WEBVTT - Smart Talks with IBM: Generative AI: Its Rise and Potential for Society 0:00:00.120 --> 0:00:02.840 Hey everyone, it's Robert and Joe here. Today we've got 0:00:02.840 --> 0:00:04.720 something a little bit different to share with you. It 0:00:04.840 --> 0:00:08.000 is a new season of the Smart Talks with IBM 0:00:08.119 --> 0:00:09.119 podcast series. 0:00:09.600 --> 0:00:11.680 Today we are witnessed to one of those rare moments 0:00:11.680 --> 0:00:14.360 in history, the rise of an innovative technology with the 0:00:14.360 --> 0:00:18.680 potential to radically transform business and society forever. The technology, 0:00:18.760 --> 0:00:22.200 of course, is artificial intelligence, and it's the central focus 0:00:22.239 --> 0:00:24.800 for this new season of Smart Talks with IBM. 0:00:25.320 --> 0:00:28.400 Join hosts from your favorite Pushkin podcasts as they talk 0:00:28.480 --> 0:00:31.640 with industry experts and leaders to explore how businesses can 0:00:31.680 --> 0:00:35.360 integrate AI into their workflows and help drive real change 0:00:35.400 --> 0:00:38.200 in this new era of AI. And of course, host 0:00:38.280 --> 0:00:40.440 Malcolm Gladwell will be there to guide you through the 0:00:40.479 --> 0:00:42.640 season and throw in his two cents as well. 0:00:43.080 --> 0:00:46.120 Look out for new episodes of Smart Talks with IBM 0:00:46.400 --> 0:00:49.519 every other week on the iHeartRadio app, Apple Podcasts, or 0:00:49.560 --> 0:00:53.360 wherever you get your podcasts, and learn more at IBM 0:00:53.479 --> 0:00:55.480 dot com slash smart Talks. 0:00:57.760 --> 0:01:03.920 All right, Welcome everybody, you guys excited, here we go. 0:01:04.280 --> 0:01:07.800 Hello, Hello, Welcome to Smart Talks with IBM, a podcast 0:01:07.800 --> 0:01:13.679 from Pushkin Industries. iHeartRadio and IBM. I'm Malcolm Gladwell. This season, 0:01:13.680 --> 0:01:17.680 we're continuing our conversations with new creators visionaries who are 0:01:17.720 --> 0:01:21.240 creatively applying technology and business to drive change, but with 0:01:21.319 --> 0:01:25.640 a focus on the transformative power of artificial intelligence and 0:01:25.680 --> 0:01:28.559 what it means to leverage AI as a game changing 0:01:28.680 --> 0:01:33.319 multiplier for your business. Today's episode is a bit different. 0:01:33.680 --> 0:01:37.560 I was recently joined on stage by Dario Gill for 0:01:37.640 --> 0:01:40.000 a conversation in front of a live audience at the 0:01:40.000 --> 0:01:44.120 iHeartMedia headquarters in Manhattan. Dario is the senior vice president 0:01:44.319 --> 0:01:47.840 and director of IBM Research, one of the world's largest 0:01:47.960 --> 0:01:52.240 and most influential corporate research labs. We discussed the rise 0:01:52.280 --> 0:01:55.760 of generative AI and what it means for business and society. 0:01:56.080 --> 0:01:59.560 He also explained how organizations that leverage AI to create 0:01:59.640 --> 0:02:04.360 value will dominate in the near future. Okay, let's get 0:02:04.360 --> 0:02:10.920 on to the conversation. Hello everyone, welcome, and I'm here 0:02:11.000 --> 0:02:14.640 with doctor Dario Gil and I wanted to say before 0:02:14.639 --> 0:02:16.280 we get started. This is something I said backstage that 0:02:16.760 --> 0:02:23.000 I feel very guilty today because you're the you know, 0:02:23.760 --> 0:02:25.800 you know, arguably one of the most important figures in 0:02:26.160 --> 0:02:28.920 AI research in the world, and we have taken you 0:02:28.960 --> 0:02:33.240 away from your job for a morning. It's like, if 0:02:33.600 --> 0:02:37.440 you know Oppenheimer's wife in nineteen forty four said let's 0:02:37.760 --> 0:02:41.600 go and have a little getaway in the Bahamas. It's 0:02:41.600 --> 0:02:43.400 that kind of thing. You know, what do you say 0:02:43.440 --> 0:02:46.519 to your wife, I can't we have got to work 0:02:46.520 --> 0:02:48.800 on this thing I can't tell you about. She's like 0:02:48.840 --> 0:02:51.200 getting me out of Los Alamos. No, So I do 0:02:51.240 --> 0:02:56.160 feel guilty. We've set back AI research by by about 0:02:56.160 --> 0:03:01.160 four hours here. But I wanted to you've been up 0:03:01.160 --> 0:03:04.400 with with ibo for twenty years, twenty years this summer. 0:03:04.800 --> 0:03:06.400 So and how old were you when you Not to 0:03:06.440 --> 0:03:08.120 give away your age, but you were how old when 0:03:08.160 --> 0:03:08.640 you started? 0:03:08.919 --> 0:03:09.679 I was twenty eight? 0:03:09.720 --> 0:03:11.480 Okay, yeah, So I want to go back to your 0:03:11.480 --> 0:03:14.400 twenty eight year old self. Now, if I asked you 0:03:14.440 --> 0:03:18.040 about artificial intelligence, I asked twenty eight year old Dario, 0:03:18.919 --> 0:03:22.200 what does the future hold for AI? How quickly will 0:03:22.639 --> 0:03:26.080 this new technology transform our world? Et cetera, et cetera. 0:03:26.160 --> 0:03:28.160 What would twenty eight year old Darigo said? 0:03:28.639 --> 0:03:30.880 Well, I think the first thing is that even though 0:03:30.919 --> 0:03:33.280 AI as a feel has been with us for a 0:03:33.280 --> 0:03:36.360 long time, since the mid nineteen fifties, at that time, 0:03:36.800 --> 0:03:40.240 AI was not a very polite word to say, meaning 0:03:40.320 --> 0:03:44.040 within the scientific community, people didn't use sort of that term. 0:03:44.080 --> 0:03:45.960 They would have said things like, you know, maybe I 0:03:46.080 --> 0:03:50.240 do things relate to machine learning, right, or statistical techniques 0:03:50.280 --> 0:03:53.400 in terms of classifiers and so on. But AI had 0:03:53.440 --> 0:03:56.520 a mixed reputation, right, it had gone through different cycles 0:03:56.520 --> 0:04:00.800 of hype, and it's also if moments of you know, 0:04:00.880 --> 0:04:05.760 a lot of negativity towards it because of lack of success. 0:04:05.920 --> 0:04:07.600 And so I think that would be the first thing 0:04:07.600 --> 0:04:10.440 we probably say, like AI is like what is that? Like, 0:04:10.480 --> 0:04:14.120 you know, respectable scientists are not working on AI the 0:04:14.200 --> 0:04:18.000 finest side and that really changed over the last fifteen years. Only, right, 0:04:18.240 --> 0:04:20.839 I would say, with the advent of deep learning over 0:04:20.880 --> 0:04:23.320 the last decade, is when that re enter again the 0:04:23.400 --> 0:04:26.240 lexicon of saying AI and that that was a legitimate 0:04:26.240 --> 0:04:28.920 thing to work on. So I would say that that's 0:04:28.960 --> 0:04:30.440 the first thing I think we would have noticed a 0:04:30.440 --> 0:04:31.480 contrast twenty years ago. 0:04:31.760 --> 0:04:35.640 Yeah, So what point in your twenty year tenure at 0:04:35.839 --> 0:04:39.599 IBM would you say you kind of snapped into present 0:04:40.240 --> 0:04:41.520 kind of wow mode. 0:04:42.880 --> 0:04:50.800 I would say in a late two thousands, when IBM 0:04:51.600 --> 0:04:56.880 was working on the Jeopardy project and just seeing the 0:04:56.920 --> 0:05:00.400 demonstrations of what could be done in question answering. 