take out all of your I can like disable all of your weapon 02:41:33.280 --> 02:41:35.117 systems. and then I send my first strike 02:41:35.200 --> 02:41:36.557 and then you don't have a second strike 02:41:36.640 --> 02:41:40.797 capability. So if that happens basically 02:41:40.880 --> 02:41:42.637 the combination 02:41:42.720 --> 02:41:46.397 of AI and nuclear be you know you you 02:41:46.480 --> 02:41:49.117 cannot deter AI plus nuclear with just 02:41:49.200 --> 02:41:53.917 nuclear. So then it forces this um that's what will force this like 02:41:54.000 --> 02:41:57.517 proliferation of AI capabilities and so even small countries are going to need 02:41:57.600 --> 02:41:59.117 to invest in lots of AI capabilities 02:41:59.200 --> 02:42:00.717 because their nuclear weapons are no 02:42:00.800 --> 02:42:04.557 longer a sufficient deterrent. Jeez. 02:42:04.640 --> 02:42:06.797 What about boweapons? 02:42:06.880 --> 02:42:09.517 >> Yeah, I this is 02:42:09.600 --> 02:42:14.637 this is the the um element that is 02:42:14.720 --> 02:42:19.437 uh really underrated right now. So, CO 02:42:19.520 --> 02:42:23.997 leaked out of a viology lab in uh Wuhan 02:42:24.080 --> 02:42:25.997 and basically shut the world down for 02:42:26.080 --> 02:42:29.277 two years. And that's like le that's 02:42:29.360 --> 02:42:33.357 like the the level one you know biorisk 02:42:33.440 --> 02:42:35.837 kind of stuff like this was relatively 02:42:35.920 --> 02:42:38.157 uh a relatively 02:42:38.240 --> 02:42:41.677 uh you know innocuous let's say um 02:42:41.760 --> 02:42:46.397 pathogen uh but it still killed you know probably at least 10 million people 02:42:46.480 --> 02:42:48.317 globally and it was still you know shut 02:42:48.400 --> 02:42:52.237 the whole world down for 2 years um well 02:42:52.320 --> 02:42:56.797 recent models new models the new AI I 02:42:56.880 --> 02:43:01.997 models are able to uh outperform 95% of 02:43:02.080 --> 02:43:04.797 MIT viologists. So the newest models 02:43:04.880 --> 02:43:07.117 from OpenI and Google are smarter than 02:43:07.200 --> 02:43:12.797 literally 95% of viologists at MIT uh 02:43:12.880 --> 02:43:15.517 based on a recent study by um the center 02:43:15.600 --> 02:43:20.477 for safety. So um so now you now whether 02:43:20.560 --> 02:43:22.557 it's right now or whether it's in a few 02:43:22.640 --> 02:43:26.797 years um it will be feasible to use AI 02:43:26.880 --> 02:43:30.077 based capabilities to help you design 02:43:30.160 --> 02:43:34.237 powerful pathogens. And what's more than that you're going to be able to design 02:43:34.320 --> 02:43:36.237 in certain characteristics of these 02:43:36.320 --> 02:43:38.077 pathogens. You know you'll be able to 02:43:38.160 --> 02:43:41.437 tune the virality tune the um lethality 02:43:41.520 --> 02:43:43.837 of them. You know, there's also due to 02:43:43.920 --> 02:43:45.197 recent advancements in synthetic 02:43:45.280 --> 02:43:49.757 biology, you now can create viruses that 02:43:49.840 --> 02:43:52.797 specifically target certain segments of 02:43:52.880 --> 02:43:57.357 DNA. So, um I could create a bioweapon 02:43:57.440 --> 02:44:02.077 that just targeted, you know, um any individual with a certain segment of 02:44:02.160 --> 02:44:06.157 DNA, which means I can target to basically like any population or any 02:44:06.240 --> 02:44:08.077 group or any subsegment of the 02:44:08.160 --> 02:44:11.757 population in the world. um which is 02:44:11.840 --> 02:44:15.757 >> uh which is really really bad. And so um 02:44:15.840 --> 02:44:21.197 the the ability so first even without AI 02:44:21.280 --> 02:44:23.117 like biology synthetic biology is making 02:44:23.200 --> 02:44:25.197 so much progress and that there's just 02:44:25.280 --> 02:44:29.037 like all sorts of inherent risk of uh 02:44:29.120 --> 02:44:31.357 like all sorts of inherent risk of 02:44:31.440 --> 02:44:34.397 bioweaponry or you know leaks of of 02:44:34.480 --> 02:44:37.117 pathogens and and viruses and whatnot. 02:44:37.200 --> 02:44:40.477 And then with AI all of a sudden you 02:44:40.560 --> 02:44:45.597 this is you know not not literally today's models but a few models a few 02:44:45.680 --> 02:44:47.117 generations down you're going to be able 02:44:47.200 --> 02:44:49.997 to use these AI systems to design or 02:44:50.080 --> 02:44:53.357 build you know uh next generation 02:44:53.440 --> 02:44:58.237 pathogens. Um so that's that's an entire 02:44:58.320 --> 02:45:01.117 I mean for good reason biological 02:45:01.200 --> 02:45:05.037 warfare is not you know one of the um is 02:45:05.120 --> 02:45:08.637 not you know there international treaties such that we don't engage in 02:45:08.720 --> 02:45:11.357 biological warfare but if you imagine 02:45:11.440 --> 02:45:16.477 these scenarios where countries you know nuclear deterrence doesn't work they 02:45:16.560 --> 02:45:19.997 don't have the resources to get to use to utilize to have large scale AI data 02:45:20.080 --> 02:45:23.437 centers um you know can you know I'm 02:45:23.520 --> 02:45:26.157 worried that countries will will 02:45:26.240 --> 02:45:28.957 turn to biological weaponry bioweapons 02:45:29.040 --> 02:45:35.117 as their deterrence mechanism which is highly destabilizing for you know the 02:45:35.200 --> 02:45:35.677 world. 02:45:35.760 --> 02:45:39.277 >> Wow, that's some scary [__] 02:45:39.360 --> 02:45:42.877 >> The flip side is there is new technology 02:45:42.960 --> 02:45:45.677 that can um 02:45:45.760 --> 02:45:48.077 that can also prevent this stuff. So 02:45:48.160 --> 02:45:51.517 there's um there's this research coming 02:45:51.600 --> 02:45:53.917 out of uh this lab in Seattle, David 02:45:54.000 --> 02:45:55.917 Baker's lab, this guy who just won a 02:45:56.000 --> 02:45:59.917 Nobel Prize on uh biological noses 02:46:00.000 --> 02:46:03.277 uh or uh digital noses, sorry. Um which 02:46:03.360 --> 02:46:06.317 is basically you have these devices that 02:46:06.400 --> 02:46:08.077 can detect 02:46:08.160 --> 02:46:11.037 chem detect proteins or chemicals or 02:46:11.120 --> 02:46:14.717 pathogens in the air um automatically. 