1 00:00:00,480 --> 00:00:03,018 Thank you all for joining the webinar today. 2 00:00:03,101 --> 00:00:05,997 Uh we're very excited to have with us a discussion 3 00:00:06,080 --> 00:00:11,193 uh between our CEO Ali Gatsi and uh via Sam Alman uh CEO of OpenAI. 4 00:00:11,276 --> 00:00:13,597 So thank you very much both for 5 00:00:13,680 --> 00:00:16,364 for being here. Thank you. Yeah, excited. Yeah. Well, 6 00:00:16,447 --> 00:00:18,557 let's get started and I hope we can have a 7 00:00:18,640 --> 00:00:23,031 fun kind of thoughtful discussion. um you know data bricks and open AAI we recently 8 00:00:23,115 --> 00:00:23,517 announced 9 00:00:23,600 --> 00:00:27,642 kind of first-of-a-kind partnership uh to make the open AAI models natively 10 00:00:27,725 --> 00:00:28,797 available within data 11 00:00:28,880 --> 00:00:32,566 bricks and agent bricks so maybe actually starting with Ali first you know what's 12 00:00:32,649 --> 00:00:33,357 the importance of 13 00:00:33,440 --> 00:00:36,849 partnering with open AI on this on this endeavor yeah I mean every one of 14 00:00:36,932 --> 00:00:37,997 our enterprise customers 15 00:00:38,080 --> 00:00:41,283 want to use openAI like you know uh they all want to have the models available they 16 00:00:41,366 --> 00:00:41,917 want to use them 17 00:00:42,000 --> 00:00:46,478 on their enterprise data uh and you know they're you know getting just those two 18 00:00:46,562 --> 00:00:47,277 things working 19 00:00:47,360 --> 00:00:51,574 together is non-trivial because it's like the data is sensitive They want to have 20 00:00:51,657 --> 00:00:54,928 their privacy, you know, they want to have auditing, they want to have GDPR rights, 21 00:00:55,011 --> 00:00:56,157 but they want to like, you know, 22 00:00:56,240 --> 00:00:59,907 they want to use, you know, the models to build agents to get insights. So, 23 00:00:59,990 --> 00:01:04,298 it's like it's really customer demand that uh, you know, has been overwhelming. Yeah, 24 00:01:04,381 --> 00:01:05,677 we're we're thrilled to do 25 00:01:05,760 --> 00:01:08,347 this. Enterprise is becoming uh one of our biggest focuses. 26 00:01:08,430 --> 00:01:10,157 We've had like 7x enterprise growth this 27 00:01:10,240 --> 00:01:13,795 year and uh you know we we really think that we're heading into the phase of AI 28 00:01:13,879 --> 00:01:14,717 where the models are 29 00:01:14,800 --> 00:01:18,567 getting so good that enterprises will need to use them want to use them bring them 30 00:01:18,650 --> 00:01:19,037 into their 31 00:01:19,120 --> 00:01:22,910 uh into their whole ecosystem and we cannot imagine a better partner than data 32 00:01:22,993 --> 00:01:23,357 bricks to 33 00:01:23,440 --> 00:01:26,395 to make that happen. So we're thrilled. We're also very happy to be a data bicks 34 00:01:26,478 --> 00:01:27,117 customer. Yeah know 35 00:01:27,200 --> 00:01:30,112 that's amazing. you know, as as you said, you know, the partnership, 36 00:01:30,195 --> 00:01:33,861 not only is of course data bricks bringing open AAI models to the M, you know, 37 00:01:33,944 --> 00:01:35,677 all of our enterprise customers, um, 38 00:01:35,760 --> 00:01:40,371 but also for some time now, OpenAI has been using data bricks for data analytics. 39 00:01:40,455 --> 00:01:41,357 So maybe starting 40 00:01:41,440 --> 00:01:44,627 with maybe Ali, like how has that partnership and that work been going? 41 00:01:44,710 --> 00:01:45,917 It's awesome working. I mean 42 00:01:46,000 --> 00:01:50,777 like extremely technical teams that push us to you know our boundaries and uh you 43 00:01:50,860 --> 00:01:51,437 know demand 44 00:01:51,520 --> 00:01:54,756 more and one more of us and uh you know it's nice that the two companies are so 45 00:01:54,840 --> 00:01:59,154 close geographically so it's just easy to walk over and iterate on the product. 46 00:01:59,237 --> 00:02:01,117 Uh but uh yeah it's a one of our 47 00:02:01,200 --> 00:02:05,583 most demanding customers in a very good way like made us just way way way better 48 00:02:05,667 --> 00:02:06,477 as a company. Um 49 00:02:06,560 --> 00:02:10,782 maybe moving separately a little bit you know Sam you and open have really 50 00:02:10,865 --> 00:02:11,597 revolutionized 51 00:02:11,680 --> 00:02:16,057 AI for consumers. Um how do you think on this new chapter as you mentioned about 52 00:02:16,141 --> 00:02:17,117 going to enterprise 53 00:02:17,200 --> 00:02:21,631 how data bricks and open AI together you know can revolutionize AI for enterprise 54 00:02:21,714 --> 00:02:22,077 as well? 55 00:02:22,160 --> 00:02:26,562 Yeah we always plan to do enterprise also but the models started with a lot 56 00:02:26,645 --> 00:02:30,761 of problems they started fairly weak. It was easier to get consumers to adopt them. 57 00:02:30,844 --> 00:02:32,477 They're coming to the point in their 58 00:02:32,560 --> 00:02:38,081 maturity where we are clearly seeing together huge enterprise demand. Um, 59 00:02:38,164 --> 00:02:40,077 and the need to figure out 60 00:02:40,160 --> 00:02:44,410 how to bring this technology to enterprises in a way that they can use it 61 00:02:44,493 --> 00:02:45,597 with all their data, 62 00:02:45,680 --> 00:02:48,887 with all their constraints, with all of the concerns they have about security 63 00:02:48,970 --> 00:02:49,357 and safety, 64 00:02:49,440 --> 00:02:56,411 but but also uh to sort of quickly jump into this new world where you'll have AI 65 00:02:56,495 --> 00:03:01,383 doing an increasing amount of the intellectual work in an enterprise. 