0:05:00.800 --> 0:05:04.000 It's literally Jeopardy. Is this crucial moment in the history 0:05:04.000 --> 0:05:04.560 of YEA. 0:05:05.240 --> 0:05:07.640 You know, there had been a long and wonderful history 0:05:08.320 --> 0:05:12.080 inside IBM on AI. So for example, like you know, 0:05:12.160 --> 0:05:15.000 in terms of like these grand challenges at the very 0:05:15.040 --> 0:05:18.000 beginning of the field founding, which is this famous Dartmouth 0:05:18.000 --> 0:05:22.440 conference that actually IBM sponsored h to create. There was 0:05:22.720 --> 0:05:27.000 an IBM and there called Nathaniel Rochester, and there were 0:05:27.040 --> 0:05:30.040 a few others who right after that they started thinking 0:05:30.040 --> 0:05:33.040 about demonstrations of this field. And then for example, they 0:05:33.080 --> 0:05:36.479 created the first you know game to play checkers and 0:05:36.560 --> 0:05:39.719 to demonstrate that you could do machine learning on that. 0:05:40.400 --> 0:05:43.039 Obviously we saw later in the nineties like chess that 0:05:43.120 --> 0:05:45.719 was a very famous example of that, Deep Blue with 0:05:45.800 --> 0:05:49.039 Deep Blue right and playing with Caspar and then but 0:05:49.120 --> 0:05:51.720 I think the moment that was really those other ones 0:05:51.760 --> 0:05:54.240 felt like, you know, kind of like brute force anticipating, 0:05:54.279 --> 0:05:56.679 sort of like moves ahead. But this aspect of dealing 0:05:56.680 --> 0:06:00.960 with language and question answering felt different, and I think 0:06:01.040 --> 0:06:03.599 for for us internally and many others, was when a 0:06:03.600 --> 0:06:06.000 moment of saying like, wow, you know, what are the 0:06:06.040 --> 0:06:09.440 possibilities here? And then soon after that connected to the 0:06:09.480 --> 0:06:12.880 sort of advancements in computing and with deep learning. The 0:06:12.960 --> 0:06:15.200 last decade has just been an all out, you know, 0:06:15.279 --> 0:06:17.440 sort of like front of advancements and that, and I 0:06:17.520 --> 0:06:19.560 just continue to be more and more impressed. And the 0:06:19.640 --> 0:06:21.360 last few years have been remarkable too. 0:06:21.560 --> 0:06:26.040 Yeah. So I'll ask you three quick conceptual questions before 0:06:26.040 --> 0:06:27.880 we dig into it, just so I sort of get 0:06:27.880 --> 0:06:32.120 a we all get a feel for the shape of AI. 0:06:32.920 --> 0:06:36.960 Question Number one is where are we in the evolution 0:06:37.080 --> 0:06:41.160 of this? So you know the obvious question. We we're 0:06:41.160 --> 0:06:43.760 all suddenly aware of it, we're talking about it. Can 0:06:43.800 --> 0:06:45.800 you give us an analogy about where we are in 0:06:45.880 --> 0:06:50.680 the kind of likely evolution of this is a technology. 0:06:50.720 --> 0:06:54.440 So I think we're on a significant inflection point that 0:06:54.839 --> 0:06:58.920 it feels the equivalent of the first browsers when they 0:06:58.960 --> 0:07:03.200 appear and people imagine the possibilities of the Internet or 0:07:03.240 --> 0:07:07.120 more imagined experience the internet. The Internet had been around, 0:07:07.320 --> 0:07:09.679 right for quite a few decades. AI has been around 0:07:09.920 --> 0:07:12.560 for many decades. I think the moment we find ourselves 0:07:12.680 --> 0:07:15.760 is that people can touch it and they can Before 0:07:15.800 --> 0:07:17.920 they were a systems that were like behind the scenes, 0:07:18.000 --> 0:07:22.120 like your search results or translation systems, but they didn't 0:07:22.160 --> 0:07:24.040 have the experience of like, this is what it feels 0:07:24.080 --> 0:07:27.040 like to interact with this thing. So that's what I mean. 0:07:27.080 --> 0:07:29.200 I think maybe that analogy of the browser is appropriate 0:07:29.240 --> 0:07:31.440 because it's all of a sudden, it's like whoa, you know, 0:07:31.640 --> 0:07:35.240 these network of machines and content can be distributed and 0:07:35.280 --> 0:07:37.920 everybody can self publish, and there was a moment that 0:07:37.960 --> 0:07:40.080 we all remember that, and I think that that is 0:07:40.200 --> 0:07:42.640 what the world has experience over the last nine months 0:07:42.720 --> 0:07:46.520 or so on. So but fundamentally, also what is important 0:07:46.600 --> 0:07:48.760 is that this moment is where the ease of the 0:07:48.840 --> 0:07:53.440 number of people that can build and use AI has skyrocketed. 0:07:54.000 --> 0:07:58.360 So over the last decade, you know, technology firms that 0:07:58.480 --> 0:08:03.120 had large research teams could build AI that worked really well, honestly, 0:08:03.560 --> 0:08:06.400 but when you went down into say hey can everybody 0:08:06.520 --> 0:08:08.800 use it? Can a data science team in a bank, 0:08:09.000 --> 0:08:11.640 you know, go and develop these applications, it was like 0:08:11.800 --> 0:08:13.960 more complicated. Some could do it, but it was more 0:08:14.040 --> 0:08:16.560 the barrier of entry was high. Now is very different 0:08:16.920 --> 0:08:20.200 because of foundation models and the implications that that has. 0:08:20.080 --> 0:08:23.720 For at the moment where the technology is being democratized. 0:08:23.080 --> 0:08:28.560 In demarketized, frankly, it works better for classes of problems 0:08:28.600 --> 0:08:31.200 like programming and other things. Is really incredibly impressive what 0:08:31.240 --> 0:08:33.600 it can do. So the accuracy and the performance of 0:08:33.640 --> 0:08:36.560 it is much better, and the ease of use and 0:08:36.600 --> 0:08:38.760 the number of use cases we can pursue it much bigger, 0:08:38.840 --> 0:08:40.560 So that democratization is a big difference. 0:08:40.559 --> 0:08:42.440 But when you say, when you make it an analogy 0:08:42.480 --> 0:08:46.760 to the first browsers, if we do another one of 0:08:46.840 --> 0:08:49.520 these time travel questions back at the beginning of the 0:08:49.520 --> 0:08:53.240 first browsers, it's safe to say many of the potential 0:08:53.440 --> 0:08:56.800 uses of the Internet and such we hadn't even begun. 0:08:56.880 --> 0:08:59.439 We couldn't even anticipate, right, Right, So we're at the 0:08:59.480 --> 0:09:02.760 point where the future direction is largely unpredictable. 0:09:03.000 --> 0:09:05.800 Yeah, I think that that is right. Because it's such 0:09:05.800 --> 0:09:10.199 a horizontal technology that the intersection of the horizontal capability, 0:09:10.280 --> 0:09:14.520 which is about expanding our productivity and tasks that we 0:09:14.559 --> 0:09:17.360 wouldn't be able to do efficiently without it, has to 0:09:17.440 --> 0:09:20.240 marry now the use cases that reflect the diversity of 0:09:20.320 --> 0:09:23.360 human experience, our institutional diversity. So as more and more 0:09:23.400 --> 0:09:26.000 institutions said, you know, I'm focused on agriculture, you know, 0:09:26.160 --> 0:09:29.240 to be able to improve seeds. You know, in these 0:09:29.320 --> 0:09:32.280 kinds of environments, they'll find their own context in which 0:09:32.320 --> 0:09:34.120 that matters that the creators of a I did not 0:09:34.160 --> 0:09:37.080 anticipate at the beginning. So I think that that is 0:09:37.120 --> 0:09:39.520 then the fruit of surprises will be like why I 0:09:39.559 --> 0:09:41.319 wouldn't even think that it could be used for that. 0:09:41.559 --> 0:09:44.760 And also clever people will create new business models as 0:09:44.880 --> 0:09:47.520 associated with that, like it happened with the Internet of 0:09:47.559 --> 0:09:50.439 course as well, and that will be its own source 0:09:50.480 --> 0:09:52.880 of transformation and change in its own right. So I 0:09:52.920 --> 0:09:55.040 think all of that is yet to unfold. Right, what 0:09:55.160 --> 0:09:58.040 we're seeing is this catalyst moment of technology that works 0:09:58.040 --> 0:09:59.760 well enough and it can be democratized. 0:10:00.080 --> 0:10:03.840 Yeah, what next sort of conceptual question? You know, we 0:10:03.840 --> 0:10:10.320 can loosely understand or categorize innovations in terms of their 0:10:10.360 --> 0:10:15.040 impact on the kind of balance of power between haves 0:10:15.080 --> 0:10:20.000 and have nots. Some innovations, you know, obviously favor those 0:10:20.040 --> 0:10:24.560 who already have make the rich richer. Some some it's 0:10:24.559 --> 0:10:28.319 arising to tie the lifts all boats, and some bias 0:10:28.320 --> 0:10:32.040 in the other direction. They close the gap between is 0:10:32.080 --> 0:10:35.440 it possible to say to predict which of those three 0:10:35.480 --> 0:10:37.200 categories AI might fall into. 0:10:38.240 --> 0:10:41.880 It's a great question, you know. A first observation I 0:10:41.920 --> 0:10:46.080 would make on your first two categories is that it 0:10:46.160 --> 0:10:49.680 will be both likely be true that the use of 0:10:49.720 --> 0:10:52.280 AI will be highly democratized, meaning the number of people 0:10:52.280 --> 0:10:55.400 that have access to its power to make improvements in 0:10:55.440 --> 0:10:58.239 terms of efficiency and so on will be fairly universal, 0:10:59.040 --> 0:11:02.400 and that the ones who are able to create AI 0:11:03.200 --> 0:11:06.400 uh may be quite concentrated. So if you look at 0:11:06.400 --> 0:11:10.000 it from the lens of who creates wealth and value 0:11:10.400 --> 0:11:14.400 over sustained periods of time, particularly it's saying a context 0:11:14.440 --> 0:11:17.920 like business, I think just being a user of AI 0:11:18.040 --> 0:11:22.280 technology is an insufficient strategy and UH. And the reason 0:11:22.360 --> 0:11:24.200 for that is like, yes, you will get the immediate 0:11:24.240 --> 0:11:27.200 productivity boost of like just making API calls, and you 0:11:27.200 --> 0:11:29.840 know that would be a new baseline for everybody, but 0:11:30.280 --> 0:11:34.040 you're not accruing value in terms of representing your data 0:11:34.160 --> 0:11:36.640 inside the AI in way that gives you a sustainable 0:11:36.640 --> 0:11:40.000 competitive advantage. So I always try to tell people is 0:11:40.040 --> 0:11:42.080 don't just be an AI user. We an you know, 0:11:42.120 --> 0:11:45.640 AI value creator. And I think that that will have 0:11:45.720 --> 0:11:48.560 a lot of consequences in terms of the haves and 0:11:48.640 --> 0:11:51.160 have nots as an example, and that will apply both 0:11:51.160 --> 0:11:54.960 to institutions and regions and countries, et cetera. So I 0:11:54.960 --> 0:11:57.480 think it would be kind of a mistake, right to 0:11:57.760 --> 0:12:00.280 just develop strategies that are just about. 