02:46:14.800 --> 02:46:17.597 And so I think what this will like, you 02:46:17.680 --> 02:46:22.477 know, the real sort of like offense defense of of bio and bioweaponry will 02:46:22.560 --> 02:46:26.077 end up looking like we're just going to have large scale deployment of digital 02:46:26.160 --> 02:46:29.197 noses effectively that in every space on 02:46:29.280 --> 02:46:31.917 every like shipping container on every 02:46:32.000 --> 02:46:34.717 plane. um you know they're just 02:46:34.800 --> 02:46:37.757 constantly sensing for all existing 02:46:37.840 --> 02:46:40.637 known pathogens, any new pathogens um 02:46:40.720 --> 02:46:42.637 that might exist and are constantly just 02:46:42.720 --> 02:46:45.757 like you know containing effect or like 02:46:45.840 --> 02:46:47.757 detecting and ultimately containing. 02:46:47.840 --> 02:46:52.157 >> Interesting. It's sniffing real time for all of that [__] 02:46:52.240 --> 02:46:54.237 >> Yeah, exactly. I mean on also on the 02:46:54.320 --> 02:47:00.520 flip side I mean I guess if AI is developing 02:47:00.960 --> 02:47:02.877 a new bioweapon 02:47:02.960 --> 02:47:05.917 CO comes out again CO too we'll just 02:47:06.000 --> 02:47:10.717 call it then RAI 02:47:10.800 --> 02:47:14.957 should also be able to figure out the the 02:47:15.040 --> 02:47:18.397 >> vaccines or >> the vaccine the antidote to it correct 02:47:18.480 --> 02:47:20.317 >> yeah totally so there will be there will 02:47:20.400 --> 02:47:25.357 be an offense defense um element to 02:47:25.440 --> 02:47:29.037 just as just as in kind as we were walking through like AI applied to 02:47:29.120 --> 02:47:32.797 command and control there's an offense defense element AI applied to cyber 02:47:32.880 --> 02:47:37.837 there's an offense defense element AI applied to bio and bioweaponry there 02:47:37.920 --> 02:47:39.757 will be an offense defense element so 02:47:39.840 --> 02:47:42.637 all these thankfully there's like you 02:47:42.720 --> 02:47:47.197 know the hope is that we end up in a in 02:47:47.280 --> 02:47:51.357 a in a global world you know the the the 02:47:51.440 --> 02:47:55.917 world agree agrees that basically we're not going to go down any of these paths 02:47:56.000 --> 02:47:59.597 like cuz there's mutual deterrence and we just you know it's not worth it for 02:47:59.680 --> 02:48:04.397 anybody in the world to destabilize you know and risk humanity like that. 02:48:04.480 --> 02:48:06.557 >> That's that's basically where we need to land. 02:48:06.640 --> 02:48:08.637 >> Wow. How concerned are you about China 02:48:08.720 --> 02:48:10.717 Taiwan? I mean we were talking about 02:48:10.800 --> 02:48:14.077 this a little bit at breakfast and I I I 02:48:14.160 --> 02:48:18.717 can't believe they have not made a move yet. I mean, I thought for sure it would 02:48:18.800 --> 02:48:20.157 happen towards the end of the last 02:48:20.240 --> 02:48:23.517 administration, but [sighs] 02:48:23.600 --> 02:48:25.997 I mean, with their chip production 02:48:26.080 --> 02:48:27.917 capabilities, I mean, how how concerned 02:48:28.000 --> 02:48:32.157 are you about China taking Taiwan? 02:48:32.240 --> 02:48:36.637 >> I think if it's going to happen, it's going to happen this decade and it's 02:48:36.720 --> 02:48:40.637 probably going to happen this administration. And um 02:48:40.720 --> 02:48:44.317 >> why do you say that? Uh 02:48:44.400 --> 02:48:47.517 I mean China at a at a macro sense they 02:48:47.600 --> 02:48:51.997 have huge demographic issues. Those are 02:48:52.080 --> 02:48:56.397 I mean there's not like that's just like the of the force of gravity in their 02:48:56.480 --> 02:48:59.837 country. They have this huge aging population. They made the wrong bet you 02:48:59.920 --> 02:49:04.397 know many decades ago to have a one child policy. Um and so they are going 02:49:04.480 --> 02:49:06.797 to have this like huge aging population. 02:49:06.880 --> 02:49:10.077 Um over then that that plays that plays 02:49:10.160 --> 02:49:14.237 out really like quite soon like over the next like a decade from now it's going 02:49:14.320 --> 02:49:17.597 to be um over time they're going to look more and more like Japan in that way 02:49:17.680 --> 02:49:21.277 where they have this like large aging population and it'll paralyze a lot of 02:49:21.360 --> 02:49:22.797 ability to make any sort of aggressive 02:49:22.880 --> 02:49:25.517 moves. So particularly when it comes to 02:49:25.600 --> 02:49:28.397 military industrial capacity etc. Um, so 02:49:28.480 --> 02:49:31.037 that's like one force of gravity uh that 02:49:31.120 --> 02:49:34.477 they have to contend with. And then um 02:49:34.560 --> 02:49:38.717 and so I think I think they're they're they're going to want to move faster 02:49:38.800 --> 02:49:41.277 sooner rather than later. And then 02:49:41.360 --> 02:49:43.597 they've I mean they've had such an 02:49:43.680 --> 02:49:46.557 insane military buildup um over the 02:49:46.640 --> 02:49:49.517 course of the past few decades. Um, you 02:49:49.600 --> 02:49:53.037 know, I don't think it's and I think, 02:49:53.120 --> 02:49:57.277 you know, we're we're currently in a situation where China has far more 02:49:57.360 --> 02:50:01.277 industrial capacity, far more manufacturing capacity than we do in the 02:50:01.360 --> 02:50:02.317 United States. 02:50:02.400 --> 02:50:05.277 >> And so, um, 02:50:05.360 --> 02:50:09.757 that is set, you know, that's a window for them. Um, 02:50:09.840 --> 02:50:13.837 >> so do you think they'll do that they're pressed to do it because of the aging 02:50:13.920 --> 02:50:16.957 population? >> I think a lot of factors. I think I 02:50:17.040 --> 02:50:20.797 think Xi is aging right. Um this will be 02:50:20.880 --> 02:50:23.997 an important component of his legacy uh 02:50:24.080 --> 02:50:28.717 as as he would view it. I think um they have an aging population which will 02:50:28.800 --> 02:50:30.957 minimize their political latitude over 02:50:31.040 --> 02:50:34.957 time um naturally and then they have um 02:50:35.040 --> 02:50:36.877 I mean they are they're in this in 02:50:36.960 --> 02:50:39.997 insane window where they have just 02:50:40.080 --> 02:50:42.717 incredible uh industrial manufacturing 02:50:42.800 --> 02:50:45.037 capabilities um compared to anywhere 02:50:45.120 --> 02:50:47.757 else in the world. you know, in 2023, 02:50:47.840 --> 02:50:51.037 uh, China deployed more industrial 02:50:51.120 --> 02:50:56.557 robots than the rest of the world combined. Um, that's like I mean, we 02:50:56.640 --> 02:51:00.157 were talking a little bit about like automated factories and automated 02:51:00.240 --> 02:51:02.637 industrials, like they're racing to that 02:51:02.720 --> 02:51:07.677 faster than any other country in the world. And so um so I think that like 02:51:07.760 --> 02:51:09.597 you can look at all these dimensions and 02:51:09.680 --> 02:51:14.237 this window um you know there's if 02:51:14.320 --> 02:51:16.797 they're if they're going to do it they're going to do it soon. 02:51:16.880 --> 02:51:21.677 >> Yeah. Yeah. I mean what what percentage of the chips that we use come from 02:51:21.760 --> 02:51:26.557 Taiwan? >> I mean 95% 02:51:26.640 --> 02:51:29.677 of the high-end chips um are manufactured in Taiwan. 