WEBVTT 00:00:07.960 --> 00:00:11.117 Thanks for having me. >> Great to have you, Sam. 00:00:11.200 --> 00:00:14.117 >> It's interesting cuz peo- people have been talking about 00:00:14.200 --> 00:00:19.277 the potential downsides uh or serious risks of AI for some time. 00:00:19.360 --> 00:00:22.757 And I think a lot of people have been saying, "Why is this the moment where it 00:00:22.840 --> 00:00:27.797 sort of escaped the bubble and is now all of a sudden an international topic 00:00:27.880 --> 00:00:32.317 of such uh intensity?" And I I think a few things 00:00:32.400 --> 00:00:36.717 have happened. One, the models have gotten so good and 00:00:36.800 --> 00:00:40.077 people for a long time looked at the models. 00:00:40.160 --> 00:00:43.637 Maybe they looked at the rate of progress, but I don't think people really felt that 00:00:43.720 --> 00:00:47.437 the models were going to come this far this fast. 00:00:47.520 --> 00:00:49.757 We will get it right, by the way. I am 00:00:49.840 --> 00:00:55.357 very confident in our I think the biggest thing that has changed is 00:00:55.440 --> 00:00:58.997 this thing that our industry has been talking about for a while, which is no, 00:00:59.080 --> 00:01:02.717 no, this technology, this approach is going to go very far. These models are 00:01:02.800 --> 00:01:08.117 in some ways going to become more capable, smarter than people. Um that finally 00:01:08.200 --> 00:01:11.677 happened. So, I think that was a big change to why now when people say, you 00:01:11.760 --> 00:01:15.077 know, my probability of doom is X or Y or whatever, 00:01:15.160 --> 00:01:18.717 it hits in a different way. Technology to do. Somebody else wants to build like 00:01:18.800 --> 00:01:22.997 a hacking bot, okay. Um the world can decide what they think about that, but 00:01:23.080 --> 00:01:27.400 I think you also see um 00:01:27.720 --> 00:01:31.837 a real fear that some of these companies or some companies developing AI could 00:01:31.920 --> 00:01:34.597 get too much power and be able to sort 00:01:34.680 --> 00:01:37.237 of exert undue influence uh on the 00:01:37.320 --> 00:01:39.757 economy, push a worldview out on people. 00:01:39.840 --> 00:01:43.477 Uh and I think the world is right to be afraid of this. So, we have these two 00:01:43.560 --> 00:01:46.637 big challenges. We have the potential of a loss of control 00:01:46.720 --> 00:01:50.797 accident or some other serious thing that could go wrong. Um we have the 00:01:50.880 --> 00:01:54.877 potential of way too much power concentration and people fear that those 00:01:54.960 --> 00:01:58.597 developing AI could exert their worldview. Um 00:01:58.680 --> 00:02:02.277 and there is a sort of narrow path we have to navigate 00:02:02.360 --> 00:02:06.717 with pragmatism, with you know, 00:02:06.800 --> 00:02:10.637 steadiness, with the ability for the the world to like trust that we will make 00:02:10.720 --> 00:02:14.917 dependable decisions, consistent decisions uh through this. 00:02:15.000 --> 00:02:17.917 And I think people are saying, "Well, you know, this now seems pretty 00:02:18.000 --> 00:02:22.037 important. Like let's let's get it right." 00:02:22.120 --> 00:02:24.957 I'm not going to get the days of the week exactly right here, but something 00:02:25.040 --> 00:02:29.677 on like a Friday or a Saturday. These models have gotten so good so fast that 00:02:29.760 --> 00:02:33.837 we need to treat the alignment and monitoring and security with a new level 00:02:33.920 --> 00:02:36.597 of rigor. And we have to be willing to 00:02:36.680 --> 00:02:38.757 pace our development of capabilities 00:02:38.840 --> 00:02:44.397 such that capabilities always such that alignment, safety, and 00:02:44.480 --> 00:02:47.837 monitoring are always ahead of capabilities. And 00:02:47.920 --> 00:02:51.157 the Hugging Face incident triggered a real reset not just in our own company, 00:02:51.240 --> 00:02:52.877 but I think the whole industry. 00:02:52.960 --> 00:02:54.597 >> You know, we're in the cyber area today, 00:02:54.680 --> 00:02:56.597 but maybe we move into another area. Is 00:02:56.680 --> 00:03:01.757 this a fair way to characterization or am I or is this unfair to characterize 00:03:01.840 --> 00:03:04.277 it like this? >> What I really care about is that we have 00:03:04.360 --> 00:03:08.637 got a great culture of accident reporting and learning. I think if we 00:03:08.720 --> 00:03:12.517 look at what happens with the FAA and the NTSB and how safe airplanes have 00:03:12.600 --> 00:03:16.997 gotten, um as I understand it from talking to people who have really studied, 00:03:17.080 --> 00:03:20.677 you know, new kinds of technology on the order of magnitude of airplanes, one of 00:03:20.760 --> 00:03:24.477 the things that has gone so well there is 00:03:24.560 --> 00:03:26.277 is the culture of accident reporting. 00:03:26.360 --> 00:03:27.957 You know, accidents are going to happen. 00:03:28.040 --> 00:03:29.717 We're going to learn as much as we can 00:03:29.800 --> 00:03:35.277 and react from each one. Um and if, you know, there's other kinds 00:03:35.360 --> 00:03:39.717 of technology where I don't think it's gone I don't think society's had the same 00:03:39.800 --> 00:03:43.877 response and I don't think we have improved as much. 00:03:43.960 --> 00:03:47.037 to pull this thing together and present you with the information to make a 00:03:47.120 --> 00:03:48.677 decision or to bl- 00:03:48.760 --> 00:03:51.877 I think this is one of the most profound 00:03:51.960 --> 00:03:56.117 changes to how we use technology that I've ever seen. 00:03:56.200 --> 00:03:59.437 I mean, sci-fi movies, as you mentioned, have been hinting at this or playing 00:03:59.520 --> 00:04:02.717 around with this for a long time, but this is one of those moments where it 00:04:02.800 --> 00:04:05.997 feels like you're in the movie. Like the sci-fi thing came true. 00:04:06.080 --> 00:04:10.277 You're talking to your personal super assistant, and you are working 10 times 00:04:10.360 --> 00:04:14.117 faster or better or more creatively than you could have before. 00:04:14.200 --> 00:04:18.077 >> Sam, you're invited back anytime you want to Dreamforce. Hold on. I'm holding 00:04:18.160 --> 00:04:20.637 it there. Thank you. Thank you. 00:04:20.720 --> 00:04:23.920 >> Please thank Sam Altman.