0:12:00.160 --> 0:12:04.240 Usage, but to to come back that question from them 0:12:04.440 --> 0:12:08.240 to give you a specific suppose I'm a I'm an 0:12:08.280 --> 0:12:12.760 industrial farmer in Iowa with ten million dollars of equipment 0:12:12.840 --> 0:12:16.400 and move and I'm comparing it to a subsistence farmer 0:12:16.840 --> 0:12:20.160 someone in the developing world who's got a cell phone. 0:12:20.920 --> 0:12:25.640 Over the next five years, who's who's well being rises 0:12:25.720 --> 0:12:27.120 by a greater amount. 0:12:28.160 --> 0:12:30.880 Yeah, I think, I mean, it's a it's a good question, 0:12:30.960 --> 0:12:32.559 but it might be hard to do a one to 0:12:32.679 --> 0:12:35.280 one sort of like attribution to just one variable in 0:12:35.280 --> 0:12:39.520 this case, which is AI. But again, provided that you 0:12:39.600 --> 0:12:42.440 have access to a phone, right and some kind to 0:12:42.679 --> 0:12:45.120 you know, be able to be connected. I do think 0:12:45.200 --> 0:12:48.200 so for example, in that context we've developed we don't 0:12:48.200 --> 0:12:50.720 work with NASA as an example, to build your spatial 0:12:50.760 --> 0:12:54.200 models using some of these new techniques, and I think 0:12:54.240 --> 0:12:57.160 for example, or ability to do flood prediction, I'll tell 0:12:57.160 --> 0:12:59.880 you an advantage of why would be a democratization force 0:13:00.080 --> 0:13:03.680 that context. Before to build a flowed model based on 0:13:03.760 --> 0:13:08.040 satellite imagery was actually so onerous and so complicated and 0:13:08.080 --> 0:13:10.680 difficult that you would just target to very specific regions 0:13:10.920 --> 0:13:14.080 and then obviously countries prioritize their own right. But what 0:13:14.120 --> 0:13:16.960 we've demonstrated is actually you can extend the technique to 0:13:17.000 --> 0:13:19.800 have like global coverage around that. So in that context, 0:13:19.840 --> 0:13:21.960 I would say it's a four stores and markeratization that 0:13:22.000 --> 0:13:25.439 everybody sort of would have access if you have some connectivity, 0:13:25.640 --> 0:13:26.359 as today. 0:13:26.080 --> 0:13:29.520 Iowa farmer might have a flood model. The guy in 0:13:29.559 --> 0:13:31.800 the developing world definitely didn't, and now he's a shot 0:13:31.800 --> 0:13:32.280 of getting one. 0:13:32.320 --> 0:13:33.800 Yeah, but now it has a shot of getting one. 0:13:33.840 --> 0:13:35.640 So there's aspects of it that so long as we 0:13:35.720 --> 0:13:38.680 provide connectivity and access to it, that they can be 0:13:38.840 --> 0:13:41.840 democratization forces. But I'll give you another example that that 0:13:41.840 --> 0:13:45.000 can be quite concerning, which is language. Right, So there's 0:13:45.160 --> 0:13:49.480 so much language in the you know, in English, and 0:13:49.960 --> 0:13:53.000 there is sort of like this reinforcement loop that happens 0:13:53.000 --> 0:13:55.480 that the more you concentrate, because it has obvious benefits 0:13:55.480 --> 0:13:59.120 for global communication and standardization, the more you can enrich 0:13:59.280 --> 0:14:02.400 like base aim models based on that capability. If you 0:14:02.520 --> 0:14:06.720 have very resource cars languages, you tend to develop less 0:14:06.760 --> 0:14:10.320 powerful AI with those languages and so on. So one 0:14:10.400 --> 0:14:14.600 has to actually worry and focus on the ability to 0:14:14.760 --> 0:14:18.120 actually represent you know that in that case is language 0:14:18.120 --> 0:14:20.920 as a piece of culture. Also in the AI sets 0:14:20.960 --> 0:14:24.000 that everybody can benefit from it too. So there's a 0:14:24.120 --> 0:14:27.640 lot of considerations in terms of equity about the data 0:14:27.680 --> 0:14:30.840 and the data sets that we accrue and what problems 0:14:30.880 --> 0:14:33.040 are we trying to solve. I mean, you mentioned agriculture 0:14:33.120 --> 0:14:35.600 or healthcare and so on. If we only solve problems 0:14:35.640 --> 0:14:38.320 that are related to marketing as an example, that would 0:14:38.320 --> 0:14:40.880 be a less rich world in terms of opportunity that 0:14:40.880 --> 0:14:43.760 if we incorporate many many other broad set of problems. 0:14:44.800 --> 0:14:46.680 Who do you think what do you think are the 0:14:46.720 --> 0:14:52.080 biggest impediments to the adoption of AI as you would 0:14:52.200 --> 0:14:54.600 like as you think AIR to be adopted. I mean, 0:14:54.640 --> 0:14:57.119 if you look, what are the sticking points that you would. 0:14:57.960 --> 0:15:00.480 Look Indiana, I'm going to give a non time technological 0:15:00.520 --> 0:15:03.080 answer as a first one has to do with workflow, right, 0:15:03.320 --> 0:15:07.920 So even if the technology is very capable, the organizational 0:15:08.040 --> 0:15:11.280 change inside a company to incorporate into the natural workflow 0:15:11.320 --> 0:15:14.840 of people and how we work is it's a lesson 0:15:14.880 --> 0:15:17.400 we have learned over the last decade is hugely important. 0:15:18.040 --> 0:15:21.720 So there's a lot of design considerations. There's a lot 0:15:21.760 --> 0:15:24.560 of how do people want to work right? How do 0:15:24.640 --> 0:15:27.000 they work today? And what is the natural entry points 0:15:27.000 --> 0:15:29.600 for AI? So that's like number one, and then the 0:15:29.680 --> 0:15:33.600 second one is, you know, for the broad value creation 0:15:33.720 --> 0:15:37.720 aspect of it is the understanding inside the companies of 0:15:37.960 --> 0:15:42.240 how you have to curate and create data to combine 0:15:42.240 --> 0:15:45.120 it with external data says that you can have powerful 0:15:45.160 --> 0:15:48.800 AI models that actually fit your need. And that aspect 0:15:48.880 --> 0:15:51.920 of what it takes to actually create and curate the 0:15:52.040 --> 0:15:56.920 data for this modern AI, it's still working progress, right. 0:15:57.240 --> 0:15:59.680 I think part of the problem that happens very often 0:15:59.720 --> 0:16:03.040 when I talk to institutions is that they say yea yeah, yah, yeah, 0:16:03.040 --> 0:16:05.600 I'm doing it. I've been doing it for a long time. 0:16:06.360 --> 0:16:08.760 And the reality is that that answer can sometimes be 0:16:08.760 --> 0:16:10.840 a little of our cop out, right, is like I 0:16:10.880 --> 0:16:13.480 know you were doing machine learning, you were doing some 0:16:13.560 --> 0:16:16.440 of these things, but actually the leader's version of AI, 0:16:16.560 --> 0:16:19.280 what what's happening with foundation models? Not only is it 0:16:19.400 --> 0:16:22.800 very new. It's very hard to do, and honestly, if 0:16:22.800 --> 0:16:25.560 you haven't been you know, assembling very large teams and 0:16:25.560 --> 0:16:28.080 spending hundreds of millions of dollars of compute, and so 0:16:28.240 --> 0:16:31.000 you're probably not doing it right. You're doing something else 0:16:31.040 --> 0:16:33.680 that is in the broad category. And I think the 0:16:33.800 --> 0:16:36.600 lessons about what it means to make this transition to 0:16:36.640 --> 0:16:39.479 this new wave is still in early phases of understanding. 