02:51:29.760 --> 02:51:33.997 >> And so what happens if if China takes Taiwan? 02:51:34.080 --> 02:51:36.317 So yeah, war game it out. So we were 02:51:36.400 --> 02:51:38.957 talking a little bit about this. So um 02:51:39.040 --> 02:51:42.317 let's say China blockades or or invades 02:51:42.400 --> 02:51:46.717 Taiwan um then then there's a qu So 02:51:46.800 --> 02:51:51.357 these fabs are incredibly incredibly valuable because 02:51:51.440 --> 02:51:55.277 >> as we were just describing if you believe in the pace of AI progress and 02:51:55.360 --> 02:51:59.437 AI technology then >> everything boils down to how much power 02:51:59.520 --> 02:52:00.957 you got, how many chips you've got. And 02:52:01.040 --> 02:52:04.237 if they own 95% of the world's 02:52:04.320 --> 02:52:08.957 ship manufacturing capability, I mean, they're going to run away with 02:52:09.040 --> 02:52:10.557 it. So then you look at that and you 02:52:10.640 --> 02:52:13.677 say, will the Taiwanese people bomb the 02:52:13.760 --> 02:52:17.277 TSMC data centers? Um, and andor will 02:52:17.360 --> 02:52:21.197 the US bomb the TSMC data centers and or will some other country bomb the the 02:52:21.280 --> 02:52:22.797 data centers? Or sorry, not the the 02:52:22.880 --> 02:52:28.317 fabs, um, the TSMC chip fabs. Um I think 02:52:28.400 --> 02:52:31.677 my personal belief I don't think uh the 02:52:31.760 --> 02:52:34.317 Taiwanese do it because even if they get 02:52:34.400 --> 02:52:36.557 blockaded or invaded that those FABs are 02:52:36.640 --> 02:52:40.877 still a huge uh component of Taiwan 02:52:40.960 --> 02:52:44.077 survivability and um Taiwan's relevance 02:52:44.160 --> 02:52:47.677 as a as a as an entity um even if they 02:52:47.760 --> 02:52:51.677 get blockade or invaded by Taiwan. So I don't think they do it. Um, China 02:52:51.760 --> 02:52:55.677 definitely doesn't do it cuz they obviously invading partially to get, you 02:52:55.760 --> 02:52:57.997 know, um, to gain those capabilities. 02:52:58.080 --> 02:53:02.237 Um, and so then, uh, does the US bomb 02:53:02.320 --> 02:53:07.117 them? If the US bombs them, that's probably World War II. I, it's hard to 02:53:07.200 --> 02:53:08.877 imagine that not just resulting in 02:53:08.960 --> 02:53:12.877 massive escalation. Um, and so you're 02:53:12.960 --> 02:53:15.597 looking at it and there's kind of no 02:53:15.680 --> 02:53:18.877 good options. Um, 02:53:18.960 --> 02:53:23.437 so I think it's I mean everyone's very 02:53:23.520 --> 02:53:25.437 focused on it obviously, but it is it is 02:53:25.520 --> 02:53:29.677 like a real powder keg of a uh 02:53:29.760 --> 02:53:33.677 >> damn >> of a region. 02:53:33.760 --> 02:53:36.477 >> How do you think this all ends? 02:53:36.560 --> 02:53:41.117 >> Um, we had a little discussion about this at 02:53:41.200 --> 02:53:41.837 breakfast. 02:53:41.920 --> 02:53:45.997 >> Yeah. Yeah. Um I mean 02:53:46.080 --> 02:53:48.557 I think 02:53:48.640 --> 02:53:52.157 I think if so let's assume that in the 02:53:52.240 --> 02:53:54.477 next handful of years next like 3 four 02:53:54.560 --> 02:53:57.597 years um there's an invasion or blockade 02:53:57.680 --> 02:53:59.357 of Taiwan 02:53:59.440 --> 02:54:01.677 and 02:54:01.760 --> 02:54:04.877 uh you know I think it's I think given 02:54:04.960 --> 02:54:07.757 how important AI is um it's hard for the 02:54:07.840 --> 02:54:09.997 US to to not take any sort of action in 02:54:10.080 --> 02:54:14.157 that scenario and then you 02:54:14.240 --> 02:54:16.237 almost all the actions you would see 02:54:16.320 --> 02:54:19.677 escalating into a major major conflict. 02:54:19.760 --> 02:54:23.800 So um 02:54:25.120 --> 02:54:28.397 best case scenario is we deter the 02:54:28.480 --> 02:54:33.197 invasion or blockade altogether and um 02:54:33.280 --> 02:54:37.400 and I think you know 02:54:37.680 --> 02:54:41.437 I think it certainly is in everyone's interest to not get into a large scale 02:54:41.520 --> 02:54:45.597 world war that's a hugely destructive and and kills lots of people. So I think 02:54:45.680 --> 02:54:47.357 like fundamentally we should be able to 02:54:47.440 --> 02:54:51.677 deter that conflict. Um, but that's uh 02:54:51.760 --> 02:54:55.197 that's why all this matters so much. We need to make sure our AI capabilities as 02:54:55.280 --> 02:54:58.477 a country are the best in the world. We need to make sure that our military AI 02:54:58.560 --> 02:55:00.157 capabilities are the best in the world. 02:55:00.240 --> 02:55:03.437 We need to make sure that you know um uh 02:55:03.520 --> 02:55:06.797 the there's clear uh economic deterrence 02:55:06.880 --> 02:55:09.437 of this kind of scenario like we need to 02:55:09.520 --> 02:55:12.157 um we need to be investing in in every 02:55:12.240 --> 02:55:16.237 way to deter this conflict such that um 02:55:16.320 --> 02:55:17.597 you know where this really will break 02:55:17.680 --> 02:55:20.157 down is if the Chinese if the CCP 02:55:20.240 --> 02:55:25.277 calculus uh uh you know diverges from 02:55:25.360 --> 02:55:28.237 our own if their calculus becomes 02:55:28.320 --> 02:55:32.717 oh no this is going to work out you we we can take this and then you know we're 02:55:32.800 --> 02:55:36.397 strong enough so they'll work out for us and then our calculus is the opposite. 