66 00:03:01,467 --> 00:03:03,517 um this feels like it's really 67 00:03:03,600 --> 00:03:08,598 the time for I think in 2026 2027 we'll see a huge transformation um of how 68 00:03:08,682 --> 00:03:10,157 enterprises think about 69 00:03:10,240 --> 00:03:14,147 this maybe the the example we can look at from 2025 is what's happened with how code 70 00:03:14,230 --> 00:03:14,717 gets written 71 00:03:14,800 --> 00:03:19,420 right yes and now if you imagine that for every other function of the enterprise uh 72 00:03:19,503 --> 00:03:20,157 that would be 73 00:03:20,240 --> 00:03:25,085 quite a big deal and thrilled to get to go do that together yeah I'm super excited I 74 00:03:25,168 --> 00:03:25,437 mean I 75 00:03:25,520 --> 00:03:29,692 think there's a lot of uh context in the data that enterprises have it's very 76 00:03:29,775 --> 00:03:30,797 proprietary right So 77 00:03:30,880 --> 00:03:34,261 uh the consumer side initially I presume started with like okay we have all 78 00:03:34,344 --> 00:03:35,277 this data in the world 79 00:03:35,360 --> 00:03:39,590 that's you know public mankind is created over thousands of years what if the AI 80 00:03:39,673 --> 00:03:40,237 could really 81 00:03:40,320 --> 00:03:43,913 sort of compress that and understand that and reason on that but this then there's 82 00:03:43,996 --> 00:03:44,317 this like 83 00:03:44,400 --> 00:03:49,460 data that's like not available to the LMS which we now can bring together so now you 84 00:03:49,543 --> 00:03:50,317 know providing 85 00:03:50,400 --> 00:03:55,129 the data that they have that's proprietary uh in the enterprise as context to 86 00:03:55,212 --> 00:03:56,317 the agents which is 87 00:03:56,400 --> 00:04:00,327 really what they need that I think is going to be like the big unlock just 88 00:04:00,410 --> 00:04:01,357 excited to uh build 89 00:04:01,440 --> 00:04:06,220 that together. Yeah. And you know on that point both Sam and Ali on sort of 90 00:04:06,303 --> 00:04:07,517 the model capability 91 00:04:07,600 --> 00:04:10,791 reaching a point where you know enterprises make a lot of sense. 92 00:04:10,874 --> 00:04:11,917 The other piece on top 93 00:04:12,000 --> 00:04:17,117 of that has been this sort of agents kind of age of agents sort of arising and I 94 00:04:17,200 --> 00:04:18,157 want to ask both 95 00:04:18,240 --> 00:04:22,957 of you a little bit more on the research side. you know now with AI research moving 96 00:04:23,040 --> 00:04:27,358 um from you know building great great frontier models to now including hey you 97 00:04:27,441 --> 00:04:28,317 mentioned context 98 00:04:28,400 --> 00:04:33,186 management right you mentioned agents and and multi-agentic patterns and systems 99 00:04:33,270 --> 00:04:33,917 where do you 100 00:04:34,000 --> 00:04:37,633 think the future innovations like across these research buckets will really kind 101 00:04:37,716 --> 00:04:38,237 of drive this 102 00:04:38,320 --> 00:04:42,807 move into into enterprise I mean I think there will be all of these different 103 00:04:42,890 --> 00:04:43,757 research buckets 104 00:04:43,840 --> 00:04:47,207 like we'll keep pushing on pre-training we will you know get these multi- aent 105 00:04:47,290 --> 00:04:47,517 systems 106 00:04:47,600 --> 00:04:51,395 that are doing complicated things the models will get smarter across the board. 107 00:04:51,478 --> 00:04:55,703 But there are maybe kind of two axes that I would think about here. 108 00:04:55,786 --> 00:04:58,077 Uh one is nothing at all to do with 109 00:04:58,160 --> 00:05:01,679 model intelligence or very little. It's about how tightly can you integrate it 110 00:05:01,762 --> 00:05:02,557 into an enterprise, 111 00:05:02,640 --> 00:05:05,400 all of the knowledge they have, all of the different data sources, 112 00:05:05,483 --> 00:05:06,477 their business processes, 113 00:05:06,560 --> 00:05:10,928 the the way that they want to work. Like you could bring a very brilliant physicist 114 00:05:11,011 --> 00:05:11,357 and drop 115 00:05:11,440 --> 00:05:14,625 them into data bricks and maybe they wouldn't get that much done on the first 116 00:05:14,709 --> 00:05:15,517 day because they'd be 117 00:05:15,600 --> 00:05:19,323 missing all of that understanding. they would know how to even though they had 118 00:05:19,406 --> 00:05:20,397 crazy intelligence. So 119 00:05:20,480 --> 00:05:26,186 can be a problem. Yeah, I'm sure. Uh so so I think there's this one layer that 120 00:05:26,270 --> 00:05:27,597 the whole ecosystem 121 00:05:27,680 --> 00:05:30,797 together this will not just be either our companies but the whole world is going 122 00:05:30,880 --> 00:05:31,197 to have to 123 00:05:31,280 --> 00:05:38,280 build kind of an integration uh into these models and sort of with these models 124 00:05:38,632 --> 00:05:40,317 and that is that is 125 00:05:40,400 --> 00:05:44,066 a lot of work to really figure out how you're going to you know teach a model 126 00:05:44,150 --> 00:05:48,491 to be a sort of productive employee in a particular enterprise or whatever. 