0:16:39.680 --> 0:16:41.280 So what would you say? I want to give you 0:16:41.280 --> 0:16:44.800 a couple of examples of people with kind of real 0:16:44.840 --> 0:16:47.880 world in real world positions of responsibility. Imagine I'm sitting 0:16:47.960 --> 0:16:50.840 right here, So imagine that I am the president of 0:16:50.840 --> 0:16:53.160 a small liberal arts college and I come to you 0:16:53.200 --> 0:16:55.200 and I say, Dario, I keep hearing about a AI 0:16:56.000 --> 0:16:58.800 my college has you know I don't make it. You know, 0:16:58.880 --> 0:17:01.840 I'm I'm not I'm making this much money. If that 0:17:01.960 --> 0:17:06.399 every year enroments declining, I feel like this maybe is 0:17:06.440 --> 0:17:09.760 an opportunity. What is the opportunity for me? What would 0:17:09.800 --> 0:17:10.600 you say? 0:17:11.720 --> 0:17:14.040 So, it's probably in a couple of segments around that. 0:17:14.280 --> 0:17:17.720 Right one has to do is well, what is the 0:17:17.720 --> 0:17:21.919 implications of this technology inside the institution itself instead of 0:17:22.000 --> 0:17:25.639 the college, And how we operate and can we improve 0:17:25.680 --> 0:17:28.480 for example, efficiency, like if you have in very low 0:17:28.880 --> 0:17:31.919 levels of sort of margin to be able to reinvest 0:17:32.520 --> 0:17:37.119 is you know you run it, you run you know infrastructure, 0:17:37.200 --> 0:17:39.240 you run many things inside the college. What are the 0:17:39.280 --> 0:17:43.560 opportunities to increase the productivity or automate and drive savings 0:17:43.560 --> 0:17:46.200 such that you can reinvest that money into the mission 0:17:46.200 --> 0:17:46.679 of education? 0:17:46.840 --> 0:17:49.320 Right as an example, So number one is operational efficiency. 0:17:49.320 --> 0:17:52.760 Operational efficiency is a big one. I think the second 0:17:52.800 --> 0:17:55.240 one is within the context of the college, there's implications 0:17:55.240 --> 0:17:58.000 for the educational mission on its own, right, How will 0:17:58.359 --> 0:18:01.000 you know how does a correct need to evolve or not? 0:18:01.440 --> 0:18:04.160 What are acceptable use policies or of someone these ai 0:18:04.280 --> 0:18:06.320 I think we've all read a lot about like what 0:18:06.359 --> 0:18:08.760 can happen in terms of exams and so on and 0:18:08.800 --> 0:18:11.200 cheating and not cheating, or what are the actually positive 0:18:11.240 --> 0:18:13.600 elements of it in terms of how curriculum should be 0:18:13.640 --> 0:18:17.280 developed and professions sustain around that. And then there's another 0:18:17.280 --> 0:18:19.960 third dimension, which is the outdoor oriented element of it, 0:18:20.000 --> 0:18:23.879 which is like prospect students right, so, which is frankly speaking, 0:18:23.920 --> 0:18:26.040 a big use case that is happening right now, which 0:18:26.080 --> 0:18:28.560 in the broader industry is called customer care or client 0:18:28.560 --> 0:18:30.959 care or citizen care. So in this question will be education, 0:18:31.080 --> 0:18:33.680 like you know, hey, are you reaching the right students 0:18:34.359 --> 0:18:37.320 around that that may apply to the college. How can 0:18:37.320 --> 0:18:39.760 you create them? For example, an environment to interact with 0:18:39.760 --> 0:18:41.800 the college and answering questions that could be a chat 0:18:41.840 --> 0:18:44.879 bought or something like that to learn about it. And personalization. 0:18:45.280 --> 0:18:47.720 So I would say there's like at least three lenses 0:18:47.800 --> 0:18:49.399 with which I would give advice, right. 0:18:49.320 --> 0:18:54.960 The positive seglee because it's really interesting. So I really 0:18:54.960 --> 0:18:58.800 can't as sign an essay anymore? Can I? 0:18:58.840 --> 0:19:00.119 Can I sign an essay? 0:19:00.280 --> 0:19:03.120 Can I say? Rend me a research paper? And come 0:19:03.119 --> 0:19:03.760 back to being three? 0:19:03.760 --> 0:19:03.840 We? 0:19:03.920 --> 0:19:04.920 Can I do that anymore? 0:19:05.000 --> 0:19:05.600 I think you can? 0:19:05.720 --> 0:19:06.520 How do I do that? 0:19:06.640 --> 0:19:10.280 And then you can that Look, there's there's two questions 0:19:10.280 --> 0:19:14.440 around that. I think that if one goes and explains 0:19:14.440 --> 0:19:16.240 in the context like what is it? Why are we here? 0:19:16.280 --> 0:19:18.280 Why are in this class? What is the purpose of this? 0:19:19.160 --> 0:19:22.679 And and one starts with assuming like an element of 0:19:22.720 --> 0:19:24.840 like decency and people are people are there like to 0:19:24.920 --> 0:19:27.280 learn and so on, and you just give it this disclaimer. Look, 0:19:27.520 --> 0:19:30.040 I know that one option you have is like just 0:19:30.119 --> 0:19:32.280 you know, put the essay question and click, go and 0:19:32.320 --> 0:19:34.280 like and give an answer. You know, but that is 0:19:34.359 --> 0:19:36.440 not why we're here, and that is not the intent 0:19:36.480 --> 0:19:38.040 of what we're trying to do. So first I would 0:19:38.080 --> 0:19:42.400 start with the sort of like the norms of intent 0:19:42.560 --> 0:19:45.560 and decency and appeal to those as step number one. 0:19:46.200 --> 0:19:48.359 Then we all know that there will be a distribution 0:19:48.400 --> 0:19:50.439 of use cases of people like that will come in 0:19:50.480 --> 0:19:52.320 one year and come out of the other and do that. 0:19:52.880 --> 0:19:55.359 And so for a subset of that, you know, I 0:19:55.400 --> 0:19:57.320 think the technology is going to have all in such 0:19:57.359 --> 0:19:59.720 a way that we will have more and more of 0:19:59.760 --> 0:20:02.719 the to discern, right, you know, when that has been 0:20:02.760 --> 0:20:06.880 AI generated right and uncreated, it won't be perfect, right, 0:20:06.960 --> 0:20:09.480 But there's some elements that you can imagine in putting 0:20:09.480 --> 0:20:11.520 the essay and you say, hey, this is likely to 0:20:11.520 --> 0:20:14.680 be generated right around that. And for example, one way 0:20:14.680 --> 0:20:16.080 you can do that, just to give you an intuition, 0:20:16.160 --> 0:20:18.480 you could just have an essay that you write with 0:20:18.600 --> 0:20:21.920 pencil and paper. At the beginning, you get a baseline 0:20:21.960 --> 0:20:24.600 of what you're writing is like, and then later when 0:20:24.640 --> 0:20:28.040 you you know generate it, there will be obvious differences 0:20:28.160 --> 0:20:30.560 around what kind of writing has been generating on the 0:20:30.560 --> 0:20:30.880 other way. 0:20:30.880 --> 0:20:34.480 But you've turned it's everything you're describing makes sense put 0:20:34.520 --> 0:20:38.399 it greatly in this respect, at least, it seems to 0:20:38.440 --> 0:20:41.400 greatly complicate the life of the teacher, whereas the other 0:20:41.440 --> 0:20:46.199 two use cases seem to kind of clarify and simplify 0:20:46.280 --> 0:20:51.119 the role. Right suddenly, you know, reaching student perspective students, 0:20:51.160 --> 0:20:53.119 sounds like I can do that much more kind of 0:20:53.160 --> 0:20:55.840 efficient in a lite. Yeah, I can bring you administration costs, 0:20:55.880 --> 0:20:57.840 but the teaching thing is tricky. 0:20:58.960 --> 0:21:02.720 Well, until we developed the new norms, right, I mean again, 0:21:02.880 --> 0:21:05.160 I mean, I know it's not abuse analogy, but calculators 0:21:05.200 --> 0:21:08.760 we deal. We've done with that too, right, And it says, well, calculator, 0:21:08.800 --> 0:21:10.199 what is the purpose of math? How are we going 0:21:10.280 --> 0:21:10.600 to do this? 0:21:10.680 --> 0:21:13.920 And so can I tell you my dad's calculator story? 0:21:14.040 --> 0:21:14.639 Yes? Please. 0:21:14.880 --> 0:21:18.879 My father was a mathematician, taught mathematics at University of Waterloo, Agada, 0:21:19.560 --> 0:21:22.879 and in the seventies when people started to get pocket calculators, 0:21:23.359 --> 0:21:25.920 his students demanded that they'd be able to use them, 0:21:26.119 --> 0:21:27.879 and he said no, and they took him to the 0:21:27.880 --> 0:21:33.720 administration and he lost. So he then changed completely throughout 0:21:33.720 --> 0:21:36.800 all of his old exams and introduced new exams where 0:21:37.160 --> 0:21:41.400 there was no calculation. It was all like deep think, 0:21:41.520 --> 0:21:43.840 you know, figure out the problem on a conceptual level 0:21:43.840 --> 0:21:47.120 and describe it to me. And they were all students 0:21:47.160 --> 0:21:49.440 deeply unhappy that he'd made their lives for computation. 