02:55:36.480 --> 02:55:38.877 That's where that's where the world war 02:55:38.960 --> 02:55:44.237 scenario happens. So um so I think it's possible to deter and I think we have to 02:55:44.320 --> 02:55:46.877 you know there's a lot of things we have to do to make sure that we deter that 02:55:46.960 --> 02:55:49.677 conflict and that should be I mean 02:55:49.760 --> 02:55:53.277 certainly I think it already is like 80% 02:55:53.360 --> 02:55:55.837 of the focus of the entire DoD. So, I 02:55:55.920 --> 02:56:01.277 mean, it's it's just we can deter, but I mean, when you're talking about an aging 02:56:01.360 --> 02:56:02.717 population, I mean, they're getting 02:56:02.800 --> 02:56:06.157 desperate, and it sounds like in order 02:56:06.240 --> 02:56:08.797 for them to to legitimately win, they 02:56:08.880 --> 02:56:11.837 have to acquire those chip fabs, 02:56:11.920 --> 02:56:16.040 correct? And so, 02:56:16.240 --> 02:56:19.677 they already have 250 times the ship building capacity. They have way more 02:56:19.760 --> 02:56:21.517 people. They have they have more power 02:56:21.600 --> 02:56:25.437 than we do. I mean, 02:56:25.520 --> 02:56:27.757 military recruitment in the US, you 02:56:27.840 --> 02:56:29.437 know, was at an all-time low. I don't 02:56:29.520 --> 02:56:32.957 know what it is today, but I mean, even 02:56:33.040 --> 02:56:37.917 if it So, I guess what I'm saying is 02:56:38.000 --> 02:56:41.677 we can you can only deter a desperate 02:56:41.760 --> 02:56:43.837 entity for so long before they throw a 02:56:43.920 --> 02:56:47.037 Hail Mary play, right? Would you agree 02:56:47.120 --> 02:56:50.877 with that? >> Yeah. And then it just depends on the 02:56:50.960 --> 02:56:53.277 >> you would have to dedicate an entire 02:56:53.360 --> 02:56:57.837 military to surround Taiwan 02:56:57.920 --> 02:57:01.677 to effectively do that in my opinion. 02:57:01.760 --> 02:57:04.477 >> Yeah. I mean, I think that the if if 02:57:04.560 --> 02:57:07.357 they assess if the CCP and the PLA 02:57:07.440 --> 02:57:11.677 assess that Taiwan is all they're like they they 02:57:11.760 --> 02:57:13.277 will focus their entire military 02:57:13.360 --> 02:57:17.757 capacity on seizing Taiwan, then 02:57:17.840 --> 02:57:20.877 that becomes a really that becomes a 02:57:20.960 --> 02:57:21.997 really tricky calculus. 02:57:22.080 --> 02:57:26.157 >> I mean, why wouldn't they? If if if if 02:57:26.240 --> 02:57:28.317 Xi believes that the winner of the AI 02:57:28.400 --> 02:57:31.837 race achieves global domination, 02:57:31.920 --> 02:57:33.837 he's getting older. You just talked 02:57:33.920 --> 02:57:36.557 about how important his legacy is to 02:57:36.640 --> 02:57:41.080 him, which I'm sure you're right. 02:57:41.280 --> 02:57:43.837 I don't know how you deter that. 02:57:43.920 --> 02:57:45.917 And then they win the AR race. 02:57:46.000 --> 02:57:49.117 >> Yeah. The only thing that we can do, 02:57:49.200 --> 02:57:50.397 I think this is a long shot, but I think 02:57:50.480 --> 02:57:53.517 it's important, is if 02:57:53.600 --> 02:57:57.837 um if ultimately we actually end up 02:57:57.920 --> 02:58:00.157 collaborating on AI. And I know that 02:58:00.240 --> 02:58:03.517 sounds kind of crazy, but um 02:58:03.600 --> 02:58:07.037 but if we're able as a country to 02:58:07.120 --> 02:58:09.677 demonstrate 02:58:09.760 --> 02:58:13.437 just we're so far ahead and there's 02:58:13.520 --> 02:58:16.237 like, you know, the one one key element 02:58:16.320 --> 02:58:19.277 of of how the whole AI thing plays out 02:58:19.360 --> 02:58:22.557 um is this idea of um AI 02:58:22.640 --> 02:58:26.717 self-improvement or uh intelligence recursion sometimes people call it. 02:58:26.800 --> 02:58:31.277 Basically, um, once AIS get sufficiently good, then 02:58:31.360 --> 02:58:36.237 you can start utilizing the AIS to help you build the next AI. You, as sci-fi as 02:58:36.320 --> 02:58:40.317 that sounds, you utilize your current generation AI to build the next 02:58:40.400 --> 02:58:41.837 generation AI faster and faster and 02:58:41.920 --> 02:58:43.517 faster and faster. 02:58:43.600 --> 02:58:47.437 >> And so, at some point, um, your AI 02:58:47.520 --> 02:58:50.797 capabilities enable you like, you know, there's some form of like, you know, 02:58:50.880 --> 02:58:53.677 just exponential takeoff. They just, they just, you know, your AI 02:58:53.760 --> 02:58:55.517 capabilities get good really, really 02:58:55.600 --> 02:58:58.957 quickly. And if somebody's even 3 to 6 02:58:59.040 --> 02:59:03.517 months behind you, then they're they're never going to catch up to you because 02:59:03.600 --> 02:59:05.917 you're running the self-improvement loop >> got 02:59:06.000 --> 02:59:10.157 >> faster than anybody else. And so this is 02:59:10.240 --> 02:59:14.797 a this is a key idea. I mean, it's I think it's um it's a little bit 02:59:14.880 --> 02:59:19.197 theoretical right now. Like it's not clear whether or not this intelligence 02:59:19.280 --> 02:59:23.277 recursion um is going to be how it plays out, but but a lot of people in AI 02:59:23.360 --> 02:59:27.357 believe it. Um, and I probably I probably believe it too that that we 02:59:27.440 --> 02:59:30.877 will be able to use AIS to help us continue training the next AIS and 02:59:30.960 --> 02:59:32.957 improve things more quickly. And if you 02:59:33.040 --> 02:59:36.797 believe that, then if we're let's say 02:59:36.880 --> 02:59:41.197 three 3 to 6 months ahead of China um, 02:59:41.280 --> 02:59:43.597 and we maintain that advantage and we 02:59:43.680 --> 02:59:46.317 take off faster, then they're going to 02:59:46.400 --> 02:59:51.117 be way behind. And then ultimately we're going to be in a great position to say, 02:59:51.200 --> 02:59:52.877 "Hey, 02:59:52.960 --> 02:59:57.117 actually, like we're way ahead and we 02:59:57.200 --> 02:59:58.957 should just, you know, you guys should 02:59:59.040 --> 03:00:02.637 quit your efforts. Um we'll give you AI 03:00:02.720 --> 03:00:07.437 for all of your economic and uh and um 03:00:07.520 --> 03:00:09.677 humanitarian uses throughout your 03:00:09.760 --> 03:00:12.237 society. Uh and we agree we're not going 03:00:12.320 --> 03:00:16.157 to battle on military AI." 03:00:16.240 --> 03:00:18.317 What would it take 03:00:18.400 --> 03:00:21.357 to take the chip building capabilities 