127 00:05:48,574 --> 00:05:50,557 Um and then the other category we 128 00:05:50,640 --> 00:05:57,640 uh we have seen a dramatic increase in how long of a horizon a model can work on 129 00:05:57,814 --> 00:05:58,717 a task for. 130 00:05:58,800 --> 00:06:02,321 Yeah. Yeah. So the way the way that I think about this is for a particular class 131 00:06:02,405 --> 00:06:06,646 of task once you've given it that sort of connection to an enterprise that kind 132 00:06:06,729 --> 00:06:07,437 of integration 133 00:06:07,520 --> 00:06:14,520 um how long of a task does the model have a 50% chance of success at gone from if we 134 00:06:15,867 --> 00:06:16,877 think about 135 00:06:16,960 --> 00:06:23,960 say coding we've gone from 5-second tasks at the launch of GPT3.5 to 5 minute tasks 136 00:06:24,440 --> 00:06:24,717 with 137 00:06:24,800 --> 00:06:31,800 various GPT4 iterations to 5-hour tasks with um GP5 and that's remarkable. 138 00:06:34,310 --> 00:06:36,797 Uh but a lot of what 139 00:06:36,880 --> 00:06:42,468 an enterprise does is tasks that require months or years sometimes and also we can 140 00:06:42,552 --> 00:06:43,437 only currently 141 00:06:43,520 --> 00:06:48,606 do this in this one vertical. So lengthening the the horizon that these 142 00:06:48,689 --> 00:06:49,917 models can work on 143 00:06:50,000 --> 00:06:53,912 um giving them all of the context that exists inside an enterprise or the world 144 00:06:53,995 --> 00:06:54,317 and then 145 00:06:54,400 --> 00:06:58,317 adding that to like many more verticals I think will be the important thrust. Yeah, 146 00:06:58,400 --> 00:07:02,968 I mean I I think that one is the 50th percentile task completion time horizon 147 00:07:03,051 --> 00:07:04,237 is super interesting. 148 00:07:04,320 --> 00:07:08,112 uh and I think if you bring more context so the first thing you mentioned also 149 00:07:08,195 --> 00:07:08,957 with more context 150 00:07:09,040 --> 00:07:13,049 uh you know you can increase that horizon like you know lengthen that so the task 151 00:07:13,133 --> 00:07:13,757 something that 152 00:07:13,840 --> 00:07:18,598 would take a human a day to do the the the agents can now do um uh not that we have 153 00:07:18,682 --> 00:07:19,357 a day yet but 154 00:07:19,440 --> 00:07:22,677 I'm saying if you add that context you can get that longer and longer and one thing 155 00:07:22,760 --> 00:07:23,117 that we saw 156 00:07:23,200 --> 00:07:27,708 that's really cool is just optimizing the context automatically like you know we we 157 00:07:27,791 --> 00:07:28,317 developed a 158 00:07:28,400 --> 00:07:33,262 technique called uh ja uh which like basically um uh inspired by genetic 159 00:07:33,345 --> 00:07:35,117 algorithms But uh basically 160 00:07:35,200 --> 00:07:38,170 gets that enterprise context that you have and gets it into the model 161 00:07:38,254 --> 00:07:39,277 but automatically so that 162 00:07:39,360 --> 00:07:42,437 you don't have to have humans that are sitting there and optimizing the context 163 00:07:42,520 --> 00:07:43,037 so that you can 164 00:07:43,120 --> 00:07:46,677 live feed the relevant context that's sitting in the different docs and inside 165 00:07:46,760 --> 00:07:47,517 of the enterprise. 166 00:07:47,600 --> 00:07:50,651 Uh so I think that's that's a really good metric to look at and see you know and 167 00:07:50,735 --> 00:07:51,357 the progression of 168 00:07:51,440 --> 00:07:56,750 it has been like insane as Sam said like from like seconds now to like many many 169 00:07:56,833 --> 00:07:57,357 hours. So 170 00:07:57,440 --> 00:08:01,420 very excited about that. It always yeah it always strikes me that the the model 171 00:08:01,503 --> 00:08:02,397 capability today is 172 00:08:02,480 --> 00:08:06,431 the dumbest it will ever be for the rest of my lifetime you know and that's like 173 00:08:06,515 --> 00:08:07,037 a incredible 174 00:08:07,120 --> 00:08:11,137 thought to have whenever you think about the potential applications of this thing. 175 00:08:11,220 --> 00:08:15,680 Um and you know as as we mentioned sort of agents getting deeper enterprise operating 176 00:08:15,763 --> 00:08:16,877 on hours and hours and 177 00:08:16,960 --> 00:08:20,111 potentially even taking actions in the enterprise what are the main things 178 00:08:20,195 --> 00:08:21,117 to think about in terms 179 00:08:21,200 --> 00:08:26,324 of ensuring kind of responsible AI and governance over the tools and the data 180 00:08:26,407 --> 00:08:27,677 and and the actions? 181 00:08:27,760 --> 00:08:32,356 Yeah, I mean we've worked very closely on that and we've made sure that uh we we 182 00:08:32,440 --> 00:08:35,466 sort of build with privacy and security from the ground up. 183 00:08:35,549 --> 00:08:38,077 So the whole uh partnership and integration builds 184 00:08:38,160 --> 00:08:41,101 on having the guardrails in place, making sure that you have audit logging 185 00:08:41,184 --> 00:08:41,837 on everything that 186 00:08:41,920 --> 00:08:45,081 happens in the enterprise so that you can track like what did it exactly do so you 187 00:08:45,165 --> 00:08:45,517 can go back 188 00:08:45,600 --> 00:08:49,648 and track those things. Having access control uh making sure I mean simple 189 00:08:49,732 --> 00:08:50,877 things for enterprises 190 00:08:50,960 --> 00:08:55,315 that maybe you know sometimes takes for granted but uh you know is it on brand 191 00:08:55,399 --> 00:08:56,397 yeah what the model 192 00:08:56,480 --> 00:08:59,298 is saying is it recommending a competitor things like that. 193 00:08:59,381 --> 00:09:01,117 So you can build those guard rails in 194 00:09:01,200 --> 00:09:05,529 from the beginning into the uh into the models and that's so if you use 35 now 195 00:09:05,612 --> 00:09:06,717 inside data books you 196 00:09:06,800 --> 00:09:10,306 get all of that out of the box. Uh so we're very excited about that. 197 00:09:10,389 --> 00:09:11,837 This is I think this is going 198 00:09:11,920 --> 00:09:15,424 to become the fundamental limiter for adoption of AI in the enterprise. 