0:21:49.800 --> 0:21:51.320 But it's to other. 0:21:51.280 --> 0:21:54.760 Point, to your point, I mean, he probably the result 0:21:54.840 --> 0:21:58.919 was probably a better education. He just removed the element 0:21:58.960 --> 0:22:01.840 that they could gain with their pocket calculators. I suppose 0:22:01.880 --> 0:22:02.600 it's a version of. 0:22:02.720 --> 0:22:04.359 I think it's a version of that, And so I 0:22:04.359 --> 0:22:06.480 think they will develop the equivalent of what your father did. 0:22:06.480 --> 0:22:07.960 And I think people say, you know what if like 0:22:08.040 --> 0:22:10.360 these kinds of things, everybody's doing it generically and none 0:22:10.400 --> 0:22:12.600 of us have any meaning because all you're doing is 0:22:12.640 --> 0:22:14.800 pressing buttons, and like the intent of this was something 0:22:14.800 --> 0:22:16.439 which was to teach you how to write or to 0:22:16.520 --> 0:22:18.760 think or something. There may be a variant of how 0:22:18.800 --> 0:22:20.920 we do all of this. I mean, obviously some version 0:22:20.960 --> 0:22:22.960 of that that has happened is like Okay, we're all 0:22:22.960 --> 0:22:24.600 going to sit down and doing with pencil on paper 0:22:24.640 --> 0:22:27.359 and computers in their classroom. But there'll be other variants 0:22:27.400 --> 0:22:29.879 of creativity that people will put forth to say, you 0:22:29.880 --> 0:22:31.919 know what, you know, that's a way to solve that 0:22:31.960 --> 0:22:32.480 problem too. 0:22:32.560 --> 0:22:35.600 But this is interesting because to stay on this analogy, 0:22:36.240 --> 0:22:41.639 we're really talking about a profound rethinking. Just using college 0:22:41.640 --> 0:22:45.400 as an example, a real profound rethinking of the way, 0:22:46.000 --> 0:22:49.760 there's no part of this college it's unaffected by aia B. 0:22:50.760 --> 0:22:53.880 In one case, I've made everyone's job easier. In one case, 0:22:53.920 --> 0:22:57.000 I've made I'm asking us to really rethink from the 0:22:57.000 --> 0:23:01.360 ground up what teaching means. In another case, I've automated 0:23:01.359 --> 0:23:03.040 systems that I didn't think of it. I mean, it's like, 0:23:03.359 --> 0:23:06.280 that's right, that's it's not that's a lot to ask 0:23:06.359 --> 0:23:10.000 someone who got a PhD in medieval language literature, you know, 0:23:10.119 --> 0:23:10.920 forty years ago. 0:23:11.520 --> 0:23:13.720 Yeah. But you know, I'll tell you a positive sort 0:23:13.760 --> 0:23:16.399 of development that I'm seeing the sciences around this, which 0:23:16.480 --> 0:23:19.840 is you're seeing as you see more and more examples 0:23:20.160 --> 0:23:24.400 of applying AI technology within the context of like historians 0:23:24.400 --> 0:23:27.879 to as an example, right, you have archival and you know, 0:23:27.920 --> 0:23:29.560 and you have all these books and being able to 0:23:29.640 --> 0:23:32.719 actually help you as an assistant right around that, but 0:23:32.760 --> 0:23:35.960 not only with text now, but with diagrams, right, And 0:23:36.320 --> 0:23:40.080 I've seen it in anthropology too, Rite and archaeology with 0:23:40.200 --> 0:23:43.439 examples of engravings and translations and things that can happen. 0:23:43.800 --> 0:23:47.360 So so as you see in diverse fields people applying 0:23:47.400 --> 0:23:49.920 these techniques to advance and how to do physics or 0:23:49.920 --> 0:23:53.239 how to do chemistry, they inspire each other, right, and 0:23:53.240 --> 0:23:55.080 they said, you know, how does it apply actually to 0:23:55.160 --> 0:23:58.760 my area? So once as that happens, it becomes less 0:23:58.760 --> 0:24:00.320 of a chore of like my god, you know, how 0:24:00.359 --> 0:24:02.439 do I have to deal with this? But actually it's 0:24:02.440 --> 0:24:06.159 triggered by curiosity, is triggered by you know, they'll be like, 0:24:06.280 --> 0:24:07.960 you know, faculty that will be like you know what, 0:24:08.040 --> 0:24:10.240 you know, let me explore what this means for my area, 0:24:10.640 --> 0:24:13.000 and they will adapt it to the local context, to 0:24:13.040 --> 0:24:16.600 the local you know, language, and the professional itself. So 0:24:16.840 --> 0:24:19.679 I see that as a positive vector that is not 0:24:19.720 --> 0:24:21.639 all going to feel like homework, you know, it's not 0:24:21.800 --> 0:24:24.000 going to feel like, oh my god, this is so overwhelming, 0:24:24.280 --> 0:24:26.639 but rather to be very practical to see what works, 0:24:26.760 --> 0:24:28.920 What have I seen others to do that is inspiring, 0:24:29.080 --> 0:24:31.120 and what am I inspired to do? You know what? 0:24:31.119 --> 0:24:33.080 What is how is this going to help my career? 0:24:33.400 --> 0:24:35.119 I think that that's going to be an interesting question 0:24:35.280 --> 0:24:37.399 for for you know, those faculty members for the. 0:24:37.400 --> 0:24:40.639 Students, the professional Sorry, I'm gonna stick with this example 0:24:40.640 --> 0:24:43.399 along because it's really interesting. I'm curious, following up on 0:24:43.440 --> 0:24:46.240 what you just said, that one of the most persistent 0:24:46.440 --> 0:24:50.679 critiques of academic but also of many of many corporate 0:24:50.720 --> 0:24:56.240 institutions in these years, has been siloing. Right, Different parts 0:24:56.280 --> 0:24:59.520 of the of the organization are going off on their own, 0:24:59.560 --> 0:25:03.520 and that's to each other. Is a potential is a 0:25:03.680 --> 0:25:08.880 real potential benefit to AI the kind of breaking down 0:25:09.440 --> 0:25:12.479 a simple tool for breaking down those kinds of barriers. 0:25:12.560 --> 0:25:14.439 Is that a very Is that an elegant way of 0:25:14.440 --> 0:25:16.359 sort of saying what. 0:25:15.760 --> 0:25:17.679 I really think? And I was actually just having a 0:25:17.680 --> 0:25:20.720 conversation with Provos stuff and very much on this topic, 0:25:20.840 --> 0:25:24.600 very recently, exactly on that which is all these this, 0:25:24.800 --> 0:25:27.600 you know, this appetite right to collaborate across disciplines. There's 0:25:27.600 --> 0:25:31.840 a lot of attempts stores a goal, right, creating interdisciplinary centers, 0:25:32.400 --> 0:25:36.520 creating dual degree programs or dual appointment programs. But actually 0:25:37.080 --> 0:25:41.200 in a lot of progress in academia happens by methodology too. 0:25:41.440 --> 0:25:44.040 Write like a new you know, when when some methodology 0:25:44.040 --> 0:25:46.960 gets adopted, I mean the most famous example of that 0:25:47.040 --> 0:25:49.399 is a scientific method as an example of that. But 0:25:49.480 --> 0:25:52.320 when you have a methodology that gets adopted, it also 0:25:52.359 --> 0:25:56.639 provides a way to speak to your colleagues across different disciplines, 0:25:56.880 --> 0:25:59.399 and I think what's happened in AI is linked to 0:25:59.440 --> 0:26:02.560 that that within the context of the scientific method as 0:26:02.560 --> 0:26:08.040 an example, the methodology about we about what we do discovery, 0:26:08.400 --> 0:26:11.119 the role of data, the role of these neural networks, 0:26:11.119 --> 0:26:13.760 of how we actually find proximity to concepts to one 0:26:13.760 --> 0:26:19.320 another is actually fundamentally different than how we've traditionally applied it. 0:26:19.720 --> 0:26:23.040 So as we see across more professions, people applying this 0:26:23.160 --> 0:26:26.520 methodology is also going to give some element of common 0:26:26.600 --> 0:26:29.960 language to each other. Right And in fact, you know, 0:26:30.200 --> 0:26:33.480 in this very high dimensional representation of information that is 0:26:33.520 --> 0:26:37.520 pressent to neural networks, we may find amazing adjacencies or 0:26:37.560 --> 0:26:41.280 connections of themes and topics in ways that the individual 0:26:41.359 --> 0:26:45.080 practitioners cannot describe, but yet will be latent in these 0:26:45.160 --> 0:26:48.280 large cal neural networks. We are going to suffer a 0:26:48.320 --> 0:26:50.880 little bit from causality, from the problem of like, hey, 0:26:50.880 --> 0:26:53.720 what's the root cause of that? Because I think one 0:26:53.760 --> 0:26:57.959 of the unsatisfying aspects that this methodology will provide is 0:26:58.040 --> 0:26:59.960 they may give you answers for which they don't give 0:26:59.960 --> 0:27:04.120 you good reasons for where the answers came from and uh, 0:27:04.160 --> 0:27:06.679 and then there will be the traditional process of discovery 0:27:06.760 --> 0:27:09.320 of saying, if that is the answer, what are the reasons? 