03:00:21.440 --> 03:00:23.757 that Taiwan has as implement that here 03:00:23.840 --> 03:00:27.517 in the US to protect it? 03:00:27.600 --> 03:00:29.197 So 03:00:29.280 --> 03:00:33.357 yeah um so the first thing is 03:00:33.440 --> 03:00:35.917 uh there's been hundreds of billions of 03:00:36.000 --> 03:00:38.477 dollars invested 03:00:38.560 --> 03:00:42.077 just into like the buildout of those 03:00:42.160 --> 03:00:45.757 fabs and the the the the call foundaries 03:00:45.840 --> 03:00:47.437 but the buildup of these of these large 03:00:47.520 --> 03:00:51.037 scale chip factories effectively and the 03:00:51.120 --> 03:00:53.597 um and all the high-end equipment and 03:00:53.680 --> 03:00:58.877 tooling inside of them. hundreds of billions of dollars of investment. So, 03:00:58.960 --> 03:01:01.357 first off, there needs to be hundreds of billions of dollars investment in the 03:01:01.440 --> 03:01:03.837 US. That's not the hard part. The second 03:01:03.920 --> 03:01:06.557 part that's that's really the hard part 03:01:06.640 --> 03:01:10.157 is um all it's basically a large scale 03:01:10.240 --> 03:01:13.517 factory operated by highly highly 03:01:13.600 --> 03:01:20.477 skilled um uh uh uh workers who 03:01:20.560 --> 03:01:23.437 are very experienced in those processes 03:01:23.520 --> 03:01:25.357 and the whole thing operates like a you 03:01:25.440 --> 03:01:28.637 know like clockwork. Um, and unless you 03:01:28.720 --> 03:01:33.517 can get those people to the US, you know, you're going to have to like 03:01:33.600 --> 03:01:36.877 rebuild all that knowhow and all that technical capability. 03:01:36.960 --> 03:01:40.317 >> And that's what takes a really long time. And that's one of the things, you 03:01:40.400 --> 03:01:43.117 know, >> why do you think we haven't done that? 03:01:43.200 --> 03:01:45.997 Why do you think we have not incentivized 03:01:46.080 --> 03:01:48.557 these brilliant minds to come here and 03:01:48.640 --> 03:01:52.317 do it for us? So TSMC, the Taiwan 03:01:52.400 --> 03:01:54.877 semiconductor, um the company that you 03:01:54.960 --> 03:01:58.077 know builds these fabs, they have stood 03:01:58.160 --> 03:02:02.237 up a few fabs in Arizona. Um but they 03:02:02.320 --> 03:02:06.797 cited issues like first there were issues around um permitting and getting 03:02:06.880 --> 03:02:08.237 enough power and they dealt with some 03:02:08.320 --> 03:02:12.397 EPA issues and then um and then they 03:02:12.480 --> 03:02:15.437 just have issues where the like uh you 03:02:15.520 --> 03:02:17.837 know the technicians working in Arizona 03:02:17.920 --> 03:02:20.157 don't aren't as skilled or don't work as 03:02:20.240 --> 03:02:23.597 hard as those working in Taiwan. Um 03:02:23.680 --> 03:02:26.157 so they've built a few fabs in the 03:02:26.240 --> 03:02:27.437 United States. So, they've tried to do 03:02:27.520 --> 03:02:31.917 it, but our our red tape and our our 03:02:32.000 --> 03:02:35.197 power is not what it needs to be to be 03:02:35.280 --> 03:02:38.477 able to do this red tape, power, um, 03:02:38.560 --> 03:02:42.877 workforce. And then there's another key thing, which is 03:02:42.960 --> 03:02:46.717 if you look at it from Taiwan Semiconductor, from from TSMC's 03:02:46.800 --> 03:02:48.797 perspective, they're not all that 03:02:48.880 --> 03:02:51.597 incentivized to stand up all these 03:02:51.680 --> 03:02:54.077 capabilities in the United States. like 03:02:54.160 --> 03:02:57.517 if as soon as they start standing up all 03:02:57.600 --> 03:03:01.757 these capabilities in the United States, the United States is not incentivized to 03:03:01.840 --> 03:03:03.677 defend Taiwan. >> Yeah. 03:03:03.760 --> 03:03:08.717 >> And it's a Taiwanese company. Um so and 03:03:08.800 --> 03:03:10.317 it's a critical part of their survival 03:03:10.400 --> 03:03:14.557 strategy. Um, so, so that's that's 03:03:14.640 --> 03:03:18.397 really where the rubber hits the road is are they actually incentivized to do a 03:03:18.480 --> 03:03:22.477 large scale buildout of of chip manufacturing capacity in the United 03:03:22.560 --> 03:03:26.157 States? Um, I think the answer is like no. 03:03:26.240 --> 03:03:28.717 >> Makes sense. I mean, there would have to 03:03:28.800 --> 03:03:30.637 be some type of a some type of a deal 03:03:30.720 --> 03:03:34.397 struck where they fall under our wing. 03:03:34.480 --> 03:03:36.157 >> Yeah. Yeah, I mean you could imagine 03:03:36.240 --> 03:03:39.997 some kind of deal with with China um 03:03:40.080 --> 03:03:43.197 between the US and China. It'd have to be like a diplomatic deal at the highest 03:03:43.280 --> 03:03:45.277 levels, which is something along the 03:03:45.360 --> 03:03:48.317 lines of, you know, 03:03:48.400 --> 03:03:50.637 hey, you guys can have Taiwan, but we 03:03:50.720 --> 03:03:55.517 need large scale fabs in, you know, we need large scale chip manufacturing in 03:03:55.600 --> 03:03:59.357 the United States or something like that. And like, you know, maybe there's 03:03:59.440 --> 03:04:01.197 worlds where that kind of deal could get 03:04:01.280 --> 03:04:05.437 could get drawn up. I don't know. Um 03:04:05.520 --> 03:04:08.157 uh but that would I mean that would also 03:04:08.240 --> 03:04:12.637 mean that the United States would just have to say hey all we care about 03:04:12.720 --> 03:04:16.157 actually at this point is is chip manufacturing and that we don't care 03:04:16.240 --> 03:04:19.037 actually about the Chinese people and 03:04:19.120 --> 03:04:23.197 the the country and all that stuff >> man. 03:04:23.280 --> 03:04:28.717 Man [clears throat] and are they working with China at all 03:04:28.800 --> 03:04:30.877 >> capacity >> the TSMC? 03:04:30.960 --> 03:04:35.437 >> Yeah. So, they're um I think they're 03:04:35.520 --> 03:04:38.557 technically not supposed to, but a lot 03:04:38.640 --> 03:04:42.557 of the the um 03:04:42.640 --> 03:04:44.717 Huawei, one of the leading companies in 03:04:44.800 --> 03:04:48.797 China, has been able to get um 03:04:48.880 --> 03:04:52.717 tons of chips from uh tons of dyes, it's 03:04:52.800 --> 03:04:54.797 called, but basically tons of chips or 03:04:54.880 --> 03:05:00.717 chip um uh uh prerequisites from Taiwan. 