199 00:09:15,507 --> 00:09:18,317 Uh it won't be about the intelligence. It won't be about the price. 200 00:09:18,400 --> 00:09:20,557 We're going to I believe the research teams and the sort 201 00:09:20,640 --> 00:09:22,957 of infrastructure teams will figure that out. 202 00:09:23,040 --> 00:09:25,677 But I think enterprises are are starting to realize 203 00:09:25,760 --> 00:09:30,881 very quickly just how critical this is and it will be the limiting reagent. 204 00:09:30,964 --> 00:09:32,477 Yeah 100%. I agree. You 205 00:09:32,560 --> 00:09:37,160 know maybe this is a question for Sam but as we and data bricks can attest you know 206 00:09:37,243 --> 00:09:37,837 the open way 207 00:09:37,920 --> 00:09:42,587 model GPOSS have been a gamecher for a lot of our companies both in terms of 208 00:09:42,670 --> 00:09:43,837 the capabilities and 209 00:09:43,920 --> 00:09:47,742 the use cases can tackle the customizability that an open way model you 210 00:09:47,825 --> 00:09:49,117 know involved and I think 211 00:09:49,200 --> 00:09:54,760 openai just recently released uh the GBD OSS uh safeguards version as well curious 212 00:09:54,843 --> 00:09:55,517 yeah what's 213 00:09:55,600 --> 00:10:00,782 the future for open source and open weights from from open AAI there's clearly 214 00:10:00,865 --> 00:10:01,997 demand for for it. 215 00:10:02,080 --> 00:10:06,695 There's honestly much less demand than there is for the most capable uh you know 216 00:10:06,778 --> 00:10:07,517 models that we 217 00:10:07,600 --> 00:10:10,983 can run in a cloud. But I think we the answer is we're you know a bigish company 218 00:10:11,067 --> 00:10:11,677 and we should do 219 00:10:11,760 --> 00:10:16,040 both. People people want a model that they control and that they can run on 220 00:10:16,124 --> 00:10:17,437 their device or in their 221 00:10:17,520 --> 00:10:21,975 own system and there needs to be some version of some way we support them. 222 00:10:22,059 --> 00:10:22,957 Yeah. Um so I am 223 00:10:23,040 --> 00:10:27,268 curious like we're seeing huge demand on GP2oss like people really want you know 224 00:10:27,352 --> 00:10:28,077 especially like 225 00:10:28,160 --> 00:10:32,713 you know an American open source model that like kind of uh uh is at the very 226 00:10:32,796 --> 00:10:33,917 frontier. Uh what is 227 00:10:34,000 --> 00:10:41,000 open opens plans going forward? Like is is G2S the last? No no I I I hope not. 228 00:10:41,248 --> 00:10:43,117 Um I we are trying to 229 00:10:43,200 --> 00:10:49,062 figure out how we can do a someday do a model that is like the quality of GPT5 uh 230 00:10:49,146 --> 00:10:50,237 running on a you 231 00:10:50,320 --> 00:10:54,450 know one device open source W model. We don't know how to do that yet but uh we 232 00:10:54,533 --> 00:10:55,517 didn't think we were 233 00:10:55,600 --> 00:11:00,454 going to get a model as good as GPT OSS running you know in 120 billion parameters 234 00:11:00,538 --> 00:11:05,527 either. Um, so again, I don't think this will be the what most people want, 235 00:11:05,610 --> 00:11:07,837 but the people who want it really 236 00:11:07,920 --> 00:11:11,513 want it. And I think we'll find we'll try to find ways to offer incredible open 237 00:11:11,596 --> 00:11:14,378 source models. Wow, that's pretty amazing. So you could run on your, you know, 238 00:11:14,462 --> 00:11:15,677 if you could get that form factor down 239 00:11:15,760 --> 00:11:19,364 running on your laptop. We we think about what it's going to look like to make 240 00:11:19,447 --> 00:11:21,917 a sort of, you know, a new computer for like the AI era. 241 00:11:22,000 --> 00:11:23,997 I think that there's a whole bunch of reasons 242 00:11:24,080 --> 00:11:26,338 why that form factor is not quite right. Yeah. 243 00:11:26,422 --> 00:11:29,037 But one part of this is it should run locally a great 244 00:11:29,120 --> 00:11:33,090 model if you wanted to. Yeah. Yeah. So if Wi-Fi is down, you know, 245 00:11:33,173 --> 00:11:35,117 you should AI if Wi-Fi is down or 246 00:11:35,200 --> 00:11:38,953 you're just like, hey, from a privacy perspective, 247 00:11:39,037 --> 00:11:42,637 I I I I believe that privacy and freedom will be 248 00:11:42,720 --> 00:11:46,734 two extremely important principles for how people use AI if it becomes as important 249 00:11:46,817 --> 00:11:52,633 in people's lives as we expect. And the current tech industry, public policy, 250 00:11:52,716 --> 00:11:55,517 cluster, whatever, leaves some area 251 00:11:55,600 --> 00:11:58,997 for concern. Yeah. And and so I think people will want good local models. 