0:27:09.760 --> 0:27:11.600 So we're going to have to do this sort of 0:27:11.720 --> 0:27:15.360 hybrid way of understanding the world. But I do think 0:27:15.400 --> 0:27:18.080 that common layer of AI is a powerful new thing. 0:27:18.400 --> 0:27:21.520 Yeah. Well, a couple of random questions. I couldn't mind 0:27:21.520 --> 0:27:24.439 as you talk. In the In the Writer's strike that 0:27:24.560 --> 0:27:27.000 just ended in Hollywood, one of the sticking points was 0:27:27.560 --> 0:27:31.639 how the studios and writers would treat AI generated content. Right, 0:27:31.800 --> 0:27:35.480 good writers get credit if their material with somehow the 0:27:35.560 --> 0:27:40.119 source for a but more broadly, did the writers need 0:27:40.160 --> 0:27:42.520 protections against the use of I could go on. You 0:27:42.520 --> 0:27:45.160 know what, probably we're familiar with all of this. Had 0:27:45.200 --> 0:27:47.440 you been I don't know whether you were, but had 0:27:47.720 --> 0:27:52.080 either side called you in for advice during that the writers, 0:27:52.080 --> 0:27:54.720 had the writers called you and said, Daria, what should 0:27:54.760 --> 0:27:57.520 we do about AI? And how should we that should 0:27:57.520 --> 0:27:59.960 be ref how should that be reflected in our content 0:28:00.000 --> 0:28:01.760 track negotiations? What would you've told. 0:28:01.600 --> 0:28:06.000 Them the way I think about that is that I died. 0:28:06.119 --> 0:28:08.360 I would divide it into two pieces. First, is what's 0:28:08.400 --> 0:28:13.159 technically possible, right, and anticipate scenarios like you know, what 0:28:13.240 --> 0:28:16.120 can you do with voice cloning for example? You know, now, 0:28:16.160 --> 0:28:20.879 for example it is possible there being dubbing right legist 0:28:20.880 --> 0:28:22.800 take that topic right around the world. There was all 0:28:22.840 --> 0:28:26.560 these folks that would dub people in other languages. Well, 0:28:26.600 --> 0:28:29.520 now you can do these incredible rendering in some I mean, 0:28:29.520 --> 0:28:31.600 I know if you've seen them, where you know you 0:28:31.720 --> 0:28:34.520 match the lips is your original voice, but speaking any 0:28:34.600 --> 0:28:37.000 language that you want. As an example, so busy that 0:28:37.080 --> 0:28:38.880 has a set of implications around that. I mean, just 0:28:38.880 --> 0:28:40.520 to give an example, So I would say, create a 0:28:40.560 --> 0:28:45.080 taxonomy that describes technical capabilities that we know of today 0:28:45.920 --> 0:28:49.960 and applications to the industry, and two examples of like, hey, 0:28:50.000 --> 0:28:51.520 you know I could film you for five minutes and 0:28:51.520 --> 0:28:53.480 I could generate two hours of content of you and 0:28:53.520 --> 0:28:55.440 I don't have to you know, then if you get 0:28:55.480 --> 0:28:57.440 paid by the hour, obviously I'm not paying you for 0:28:57.480 --> 0:29:00.400 that other thing. So I would say technological capability and 0:29:00.440 --> 0:29:03.920 then map with their expertise consequences of how it changes 0:29:03.960 --> 0:29:06.320 the way they work or the way they interact or 0:29:06.360 --> 0:29:08.360 the way they negotiate and so on. So that would 0:29:08.360 --> 0:29:10.640 be one element of it, and then the other one 0:29:10.760 --> 0:29:13.320 is like a non technology related matter, which is an 0:29:13.320 --> 0:29:16.080 element of almost of distributed justice, is like who deserves 0:29:16.120 --> 0:29:18.400 what right and who has the power to get what? 0:29:19.280 --> 0:29:23.000 And then that's a completely different discussion. That is to say, well, 0:29:23.040 --> 0:29:25.840 if this is the scenario of what's possible, you know, 0:29:25.920 --> 0:29:28.920 what do we want and what are we able to get? 0:29:29.080 --> 0:29:31.200 And I think that that's a different discussion, which is 0:29:31.400 --> 0:29:32.320 which we all as. 0:29:32.200 --> 0:29:33.320 Life, which when do you do? 0:29:33.400 --> 0:29:38.320 First? I think it's very helpful to have an understanding 0:29:38.600 --> 0:29:42.520 of what's possible and how it changes a landscape as 0:29:42.600 --> 0:29:46.480 part of a broader discussion, right, and a broad negotiation 0:29:47.080 --> 0:29:50.800 because you also have to see the opportunities because there 0:29:50.800 --> 0:29:53.480 will be a lot of ground to say, actually, you know, 0:29:53.880 --> 0:29:56.360 if we can do it in this way and we 0:29:56.400 --> 0:29:58.760 can all be that much more efficient in getting this 0:29:58.920 --> 0:30:02.640 piece work done on this but we have a reasonable 0:30:02.720 --> 0:30:06.520 agreement about how we both sides benefit from it, right, 0:30:07.280 --> 0:30:09.600 then that's a win win for everybody. Yeah, Right, So 0:30:09.680 --> 0:30:12.640 that's I think that would be a golden triangle, right. 0:30:12.560 --> 0:30:15.120 Here's my reading, and I would like you to correct 0:30:15.120 --> 0:30:16.880 me if I'm wrong, and I'm likely to be wrong. 0:30:18.000 --> 0:30:20.120 When I looked at that strike, I said, if they're 0:30:20.120 --> 0:30:24.160 worried about AI, the writers are worried about AI. That 0:30:24.280 --> 0:30:26.840 seems silly. It should be the studios who are worried 0:30:26.840 --> 0:30:29.280 about the economic impact of AI. Does it in the 0:30:29.320 --> 0:30:32.360 long run AI put the studios out of business long 0:30:32.400 --> 0:30:34.320 before it puts the writers out of business. I only 0:30:34.360 --> 0:30:38.360 need the studio because the costs of production are as 0:30:38.440 --> 0:30:41.000 high as the sky, and the cost of production are overwhelming. 0:30:41.040 --> 0:30:44.280 And whereas if I don't, if I have a tool 0:30:44.320 --> 0:30:49.680 which brings introduces massive technological efficiencies to the production of movies, 0:30:50.120 --> 0:30:52.600 then I don't. Why don't need a studio? Why would 0:30:52.600 --> 0:30:53.440 they the scared ones? 0:30:53.600 --> 0:30:55.640 Or maybe maybe you need like a different kind of 0:30:55.680 --> 0:30:57.560 studio or a different kind of different kind of study. 0:30:57.640 --> 0:31:01.320 But I mean the in the but in in this strike, 0:31:01.400 --> 0:31:04.680 the fright the frightened ones with the writers and the 0:31:05.200 --> 0:31:08.200 you know, with the studios. Wasn't that backwards? 0:31:09.600 --> 0:31:11.960 I haven't thought about it. Uh, it can be about 0:31:11.960 --> 0:31:13.800 the implications of it. It goes back to we're talking 0:31:13.840 --> 0:31:16.960 before the implications because are so horizontal. It is right 0:31:17.000 --> 0:31:18.520 to think about it, like what does it do to 0:31:18.560 --> 0:31:19.680 the studios as well? Right? 0:31:19.800 --> 0:31:21.000 Yeah? 0:31:21.040 --> 0:31:23.680 But then you know, the reason why that happens is 0:31:23.720 --> 0:31:28.160 that it's the order of either negotiations or who first 0:31:28.160 --> 0:31:31.840 got concerned about it and did something about it, right, 0:31:31.880 --> 0:31:34.680 which is in the context of the strike. You know, 0:31:34.800 --> 0:31:37.160 I don't know what the equivalent conversations are going inside 0:31:37.200 --> 0:31:39.120 the studio and whether they have a workroom saying what 0:31:39.160 --> 0:31:41.160 this is going to mean to us? Right, but it 0:31:41.200 --> 0:31:43.720 doesn't get exercise through a strike, but maybe through a 0:31:43.840 --> 0:31:46.160 task force inside you know, the companies about what are 0:31:46.160 --> 0:31:46.920 they going to do? Right? 