03:05:00.800 --> 03:05:03.117 And they usually do it through like they 03:05:03.200 --> 03:05:08.077 like start some cutout company that doesn't seem associated with them in 03:05:08.160 --> 03:05:13.117 like Singapore and then that Singaporean company buys a bunch of um uh or 03:05:13.200 --> 03:05:16.397 Malaysia and that or the Singaporean Malaysian companies buy a bunch of chips 03:05:16.480 --> 03:05:18.397 from TSMC and then they mail it back or 03:05:18.480 --> 03:05:21.277 something. But there's clearly been uh 03:05:21.360 --> 03:05:25.597 there's been a lot of TSMC high-end 03:05:25.680 --> 03:05:28.717 um outputs that have gone to to the Chinese companies. 03:05:28.800 --> 03:05:33.870 >> Wow. Wow. [sighs and gasps] 03:05:35.120 --> 03:05:38.237 Scary [__] man. It get it gets I mean I 03:05:38.320 --> 03:05:42.717 think this is where um 03:05:42.800 --> 03:05:44.797 you have to believe like right now if 03:05:44.880 --> 03:05:47.677 you look at the you know just as we were 03:05:47.760 --> 03:05:49.837 right now like if you look at the 03:05:49.920 --> 03:05:53.517 situation and all of the 03:05:53.600 --> 03:05:57.677 um all the dynamics at play right now 03:05:57.760 --> 03:05:59.917 it's it's like it's a powder cake. It's 03:06:00.000 --> 03:06:04.397 like very very very volatile. Um, highly 03:06:04.480 --> 03:06:06.477 problematic in many ways. And this is 03:06:06.560 --> 03:06:08.477 where I mean you just ultimately have to 03:06:08.560 --> 03:06:14.477 believe that there's there's got to be some effort towards diplomatic 03:06:14.560 --> 03:06:15.917 solutions. >> Yeah. 03:06:16.000 --> 03:06:21.597 >> Because it is definitely true like war will be really bad for both sides. 03:06:21.680 --> 03:06:24.717 >> Yeah. >> Yeah. How do we coordinate with China 03:06:24.800 --> 03:06:29.597 with the AI? >> Yeah. So, um, 03:06:29.680 --> 03:06:31.277 >> what does that look like? 03:06:31.360 --> 03:06:33.757 >> So, yeah, right now, right now, we're 03:06:33.840 --> 03:06:37.517 definitely, um, US and China, um, we're 03:06:37.600 --> 03:06:41.517 definitely in an allout race dynamic. 03:06:41.600 --> 03:06:44.637 And, you know, we're going to race, and I think this correct, we're going to 03:06:44.720 --> 03:06:46.397 race to build the best AI systems. 03:06:46.480 --> 03:06:50.237 They're going to race to build the best AI systems. Um 03:06:50.320 --> 03:06:55.437 uh and we're both all in on this approach and um we're both all in on on 03:06:55.520 --> 03:06:58.797 racing towards building the most advanced AI capabilities, the largest 03:06:58.880 --> 03:07:00.877 data centers, largest capacity, etc., 03:07:00.960 --> 03:07:07.277 etc. Um the uh and this is if you know 03:07:07.360 --> 03:07:10.237 if you recall kind of how um how nuclear 03:07:10.320 --> 03:07:13.837 was um like you know in nuclear uh 03:07:13.920 --> 03:07:15.517 nuclear war as well as application of 03:07:15.600 --> 03:07:19.997 nuclear towards um uh towards a power 03:07:20.080 --> 03:07:24.957 production. It was kind of you know all systems go like everyone racing um 03:07:25.040 --> 03:07:26.637 towards building capacity building 03:07:26.720 --> 03:07:30.477 capability and then uh Chernobyl and 03:07:30.560 --> 03:07:35.117 ThreeM island happen um and it creates 03:07:35.200 --> 03:07:37.757 large scale constrnation around the 03:07:37.840 --> 03:07:40.317 technology um and the risks of those 03:07:40.400 --> 03:07:44.237 technologies and um there were uh a 03:07:44.320 --> 03:07:48.157 bunch of international treaties and there's a large international response 03:07:48.240 --> 03:07:50.557 towards coordinating on nuclear 03:07:50.640 --> 03:07:53.517 technology. Now all said and done if you 03:07:53.600 --> 03:07:58.717 really you know if you look at nuclear like uh that set our country back set 03:07:58.800 --> 03:08:01.357 many countries back you know many generations in terms of power 03:08:01.440 --> 03:08:03.837 generation. But what it took was 03:08:03.920 --> 03:08:07.037 effectively these like smallcale 03:08:07.120 --> 03:08:11.997 disasters um to take place that effectively uh 03:08:12.080 --> 03:08:13.917 that effectively were the forcing 03:08:14.000 --> 03:08:16.957 function for international cooperation. 03:08:17.040 --> 03:08:21.277 Um you can you can imagine a scenario 03:08:21.360 --> 03:08:24.877 with AI where because of all the things 03:08:24.960 --> 03:08:28.237 that we've been talking about um uh 03:08:28.320 --> 03:08:31.277 there's some scenario where um maybe 03:08:31.360 --> 03:08:33.677 some terrorist group or some Nazi actor 03:08:33.760 --> 03:08:36.797 or some you know North Korea or whomever 03:08:36.880 --> 03:08:41.117 somebody decides to use it for um in a 03:08:41.200 --> 03:08:44.317 particularly adversarial or you know 03:08:44.400 --> 03:08:47.437 inhumane way and create and that 03:08:47.520 --> 03:08:51.117 disaster has some large scale fallout. 03:08:51.200 --> 03:08:53.997 So uh it create you know you take out 03:08:54.080 --> 03:08:59.757 the the you take out power in uh like 03:08:59.840 --> 03:09:04.077 one of the largest cities in the world and tons of people die or you take out 03:09:04.160 --> 03:09:07.677 or there's some pathogen that gets released and like tens of millions of 03:09:07.760 --> 03:09:11.357 people die or you know some one of these things happens that causes the 03:09:11.440 --> 03:09:14.557 international community and everyone in the world to realize oh shoot we have to 03:09:14.640 --> 03:09:17.277 be coordinating on this and you know we 03:09:17.360 --> 03:09:19.597 should be collaborating for AI to 03:09:19.680 --> 03:09:22.637 improve our societies and improve our 03:09:22.720 --> 03:09:27.197 economies and improve the lives of our people. But we shouldn't it, you know, 03:09:27.280 --> 03:09:30.797 we need to we need to coordinate on its use towards, 03:09:30.880 --> 03:09:32.317 for lack of a better term, scary things 03:09:32.400 --> 03:09:35.037 like bio or cyber warfare or, you know, 03:09:35.120 --> 03:09:38.717 the list goes on. Um, so 03:09:38.800 --> 03:09:40.477 long story short, I think the path 03:09:40.560 --> 03:09:44.717 really is um some kind of, you know, we 03:09:44.800 --> 03:09:46.317 sometimes we talk about like an AI oil 03:09:46.400 --> 03:09:49.757 spill or some