252 00:11:59,080 --> 00:12:03,275 Oh, that's super exciting. Yeah, looking forward to that future. Um, you know, 253 00:12:03,358 --> 00:12:05,837 maybe this is a question for Ali, but uh we 254 00:12:05,920 --> 00:12:09,113 as a community today are so focused on vibe coding and, you know, 255 00:12:09,196 --> 00:12:10,877 these small kind of agentic apps in 256 00:12:10,960 --> 00:12:13,932 the short term. What are we underestimating about this technology 257 00:12:14,016 --> 00:12:15,437 in the long term, especially for 258 00:12:15,520 --> 00:12:19,028 for the enterprise? Yeah, I mean, I think Sam touched on this. I mean, 259 00:12:19,111 --> 00:12:20,157 we're going to see it. 260 00:12:20,240 --> 00:12:23,463 We're going to have co-workers that are agents and they're going to there there's 261 00:12:23,546 --> 00:12:26,662 so many different aspects of the work that they're going to be involved in. 262 00:12:26,745 --> 00:12:28,157 Like even on coding now we're like 263 00:12:28,240 --> 00:12:32,127 just thinking about okay can you produce the code but you know coding at databbooks 264 00:12:32,210 --> 00:12:32,797 is only 20% of 265 00:12:32,880 --> 00:12:36,133 what R&D or an engineer does. We measure it it's only 20% of their time what are 266 00:12:36,216 --> 00:12:36,717 they doing the 267 00:12:36,800 --> 00:12:41,592 rest of the time but it's like design docs it's PRDS it's you know discussions 268 00:12:41,675 --> 00:12:42,717 meetings but every 269 00:12:42,800 --> 00:12:45,295 one of them when you start looking at it it's like there's a place for agents too 270 00:12:45,378 --> 00:12:45,677 and actually 271 00:12:45,760 --> 00:12:48,524 if agents were involved in the whole process it would probably also make 272 00:12:48,607 --> 00:12:49,117 the coding much 273 00:12:49,200 --> 00:12:52,372 much much better. It's context that's missing when they're writing the code. 274 00:12:52,455 --> 00:12:55,322 So, we're going to see, you know, I think end to end for, you know, 275 00:12:55,406 --> 00:12:57,837 all of this kind of productive work. We're going to see 276 00:12:57,920 --> 00:13:01,138 them uh show up and help accelerate everything that's happening inside of 277 00:13:01,221 --> 00:13:04,488 the enterprise. And it's going to happen, you know, slowly. It will take time, 278 00:13:04,571 --> 00:13:06,077 but that task horizon is just getting 279 00:13:06,160 --> 00:13:08,877 longer and longer and longer, which means they can take on more and more complicated 280 00:13:08,960 --> 00:13:14,312 tasks over time. So, it's exciting times. Yeah. Yeah. Um, and then maybe Sam, 281 00:13:14,395 --> 00:13:18,906 for you, I think you've said before that we're in this like once in a history kind 282 00:13:18,990 --> 00:13:20,397 of transition, right? Okay, 283 00:13:20,480 --> 00:13:23,633 now the world is changing much faster than than we think. 284 00:13:23,716 --> 00:13:25,677 Um maybe if you zoom out 5 10 years, 285 00:13:25,760 --> 00:13:28,996 what's like the most I don't know surprising institution, right? 286 00:13:29,080 --> 00:13:33,627 Schools, cities, religions, hospitals that you think will look unrecognizable, right, 287 00:13:33,710 --> 00:13:35,837 with those generations. I think once in 288 00:13:35,920 --> 00:13:39,095 history is too big of a statement. Okay, I can say once in a generation. 289 00:13:39,178 --> 00:13:42,821 Once in a generation. Yeah. Um the econ I think the economy changes a lot. 290 00:13:42,904 --> 00:13:45,597 I I I don't think it's I don't think we're going to 291 00:13:45,680 --> 00:13:49,245 like run out of meaning or purpose or fulfillment or anything like that. Um, 292 00:13:49,328 --> 00:13:50,157 I think humans have 293 00:13:50,240 --> 00:13:55,458 such main character energy. We're so able to like not care about what the machines 294 00:13:55,541 --> 00:13:56,557 do. Like already, 295 00:13:56,640 --> 00:14:00,231 you know, GPT5 is probably smarter than most people and they seem entirely 296 00:14:00,314 --> 00:14:03,602 unbothered. I feel pretty unbothered even though I'm like, "Oh, you know, 297 00:14:03,685 --> 00:14:05,357 I just had said host or whatever." Um, 298 00:14:05,440 --> 00:14:10,915 I so I don't think we have any of those crisis of meanings in quite the way 299 00:14:10,998 --> 00:14:14,477 that or I'm I'm less convinced. Maybe we do. Future's hard to predict. 300 00:14:14,560 --> 00:14:16,557 I'm less convinced than other people are, but the 301 00:14:16,640 --> 00:14:19,988 economy feels like it should just absolutely radically transform and I you 302 00:14:20,071 --> 00:14:20,637 know maybe the 303 00:14:20,720 --> 00:14:24,742 whole like maybe the structure of it changes in a pretty big way. 304 00:14:24,825 --> 00:14:26,637 Any uh predictions or thoughts 305 00:14:26,720 --> 00:14:30,850 on your end on this Ali? I think you know Sam has you know he he's the ones who 306 00:14:30,933 --> 00:14:31,917 touched on are those 307 00:14:32,000 --> 00:14:36,011 are excellent. I'll give boring enterprise you know what I see in my kind of vicinity 308 00:14:36,095 --> 00:14:36,397 which is 309 00:14:36,480 --> 00:14:40,365 uh you know I think uh you know there's like this whole um you know in sales 310 00:14:40,448 --> 00:14:41,357 for instance in B2B 311 00:14:41,440 --> 00:14:46,968 organizations there's sales engineering you know but I think that's actually even 312 00:14:47,052 --> 00:14:48,077 more than R&D uh 313 00:14:48,160 --> 00:14:52,211 this kind of technology agents is applicable there uh the marketing function 314 00:14:52,294 --> 00:14:53,517 inside of a B2B function 315 00:14:53,600 --> 00:14:57,043 like there's so much that the agents already doing at data bricks like you know 316 00:14:57,126 --> 00:14:57,757 in the marketing 317 00:14:57,840 --> 00:15:03,813 funnel like preparing u campaigns and you know preparing the material that's for 318 00:15:03,896 --> 00:15:07,523 the different audiences all that that's already being done by agents. 