0:31:47.120 --> 0:31:49.360 Well, And to go back to your thing you said, 0:31:49.400 --> 0:31:50.680 the first thing you do is you make a list 0:31:50.680 --> 0:31:55.400 of what technological capabilities are. But don't technological capabilities change 0:31:55.400 --> 0:31:58.960 every I mean they do. You're raising ahead so fast, 0:31:59.160 --> 0:32:01.959 so you can't. Can you you have a contract? I'm 0:32:01.960 --> 0:32:04.080 sorry for getting into little weeds here, but this is interesting. 0:32:04.520 --> 0:32:07.360 Can you you can't have a five year contract if 0:32:07.360 --> 0:32:11.080 the contract is based on an assessment of technological capabilities 0:32:11.120 --> 0:32:13.840 in twenty twenty three, because by the time get to 0:32:13.880 --> 0:32:18.680 twenty eight, twenty three eight, it's totally. 0:32:18.320 --> 0:32:21.120 Different, right, yeah, But like you know, I mean where 0:32:21.120 --> 0:32:24.720 I was going is like there are some abstractions around 0:32:24.720 --> 0:32:27.560 that is like, you know, one can we do with 0:32:27.760 --> 0:32:28.360 my image? 0:32:28.440 --> 0:32:28.640 Right? 0:32:28.680 --> 0:32:31.240 Like if I generally get the category that my image 0:32:31.240 --> 0:32:34.080 can be reproduced, generated contents and so on, it's like, 0:32:34.200 --> 0:32:36.760 let's talk about the abstract notion about who has rights 0:32:36.800 --> 0:32:39.600 to that or do we both get to benefit from that? 0:32:39.800 --> 0:32:42.600 If you get that straight, Yes, the nature of how 0:32:42.640 --> 0:32:46.000 the image gets alter created as something will change underneath, 0:32:46.160 --> 0:32:48.760 but the concept will stay the same. And so I 0:32:48.760 --> 0:32:50.960 think it's what's important is to get the categories right. 0:32:51.280 --> 0:32:56.480 Yeah. Yeah, If you just think about the biggest technological 0:32:58.000 --> 0:33:02.480 revolutions of the post war era last seventy five years, 0:33:03.400 --> 0:33:05.280 you can all come up with a list. Actually, it's 0:33:05.280 --> 0:33:06.640 really fun to come up with the list. I was 0:33:06.720 --> 0:33:10.360 thinking about this when we were you know, containerized shipping 0:33:10.440 --> 0:33:17.280 is my favorite, the green revolution, the internet is Where 0:33:17.360 --> 0:33:18.400 is the I in that list? 0:33:21.080 --> 0:33:23.880 So I would put it first in that context that 0:33:23.920 --> 0:33:28.000 you put forth over since World War Two, undoubtedly, like 0:33:28.200 --> 0:33:32.320 computing as a category is one of those trajectories that 0:33:32.520 --> 0:33:36.400 has reshaped right or world. And I think we think computing, 0:33:36.960 --> 0:33:41.040 I would say the role that semiconductors have had has 0:33:41.120 --> 0:33:45.160 been incrowdly defining. I would say AI is the second 0:33:45.400 --> 0:33:49.719 example of that as a core architecture that is going 0:33:49.760 --> 0:33:52.680 to have an equivalent level of impact. And then the 0:33:52.720 --> 0:33:54.520 third leg I would put to that equation will be 0:33:54.600 --> 0:33:57.560 quantum and quantum information. And that's sort of like I 0:33:57.640 --> 0:33:59.720 like to summarize that the future of computing it's bits, 0:33:59.760 --> 0:34:03.360 neural and cubits, and is that idea of high precision computation, 0:34:04.080 --> 0:34:06.720 the world of neural networks and artificial intelligence, and the 0:34:06.760 --> 0:34:10.239 world of quantum and the combination of those things is 0:34:10.239 --> 0:34:12.240 going to be the defining force over the next hundred 0:34:12.320 --> 0:34:14.960 years in that category of computing. But it makes a 0:34:15.040 --> 0:34:15.719 list for sure. 0:34:15.960 --> 0:34:18.360 If it's that high up on the list. This is 0:34:18.360 --> 0:34:22.200 a total hypothetical. Would you if you were starting over, 0:34:22.680 --> 0:34:26.160 if you're starting IBM right now, would you say, oh, 0:34:26.440 --> 0:34:29.920 our AI operations actually should be way bigger, Like how 0:34:29.920 --> 0:34:31.960 many how many thousands of people working for you? 0:34:32.840 --> 0:34:36.160 So within the research division it's about like three thousand, 0:34:36.200 --> 0:34:37.120 five hundred scientists. 0:34:37.120 --> 0:34:39.279 So in a perfect world, would you if it's that big, 0:34:39.400 --> 0:34:42.400 isn't that too small as a group? 0:34:42.800 --> 0:34:44.920 Yeah, Well, that's like in the RICAR division. I mean 0:34:45.000 --> 0:34:48.879 IBM overall, But I mean. 0:34:48.840 --> 0:34:51.680 Like, so starting from first, so you have a you 0:34:51.880 --> 0:34:56.400 we've got a technology that you're ranking with Compute and 0:34:56.719 --> 0:34:58.759 you know, up there with as in terms of a 0:34:59.080 --> 0:35:04.120 world changer. Are we So what I'm basically asking is 0:35:04.160 --> 0:35:07.560 are we underinvested in this huge you know? 0:35:07.680 --> 0:35:10.360 But so so yeah, it's a good question. So like 0:35:10.400 --> 0:35:12.239 what I would say is that I think we should 0:35:12.320 --> 0:35:15.880 segment how many people do you need on the creation 0:35:16.680 --> 0:35:19.400 of the technology itself, and what is the right size 0:35:19.440 --> 0:35:22.239 of research and engineers and compute to do that? And 0:35:22.360 --> 0:35:24.600 how many people do you need in the sort of 0:35:24.719 --> 0:35:29.760 application of the technology to create better products, to deliver 0:35:29.920 --> 0:35:33.319 services and consulting and then ultimately to diffuse it through 0:35:33.440 --> 0:35:35.920 you know, sort of all feheres of society. And the 0:35:36.040 --> 0:35:38.520 numbers are very different, and that is not different than 0:35:38.560 --> 0:35:41.040 anywhere else. I mean, I mean, if you give examples 0:35:41.040 --> 0:35:43.279 of since you were talking about in context of World 0:35:43.280 --> 0:35:45.600 War two, how many people does it take to create, 0:35:45.920 --> 0:35:48.640 you know, an atomic weapon as an example, it's a 0:35:48.719 --> 0:35:51.120 large number. I mean, it wasn't just Los animals. There's 0:35:51.120 --> 0:35:53.200 a lot of people in Okay, it's a large number, 0:35:53.280 --> 0:35:56.839 but it wasn't a million people, right, So you could 0:35:56.880 --> 0:36:01.520 have highly concentrated teams of people that, with enough resources, 0:36:01.520 --> 0:36:05.680 can do extraordinary scientific and technological achievements. And that's always, 0:36:05.719 --> 0:36:07.680 by definition, is going to be a fraction of like 0:36:07.760 --> 0:36:10.400 one percent compared to the total volume that is going 0:36:10.440 --> 0:36:11.719 to require to then deal with it. 0:36:11.960 --> 0:36:15.040 Yeah, but the application side is infinite almost. 0:36:14.840 --> 0:36:17.520 That's exactly. So that is where like in the end 0:36:17.560 --> 0:36:21.440 the bottleneck really is. So with you know, thousands of 0:36:21.600 --> 0:36:25.279 scientists and engineers, you can create world class AI, right, 0:36:25.640 --> 0:36:27.799 And so no, you don't need ten thousand to be 0:36:27.840 --> 0:36:30.240 able to create the large language model and the generatic 0:36:30.280 --> 0:36:33.360 model and some but you need thousands, and you need 0:36:33.520 --> 0:36:35.680 you know, very significant amount of computer and data. You 0:36:35.760 --> 0:36:39.880 need that The rest is Okay, I build software, I 0:36:39.880 --> 0:36:43.280 build databases, or I build a software product that allows 0:36:43.280 --> 0:36:46.080 you to do inventory management, or I build you a 0:36:46.120 --> 0:36:51.960 photo editor and so on. Now that product incorporating the AI, modifying, 0:36:52.080 --> 0:36:54.920 expanding it and so on. Well, now you're talking about 0:36:55.000 --> 0:36:57.920 the entire software industries. So now you're talking about millions 0:36:57.920 --> 0:37:00.600 of people right who are nested, you know, who are 0:37:00.640 --> 0:37:03.799 required to bring AI into their product. Then you go 0:37:03.840 --> 0:37:07.280 on a step beyond the technology creators in terms of software, 0:37:07.280 --> 0:37:09.799 and you say, well, okay, now what the skills to 0:37:09.880 --> 0:37:13.960 help organizations go undeployed in the department of you know, 0:37:14.040 --> 0:37:16.399 the interior, right, And then I said, okay, well, now 0:37:16.400 --> 0:37:19.839 you need like consultants and experts and people to work 0:37:19.880 --> 0:37:22.239 they are to integer into the workflow. So now you're 0:37:22.280 --> 0:37:25.040 talking into the many tens of millions of people around that. 0:37:25.320 --> 0:37:27.719 So I see it as these concentric circles of it, 0:37:28.160 --> 0:37:31.120 but to some degree in many of these core technology areas, 0:37:31.160 --> 0:37:32.680 just saying like well, I need a team of like 0:37:32.680 --> 0:37:35.040 one hundred thousand people to create like AI or a 0:37:35.360 --> 0:37:37.879 or a new transistor or a new quantum computer. It's 0:37:37.920 --> 0:37:40.120 actually a diminished in return right in the end, like 0:37:40.200 --> 0:37:42.560 too many people connecting with each other's very difficult. 