kind of of incident that 03:09:49.840 --> 03:09:53.917 really causes the international community to realize like, hey, we we 03:09:54.000 --> 03:09:55.597 have to we have to start coordinating on 03:09:55.680 --> 03:09:58.397 this. I mean, it's you say 03:09:58.480 --> 03:10:01.677 China's all out, you know, gone all in 03:10:01.760 --> 03:10:06.557 on the race day ice and the US has gone all out on the race day, but we're 03:10:06.640 --> 03:10:08.557 kneecapping ourselves. I mean, you just 03:10:08.640 --> 03:10:12.157 mentioned the red tape, the EPA, 03:10:12.240 --> 03:10:16.717 the permitting, and the power and we're not producing 03:10:16.800 --> 03:10:20.957 more power. We're flatlined. We've established that, 03:10:21.040 --> 03:10:23.037 as far as I know, we're not getting rid 03:10:23.120 --> 03:10:25.997 of the red tape, you know, to to to jet 03:10:26.080 --> 03:10:33.080 launch this. And I mean, it just seems like 03:10:34.640 --> 03:10:36.237 we're cutting ourselves off at the knees 03:10:36.320 --> 03:10:38.637 here, 03:10:38.720 --> 03:10:42.077 right? Right now, I mean, we have a lot 03:10:42.160 --> 03:10:43.917 of work to do for sure. We have to we 03:10:44.000 --> 03:10:46.717 have to build strategies to win to have 03:10:46.800 --> 03:10:49.117 energy dominance to have data dominance. 03:10:49.200 --> 03:10:53.437 Um to on the algorithms I think we'll be okay. They're gonna espionage but I 03:10:53.520 --> 03:10:57.437 think we'll be okay on algorithms. Um we need to ensure we have chip dominance 03:10:57.520 --> 03:11:01.837 long term. Uh we need to make sure all this lends itself to military dominance. 03:11:01.920 --> 03:11:03.197 I totally agree with you. I mean we need 03:11:03.280 --> 03:11:05.517 to 03:11:05.600 --> 03:11:10.077 we need to today ensure that we have the 03:11:10.160 --> 03:11:12.477 proper strategies in place so that we 03:11:12.560 --> 03:11:14.637 stay ahead on all these areas. The worst 03:11:14.720 --> 03:11:17.677 case scenario for the United States is 03:11:17.760 --> 03:11:20.557 the following which is 03:11:20.640 --> 03:11:23.197 um CCP 03:11:23.280 --> 03:11:25.677 uh does a large scale Manhattan style 03:11:25.760 --> 03:11:30.797 project inside their country um realizes they can start because of 03:11:30.880 --> 03:11:33.597 all the factors that we've talked about they they realize they can start 03:11:33.680 --> 03:11:36.557 overtaking the US on AI that lends 03:11:36.640 --> 03:11:39.757 itself to extreme hyper military 03:11:39.840 --> 03:11:44.637 advantage and they use that to take over 03:11:44.720 --> 03:11:46.637 the world. That's like that's like worst 03:11:46.720 --> 03:11:50.317 case scenario for the US. If US and AI 03:11:50.400 --> 03:11:53.517 AI US and China AI capabilities are even 03:11:53.600 --> 03:11:57.677 just roughly on par, I think you have deterrence. I don't think either country 03:11:57.760 --> 03:12:00.397 will take the risk. I think if US is way 03:12:00.480 --> 03:12:03.517 ahead of China, I think you maintain US 03:12:03.600 --> 03:12:07.837 leadership and that's a pretty safe world. So the the the worst case 03:12:07.920 --> 03:12:10.317 scenario is they get ahead of us. 03:12:10.400 --> 03:12:15.757 Are there any other players other than the US and China involved in this? 03:12:15.840 --> 03:12:18.957 Who else do we need to be watching out for? 03:12:19.040 --> 03:12:21.837 >> So yeah, right now definitely US and 03:12:21.920 --> 03:12:26.477 China. Um the a lot of other countries 03:12:26.560 --> 03:12:31.037 will matter. Um but not all of them have enough 03:12:31.120 --> 03:12:32.797 ingredients to really properly be AI 03:12:32.880 --> 03:12:35.597 superpowers. So um but other countries 03:12:35.680 --> 03:12:39.117 are going to they have they have key 03:12:39.200 --> 03:12:46.200 ingredients. So um to to name a few a uh 03:12:46.800 --> 03:12:50.957 everything we've talked about with with cyber warfare and information warfare, 03:12:51.040 --> 03:12:52.957 information operations, 03:12:53.040 --> 03:12:55.117 Russia has very advanced operations in 03:12:55.200 --> 03:12:59.437 in those areas. Um and that could end up 03:12:59.520 --> 03:13:01.757 mattering a lot if they ally with the 03:13:01.840 --> 03:13:04.637 with the CCP. Um there's a lot of ways 03:13:04.720 --> 03:13:07.437 they can team up and and have uh and 03:13:07.520 --> 03:13:11.037 that could be pretty bad. Um 03:13:11.120 --> 03:13:14.877 there's uh you know the the countries in 03:13:14.960 --> 03:13:17.197 the Middle East will be very important 03:13:17.280 --> 03:13:20.477 because they have um incredible amounts 03:13:20.560 --> 03:13:23.597 of capital and they have lots of energy. 03:13:23.680 --> 03:13:26.717 Uh and so 03:13:26.800 --> 03:13:28.637 that's these are you know they're 03:13:28.720 --> 03:13:31.357 critical um players in how all this 03:13:31.440 --> 03:13:34.717 plays out. India matters a lot. India 03:13:34.800 --> 03:13:37.037 has a lot of high-end technical talent. 03:13:37.120 --> 03:13:41.197 Um uh I don't know if I think right now 03:13:41.280 --> 03:13:44.557 I don't know if between India and China which has more high-end technical talent 03:13:44.640 --> 03:13:47.437 but there's a lot in India for sure. Um 03:13:47.520 --> 03:13:50.237 massive population um also starting to 03:13:50.320 --> 03:13:53.757 industrialize in a real way. Um and 03:13:53.840 --> 03:13:55.757 right next to right next to China so 03:13:55.840 --> 03:14:00.877 India will matter a lot. And then um you know there's a lot of there's a lot of 03:14:00.960 --> 03:14:04.077 technical talent in Europe as well. I think it's unclear exactly how this 03:14:04.160 --> 03:14:07.197 plays out um with the European 03:14:07.280 --> 03:14:09.997 capabilities. I mean they have to 03:14:10.080 --> 03:14:12.637 um it seems like there's some efforts 03:14:12.720 --> 03:14:16.237 now for Europe to try to uh build up 03:14:16.320 --> 03:14:17.917 large scale power, build up large data 03:14:18.000 --> 03:14:21.677 centers, you know, um 03:14:21.760 --> 03:14:25.517 uh make a play. I think yet to be seen how effective those efforts are going to 03:14:25.600 --> 03:14:27.837 be, but um you can clearly see some 03:14:27.920 --> 03:14:31.357 scenarios where if