319 00:15:07,606 --> 00:15:09,357 Uh so I think we're going to see that 320 00:15:09,440 --> 00:15:12,935 uh change a lot of the inside of an enterprise like data bricks there's so 321 00:15:13,018 --> 00:15:13,757 many different op 322 00:15:13,840 --> 00:15:17,507 ops functions and the people that basically are analysts they're very smart 323 00:15:17,590 --> 00:15:18,557 people with financial 324 00:15:18,640 --> 00:15:22,483 backgrounds you know they're typically sitting in excel every day um a lot 325 00:15:22,567 --> 00:15:23,757 of that is already being 326 00:15:23,840 --> 00:15:26,950 completely transformed and we're going to see that actually be even more 327 00:15:27,033 --> 00:15:27,837 revolutionized so at 328 00:15:27,920 --> 00:15:31,805 least when I look at inside of a company like data bricks it's pretty much every 329 00:15:31,888 --> 00:15:32,797 function is going to 330 00:15:32,880 --> 00:15:36,372 be completely transformed that doesn't answer the society question but but you 331 00:15:36,455 --> 00:15:37,197 know there's a lot 332 00:15:37,280 --> 00:15:40,411 of enterprises out there that kind of a little bit look like data bricks and I 333 00:15:40,495 --> 00:15:41,277 think already they're 334 00:15:41,360 --> 00:15:45,224 they're going to be vastly different in five years or maybe three. Yeah. Yeah. 335 00:15:45,308 --> 00:15:46,237 And maybe you know I 336 00:15:46,320 --> 00:15:50,345 know we talked about 5 10 years from now but what are the most exciting enterprise 337 00:15:50,428 --> 00:15:51,197 AI use cases that 338 00:15:51,280 --> 00:15:56,367 that you both see today. I mean there's there's so many uh and you know it starts 339 00:15:56,450 --> 00:15:57,197 like it we're 340 00:15:57,280 --> 00:16:00,609 taking on the smaller tasks. So like when you zoom in it might not look like 341 00:16:00,692 --> 00:16:04,040 that exciting but it's like what Sam said with the task horizons. 342 00:16:04,124 --> 00:16:07,357 Uh but you know like one one one thing that I thought was 343 00:16:07,440 --> 00:16:11,174 interesting. Astroenica had a use case with agents that basically sift through 344 00:16:11,257 --> 00:16:14,923 400,000 documents which is something no human could have done. Yeah. 345 00:16:15,006 --> 00:16:16,797 Uh and it's uh you know our in in 346 00:16:16,880 --> 00:16:19,889 finance we have a lot of financial organizations that are now going through 347 00:16:19,973 --> 00:16:20,797 all of the SEC filings 348 00:16:20,880 --> 00:16:25,133 all the related documents and they're gleaning out alpha for investments 349 00:16:25,216 --> 00:16:28,253 that then they're giving to analysts that are then using them. 350 00:16:28,336 --> 00:16:30,397 Humans that are using them but it's just this 351 00:16:30,480 --> 00:16:34,330 would be extremely tedious. human would have to do that before and now it's just 352 00:16:34,413 --> 00:16:35,117 you know we take 353 00:16:35,200 --> 00:16:37,652 for granted it's like yeah of course sift through all those thousands of pages 354 00:16:37,735 --> 00:16:38,237 and give give back 355 00:16:38,320 --> 00:16:41,994 the response quickly to me. uh in insurance we're seeing it you know 356 00:16:42,078 --> 00:16:43,597 underwriting again like in in 357 00:16:43,680 --> 00:16:46,886 in the healthcare space there's if you think about if you go to a hospital 358 00:16:46,969 --> 00:16:47,997 there's like thousands of 359 00:16:48,080 --> 00:16:52,008 thousands of pages of papers that you have to sign that are produced for every 360 00:16:52,091 --> 00:16:53,037 hospital visit and I 361 00:16:53,120 --> 00:16:57,675 don't know if anyone ever even looks up those uh but now in the healthcare sector 362 00:16:57,758 --> 00:16:58,477 they're having 363 00:16:58,560 --> 00:17:01,520 agents more and more companies and startups are having actually all of 364 00:17:01,603 --> 00:17:02,477 that being fed to LLMs 365 00:17:02,560 --> 00:17:07,909 that are going through it and you know extracting tidbits figuring out risk uh so 366 00:17:07,993 --> 00:17:09,117 so much of society 367 00:17:09,200 --> 00:17:12,281 is going to be changed uh but it's going to happen you know as you said with those 368 00:17:12,364 --> 00:17:12,404 you 369 00:17:12,480 --> 00:17:16,460 know simpler and simpler tasks that gets more and more complicated and you know 370 00:17:16,543 --> 00:17:17,437 over time increases 371 00:17:17,520 --> 00:17:23,030 I I super agree with that answer but just to add two other directions I find very 372 00:17:23,113 --> 00:17:23,997 interesting um 373 00:17:24,080 --> 00:17:27,453 obviously it's great that people are doing things you know faster and better 374 00:17:27,536 --> 00:17:28,317 and cheaper but the 375 00:17:28,400 --> 00:17:32,101 categories of things that you just wouldn't do at all before AI or and you 376 00:17:32,184 --> 00:17:33,277 already touched on this 377 00:17:33,360 --> 00:17:37,362 but things that you just simply couldn't do are are both interesting to me so 378 00:17:37,445 --> 00:17:41,165 there's more focus on stuff you couldn't do like read 4,000 pages of documents. 