they make a hard a 03:14:31.440 --> 03:14:33.597 hard turn and and go all in, they could 03:14:33.680 --> 03:14:35.837 be relevant as well. 03:14:35.920 --> 03:14:38.717 Is there a world where AI takes on a 03:14:38.800 --> 03:14:43.517 mind of its own? So, 03:14:43.600 --> 03:14:47.997 uh, you know, obviously you can hypothetically paint the scenario where 03:14:48.080 --> 03:14:49.917 like, you know, you have super 03:14:50.000 --> 03:14:51.597 intelligence or you have really powerful 03:14:51.680 --> 03:14:55.917 AI and then, um, you know, it realizes 03:14:56.000 --> 03:14:57.917 at some point that humans are kind of 03:14:58.000 --> 03:15:01.357 annoying and takes us all out. But uh 03:15:01.440 --> 03:15:03.517 but I think I think it's a very like 03:15:03.600 --> 03:15:06.717 that's so preventable um uh as an 03:15:06.800 --> 03:15:09.597 outcome because 03:15:09.680 --> 03:15:11.917 first of all all the things we just 03:15:12.000 --> 03:15:16.877 talked about are like the very real things that happen long before you have 03:15:16.960 --> 03:15:20.637 you know this hyper advanced AI that takes everyone out. That's first that's 03:15:20.720 --> 03:15:22.397 first thing. So we have lots of things 03:15:22.480 --> 03:15:24.717 we have to get right before then. And 03:15:24.800 --> 03:15:31.197 then second is um you know for AI to 03:15:31.280 --> 03:15:35.357 actually um be capable of you know 03:15:35.440 --> 03:15:38.717 having a mind of its own and taking all humans out like we'd have to give it 03:15:38.800 --> 03:15:41.757 just incredible amounts of control. Like 03:15:41.840 --> 03:15:46.077 it would have to just basically be running everything and we're just sort 03:15:46.160 --> 03:15:48.077 of like along for the ride. >> Mhm. 03:15:48.160 --> 03:15:51.757 >> And that's a choice. We have this choice of 03:15:51.840 --> 03:15:56.557 whether or not to like give all of our control to AI systems. And as I was 03:15:56.640 --> 03:15:57.757 talking about before with like human 03:15:57.840 --> 03:16:01.677 sovereignty, I my belief is we should 03:16:01.760 --> 03:16:06.157 not seed control of our most critical systems. Like we should we should design 03:16:06.240 --> 03:16:10.717 all the systems such that human decision-m human control is really 03:16:10.800 --> 03:16:12.477 really important. Human oversight is 03:16:12.560 --> 03:16:15.277 really important. Um, this is one of the 03:16:15.360 --> 03:16:18.957 things that I actually think is is um, 03:16:19.040 --> 03:16:22.717 one of the things that we're working on as a company. So, honestly, one of like 03:16:22.800 --> 03:16:25.037 as I think about like long-term missions, one of the most important 03:16:25.120 --> 03:16:28.397 things is creating human sovereignty. So 03:16:28.480 --> 03:16:30.557 first is how do we make sure all the 03:16:30.640 --> 03:16:33.597 data that goes into these AI models um 03:16:33.680 --> 03:16:35.597 increases human sovereignty such that 03:16:35.680 --> 03:16:37.757 the models are going to do what we tell 03:16:37.840 --> 03:16:41.197 them are aligned with humans and aligned 03:16:41.280 --> 03:16:45.117 with um uh our objectives and two is 03:16:45.200 --> 03:16:48.797 that we create oversight. So as AI 03:16:48.880 --> 03:16:53.677 starts doing more and more actions, doing more planning, you know, taking 03:16:53.760 --> 03:16:57.837 out, you know, carrying out more things in in the in the world, in the economy, 03:16:57.920 --> 03:16:59.997 in military, etc. that humans are 03:17:00.080 --> 03:17:01.517 watching and supervising every one of 03:17:01.600 --> 03:17:05.437 those actions. So that's that's how we 03:17:05.520 --> 03:17:07.517 maintain control and that's how we 03:17:07.600 --> 03:17:09.757 prevent, you know, the Terminator 03:17:09.840 --> 03:17:14.637 scenarios or the, you know, AI takes us out kind of scenarios. 03:17:14.720 --> 03:17:18.797 >> Interesting. Well, Alex, wrapping up the interview 03:17:18.880 --> 03:17:20.797 here, but man, what a fascinating 03:17:20.880 --> 03:17:25.997 discussion. Thank you. Thank you for being here. One last question. If you 03:17:26.080 --> 03:17:27.517 had three guests you'd like to see on 03:17:27.600 --> 03:17:30.077 the show, who would it be? 03:17:30.160 --> 03:17:34.237 Oh, it's a good question. Um, 03:17:34.320 --> 03:17:37.357 who would I like to see? 03:17:37.440 --> 03:17:40.157 Uh, well, I really like what you've been 03:17:40.240 --> 03:17:41.997 doing recently, which is getting more 03:17:42.080 --> 03:17:46.637 tech folks on the on the pod. Um, so, 03:17:46.720 --> 03:17:49.037 uh, so I go in that direction. I mean, I 03:17:49.120 --> 03:17:50.397 think Elon would be great to see on the 03:17:50.480 --> 03:17:55.197 show. I think, uh, I think, 03:17:55.280 --> 03:17:58.477 um, we were talking about this. Zach would be would be cool to see on the 03:17:58.560 --> 03:18:01.517 show. I think, uh, Sam Alman would be 03:18:01.600 --> 03:18:03.117 cool to see on the show. So, definitely 03:18:03.200 --> 03:18:06.397 like, um, more people in tech. Outside 03:18:06.480 --> 03:18:09.997 of that, um 03:18:10.080 --> 03:18:13.517 I think uh and we were talking about 03:18:13.600 --> 03:18:17.197 some of this like um 03:18:17.280 --> 03:18:20.957 international leadership like international like 03:18:21.040 --> 03:18:25.997 leaders of other countries is super important because um we talk about all 03:18:26.080 --> 03:18:29.997 these scenarios like international cooperation is going to matter so much. 03:18:30.080 --> 03:18:32.957 >> Right on. we'll we'll reach out to them 03:18:33.040 --> 03:18:36.237 and uh you know as far as world leaders 03:18:36.320 --> 03:18:39.517 is concerned we're we're on it but um 03:18:39.600 --> 03:18:42.397 well Alex thanks again for coming man 03:18:42.480 --> 03:18:45.677 fascinating discussion I'm just super 03:18:45.760 --> 03:18:47.917 happy to see all the success that you've 03:18:48.000 --> 03:18:51.357 amassed throughout your 28 years it's it 03:18:51.440 --> 03:18:55.677 is I love seeing it. So thank you for being 03:18:55.760 --> 03:18:59.757 here. I know you're a busy guy. So >> yeah, thanks for having me. It was fun. 03:18:59.840 --> 03:19:03.460 Heat. Heat. [music] 03:19:08.650 --> 03:19:10.670 [music]