379 00:17:41,248 --> 00:17:42,637 And I think that's amazing that 380 00:17:42,720 --> 00:17:47,417 AI is just now churning through this. And um it's so superhuman at literature search 381 00:17:47,500 --> 00:17:48,157 as a specific 382 00:17:48,240 --> 00:17:53,086 example of what that's meant. Um but this wouldn't do category just doesn't get 383 00:17:53,170 --> 00:17:54,397 as much attention. So 384 00:17:54,480 --> 00:17:58,631 if you I use coding again because it's the kind of thing I have the best mental 385 00:17:58,714 --> 00:17:59,757 personal mental model 386 00:17:59,840 --> 00:18:05,090 for. Um coders they can do their work faster but they also there's just features 387 00:18:05,173 --> 00:18:06,157 they would never 388 00:18:06,240 --> 00:18:08,806 think they would never try. that that would be at the bottom of someone's 389 00:18:08,889 --> 00:18:11,855 forever. I will never do do that. I'm just not ever going to get there. 390 00:18:11,938 --> 00:18:14,317 And now if I can just like kick off one extra agent, 391 00:18:14,400 --> 00:18:18,652 I am willing to do that. So seeing there's just like there were these good ideas 392 00:18:18,735 --> 00:18:22,612 that you just wouldn't do. And if you make the activation energy low enough, 393 00:18:22,695 --> 00:18:24,397 you make it cheap enough, you just 394 00:18:24,480 --> 00:18:27,769 you say, "Okay, fine. I'll try it." That's been like an interesting thing for me 395 00:18:27,852 --> 00:18:28,317 to see. Yeah, 396 00:18:28,400 --> 00:18:33,919 that's awesome. Yeah. Yeah. So yeah, the ability to launch async parallel basically 397 00:18:34,002 --> 00:18:34,797 agentic jobs, 398 00:18:34,880 --> 00:18:38,747 right? uh really changes how you think about the process of actually doing work 399 00:18:38,830 --> 00:18:39,597 which is which is 400 00:18:39,680 --> 00:18:43,301 very exciting. Yes. I mean it's uh it's so cool because it's so fast and you can do 401 00:18:43,384 --> 00:18:43,837 so much with 402 00:18:43,920 --> 00:18:47,527 it but now I feel like I'm so frustrated like why is it not why is it not like why 403 00:18:47,610 --> 00:18:50,117 it's taking so long like a minute this to write 20,000 lines. 404 00:18:50,200 --> 00:18:52,077 You know this is an area that I think we as an 405 00:18:52,160 --> 00:18:57,619 industry are underinvested in is how how we make this stuff dramatically faster. 406 00:18:57,702 --> 00:19:00,872 Yeah. 100%. We're going to push on that. Yeah. Exciting. Yeah. 407 00:19:00,955 --> 00:19:03,277 I mean we've done this in our database why Lindy is 408 00:19:03,360 --> 00:19:07,169 king right for any interactive experience whether it be a dashboard and the same 409 00:19:07,253 --> 00:19:07,997 principle applies 410 00:19:08,080 --> 00:19:11,988 to agents as well industry took a bet that that uh which I think was the right bet if 411 00:19:12,071 --> 00:19:12,317 you had 412 00:19:12,400 --> 00:19:15,674 to do one or the other which is we were going to prioritize cost per token over 413 00:19:15,758 --> 00:19:16,397 you know token to 414 00:19:16,480 --> 00:19:20,764 token latency and design most hardware that way there's clearly demand for a much 415 00:19:20,847 --> 00:19:21,357 higher cost 416 00:19:21,440 --> 00:19:24,922 per token at a much lower latency interesting is that now we're getting 417 00:19:25,005 --> 00:19:26,077 peaks into the hardware 418 00:19:26,160 --> 00:19:30,119 you're building the hardware that we'll eventually build not the first thing we'll 419 00:19:30,202 --> 00:19:33,752 do but We want to we want to do that at some Yeah. Yeah. Great. 420 00:19:33,836 --> 00:19:36,797 Uh we have a lot of uh folks dialing in here from 421 00:19:36,880 --> 00:19:41,677 uh you know where data scientists or data engineers or some even cos and leaders at 422 00:19:41,760 --> 00:19:46,816 at these enterprise companies. Um with this sort of generative AI revolution 423 00:19:46,900 --> 00:19:48,237 maybe what's like one 424 00:19:48,320 --> 00:19:52,730 thing that you know CEO or leader should do like immediately now to like prepare 425 00:19:52,814 --> 00:19:53,517 yourself right 426 00:19:53,600 --> 00:19:58,010 for this wave that's already here and it's like continuing to to roll to roll 427 00:19:58,093 --> 00:19:58,477 forward. 428 00:19:58,560 --> 00:20:03,569 I think uh you know I'll start with the boring stuff because it's uh it's people 429 00:20:03,652 --> 00:20:04,397 people um you 430 00:20:04,480 --> 00:20:08,127 know they're excited about you know what they can accomplish with AI which is great 431 00:20:08,210 --> 00:20:08,397 uh but 432 00:20:08,480 --> 00:20:11,108 before you get there you have to have a good data foundation and you have to make 433 00:20:11,191 --> 00:20:11,677 sure that you can 434 00:20:11,760 --> 00:20:15,390 secure it and you can make sure that you have the right guard guardrails in place 435 00:20:15,473 --> 00:20:18,518 and then also getting access to as much of your data in the enterprise. 436 00:20:18,601 --> 00:20:20,077 there's a lot of data that's siloed 437 00:20:20,160 --> 00:20:24,514 and tucked away like that's sitting in a database that's not accessible or that's 438 00:20:24,597 --> 00:20:25,117 you know um 439 00:20:25,200 --> 00:20:28,782 actually not even this is on premise it's not in the cloud and so on so making sure 440 00:20:28,866 --> 00:20:29,357 that you just 441 00:20:29,440 --> 00:20:34,409 can get all of that in one place uh so that it's accessible as context to the AI 442 00:20:34,493 --> 00:20:35,357 will be kind of 443 00:20:35,440 --> 00:20:38,489 like that differential whether the AI immediately gets it right whether it's 444 00:20:38,572 --> 00:20:39,437 within the task horizon 445 00:20:39,520 --> 00:20:43,296 or whether it's uh you know it's just trying many many things because you just 446 00:20:43,380 --> 00:20:44,237 didn't give it that 447 00:20:44,320 --> 00:20:47,659 little context that's sitting tucked away in that silo data set that's exactly what 448 00:20:47,743 --> 00:20:48,576 I was going to 449 00:20:49,200 --> 00:20:52,205 I think that's I think that's it sounds so boring but it's so important to get 450 00:20:52,288 --> 00:20:52,957 in this work. Yeah. 451 00:20:53,040 --> 00:20:57,321 Yeah. And and you know we've seen sort of natural language is great but it's also 452 00:20:57,404 --> 00:20:58,237 ambiguous. So you 453 00:20:58,320 --> 00:21:03,597 know when when our Ali knows this or when our customers ask hey please compute the 454 00:21:03,680 --> 00:21:05,716 uh churn for me. Well what is what is churn? 455 00:21:05,800 --> 00:21:08,077 Like it's it's really dependent on the enterprise 456 00:21:08,160 --> 00:21:11,566 context and and the enterprise use. Yeah it is I mean it is that definition exists 457 00:21:11,649 --> 00:21:12,077 somewhere in 458 00:21:12,160 --> 00:21:16,299 the organization in a company. how you define revenue or how you define churn 459 00:21:16,382 --> 00:21:16,957 or what when 460 00:21:17,040 --> 00:21:21,714 a fiscal year starts or what the quarters are um and you know so it's just can you 461 00:21:21,798 --> 00:21:22,237 find that 462 00:21:22,320 --> 00:21:26,400 as a context is is that is the AI connected to that if it is it will be able 463 00:21:26,483 --> 00:21:27,277 to figure it out 464 00:21:27,360 --> 00:21:30,917 because the models are so freaking smart now and they're getting smarter every 465 00:21:31,000 --> 00:21:31,757 release that comes 466 00:21:31,840 --> 00:21:34,973 u they're getting smarter and smarter so we just have to make sure we can feed 467 00:21:35,056 --> 00:21:35,757 that context though 468 00:21:35,840 --> 00:21:39,689 there's a little bit debate raging now on like whether these agent patterns are just 469 00:21:39,772 --> 00:21:40,157 agents and 470 00:21:40,240 --> 00:21:44,809 tools in a loop or whether you got to handgineer all the individual uh pieces 471 00:21:44,893 --> 00:21:49,328 of the agents and wire them into some deterministic kind of kind of workflow. 472 00:21:49,411 --> 00:21:51,517 the models what you see if models can 473 00:21:51,600 --> 00:21:54,448 like do novel math they're certainly smart enough to figure out how to do this out 474 00:21:54,531 --> 00:21:54,877 without hand 475 00:21:54,960 --> 00:21:58,491 engineering may take us a little bit of time and again you do need in the same way 476 00:21:58,575 --> 00:21:59,197 that an employee 477 00:21:59,280 --> 00:22:01,997 needs some training needs to know where to look where the enterprise stuff is you'll 478 00:22:02,080 --> 00:22:02,397 have to like 479 00:22:02,480 --> 00:22:06,044 make that knowledge available but they will not need custom training for it yeah 480 00:22:06,127 --> 00:22:06,637 totally agree 481 00:22:06,720 --> 00:22:10,546 isn't a bitter lesson uh it is it is a bit lesson bitter lesson you know it's like 482 00:22:10,629 --> 00:22:11,357 and we've seen it 483 00:22:11,440 --> 00:22:14,866 with most of the agents that people are building the successful agents when when 484 00:22:14,950 --> 00:22:15,437 you when when 485 00:22:15,520 --> 00:22:18,583 the people built them talk about it's like hey I actually did the simplest thing I 486 00:22:18,666 --> 00:22:19,197 didn't have lots 487 00:22:19,280 --> 00:22:23,149 of lots of lots of agents that call each other and lots of hand tune things. 488 00:22:23,232 --> 00:22:24,397 I just let the AI figure 489 00:22:24,480 --> 00:22:30,054 it out. Yeah. Yeah. Great. Um, thank you both for uh joining us today. 490 00:22:30,138 --> 00:22:35,018 We really really appreciate it. Uh, for the folks who are dial in here, you know, 491 00:22:35,101 --> 00:22:37,117 if you want to get ready and get started 492 00:22:37,200 --> 00:22:41,721 uh using uh open a on data bricks, uh please check out uh AI uh playground 493 00:22:41,804 --> 00:22:43,277 and also our agent bricks 494 00:22:43,360 --> 00:22:47,714 product uh to get starting building agents uh with open AAI and the GBT5 models 495 00:22:47,797 --> 00:22:48,640 on data bricks. 496 00:22:49,200 --> 00:22:52,950 Thank you both for being here. Thanks so much. Great discussion. Thank you.