0:37:42.560 --> 0:37:45.399 But on the application side of just to go back 0:37:45.440 --> 0:37:51.200 to our example of that college, just the task of 0:37:51.280 --> 0:37:55.799 sitting down with a faculty and working with them to 0:37:55.880 --> 0:37:59.520 reimagine what they do with these new set of tools 0:37:59.600 --> 0:38:02.200 in mind, and with the understanding that the students coming 0:38:02.200 --> 0:38:03.759 in are probably going to know more about it than 0:38:03.800 --> 0:38:06.840 they do that a lot, I mean, that's a that 0:38:07.000 --> 0:38:09.759 is a curricullion people problem. 0:38:09.960 --> 0:38:11.960 It's a people problem. Yeah, that's why I started in 0:38:12.080 --> 0:38:13.880 terms of the barriers of adoption of that. I mean 0:38:13.920 --> 0:38:17.160 the context of IBM an example, that's why we have 0:38:17.640 --> 0:38:21.440 a consulting organization, IVAN Consulting that complements ib AND technology, 0:38:21.960 --> 0:38:24.759 and the IVAN Consulting Organization has over one hundred and 0:38:24.800 --> 0:38:28.480 fifty thousand employees because of this question, right, because you 0:38:28.560 --> 0:38:30.960 have to sit down and you say, Okay, what problem 0:38:30.960 --> 0:38:33.520 are you trying to solve, what is the methodology we're 0:38:33.520 --> 0:38:35.480 going to do, and here's the technology options that we 0:38:35.600 --> 0:38:37.680 have to be able to bring into the table. In 0:38:37.760 --> 0:38:42.960 the end, the adoption across or society will be limited 0:38:43.200 --> 0:38:46.560 by this part. The technology is going to make it easier, 0:38:46.640 --> 0:38:51.479 more cost effective to implement those solutions, but you first 0:38:51.560 --> 0:38:53.480 have to think about what you want to do, how 0:38:53.520 --> 0:38:55.040 you're going to do it, and how are you going 0:38:55.080 --> 0:38:56.719 to bring it into a life of this in this 0:38:56.800 --> 0:39:00.480 context faculty member or you know, the administrator and so 0:39:00.600 --> 0:39:01.360 on in these colleges. 0:39:01.600 --> 0:39:04.360 With that Hollywood that that notion I thought, which was 0:39:04.400 --> 0:39:09.680 absolutely I thought really interesting that in a Hollywood strike 0:39:09.760 --> 0:39:13.240 you have to have this conversation about a distributive justice, 0:39:13.440 --> 0:39:16.920 conversation about how do we that's it's a really hard conversation, 0:39:17.360 --> 0:39:19.719 right to have. And uh so this brings me to 0:39:19.719 --> 0:39:21.399 my nee, which is that you we were talking about 0:39:21.440 --> 0:39:25.480 stage you have. You have two daughters, one in college, 0:39:25.480 --> 0:39:28.359 one about to go to college. That's right, so they're 0:39:28.360 --> 0:39:32.799 both science minded. So tell me about the conversations you 0:39:32.800 --> 0:39:35.680 you have with your daughter. You have a unique conversation 0:39:35.760 --> 0:39:38.839 with your daughters because your conversation, your advice to them 0:39:38.960 --> 0:39:42.360 is is influenced by what you do for a living. 0:39:42.640 --> 0:39:43.480 Yes, it's true. 0:39:43.640 --> 0:39:48.000 So did you warn your daughters away from certain fields? 0:39:48.040 --> 0:39:53.600 Did you say whatever you do, don't be you know, no. 0:39:52.560 --> 0:39:55.120 No, no, that's not my style. I mean for me, no, 0:39:55.400 --> 0:39:57.520 I try not to be like you know, preachy about that. 0:39:58.480 --> 0:40:01.000 So for me, it was just about by example of 0:40:01.120 --> 0:40:04.160 things I love, right and things I care about, and 0:40:04.200 --> 0:40:06.360 then you know, bringing them to the lab and seeing 0:40:06.440 --> 0:40:09.680 things and then the natural conversations of things working on 0:40:10.000 --> 0:40:13.040 or interesting people I meet. So to the extent that 0:40:13.080 --> 0:40:15.400 they have chosen that and obviously this has an influence 0:40:15.440 --> 0:40:19.040 on them. It has been through seeing it, you know, 0:40:19.080 --> 0:40:21.040 perhaps through my eyes, right, and what do you see 0:40:21.080 --> 0:40:22.960 me do? And that I like my profession, right. 0:40:22.760 --> 0:40:25.680 But one of your daughters. You said, is thinking that 0:40:25.760 --> 0:40:29.000 she wants to be a doctor. But being a doctor 0:40:29.120 --> 0:40:31.600 in a post AI world it's surely a very different 0:40:31.600 --> 0:40:34.759 proposition than being a doctor in a PREAI world. Do 0:40:34.800 --> 0:40:38.800 you think have you tried to prepare her for that difference? 0:40:39.000 --> 0:40:41.040 Have you explained to her what you think will happen 0:40:41.080 --> 0:40:42.399 to this profession she might enter. 0:40:42.920 --> 0:40:45.840 Yeah, I mean not in like, you know, incredible amount 0:40:45.880 --> 0:40:49.640 of detail, but yes, at the level of understanding what 0:40:49.760 --> 0:40:53.239 is changing, like this lens of the information, lens with 0:40:53.280 --> 0:40:55.279 which you can look at the world and what is 0:40:55.480 --> 0:40:58.839 possible and what it can do, Like what is our 0:40:58.960 --> 0:41:01.279 role and what is all of the technology and how 0:41:01.320 --> 0:41:04.399 that shapes At that level of abstraction, for sure, but 0:41:04.440 --> 0:41:06.840 not at the level of like, don't be a radiologist, 0:41:06.960 --> 0:41:08.719 you know, because this is what we want for you. 0:41:08.760 --> 0:41:10.600 I was gonna say, if you're not't happy with your 0:41:10.640 --> 0:41:13.160 current job, you could do a podcast called Parenting Tips 0:41:13.160 --> 0:41:17.279 with Dario, which is just an AI person gives you 0:41:17.360 --> 0:41:19.880 advice on what your kids should do based on exactly this, 0:41:20.040 --> 0:41:22.920 Like should I be a readiologist? Dario tell me, like 0:41:23.160 --> 0:41:26.799 it seems to be a really important question. Yeah, let 0:41:26.840 --> 0:41:28.560 me ask this question in a more I'm joking, but 0:41:28.640 --> 0:41:32.880 in a more serious way. Surely it would if I 0:41:32.880 --> 0:41:34.520 don't mean to use your daughter as an example. But 0:41:34.560 --> 0:41:36.759 let's imagine we're giving advice to somebody who wants to 0:41:36.880 --> 0:41:41.719 enter medicine. A really useful conversation to have is what 0:41:41.800 --> 0:41:46.279 are the skills that are will be most prized in 0:41:46.320 --> 0:41:49.760 that profession fifteen years from now, and are they different 0:41:49.800 --> 0:41:52.000 from the skills that are prized now. How would you 0:41:52.040 --> 0:41:52.960 answer that question? 0:41:53.840 --> 0:41:58.080 Yeah, I think for example, this goes back to how 0:41:58.160 --> 0:42:01.080 is this scientific method in this context, like the practice 0:42:01.120 --> 0:42:03.759 of medicine going to change? I think we will see 0:42:03.760 --> 0:42:06.279 more changes on how we practice a scientific method and 0:42:06.320 --> 0:42:10.160 so on as a consequence of what is happening with 0:42:10.360 --> 0:42:13.440 the world of computing and information, how we represent information, 0:42:13.560 --> 0:42:16.840 how we represent knowledge, how we extract meaning from knowledge 0:42:17.000 --> 0:42:20.040 as a method than we have seen in the last 0:42:20.080 --> 0:42:23.480 two hundred years. So therefore, what I would like strongly 0:42:23.560 --> 0:42:25.920 encourage is not about like, hey, use this tool for 0:42:25.960 --> 0:42:29.040 doing this or doing that, but in the curriculum itself, 0:42:29.080 --> 0:42:32.840 in understanding how we do problems solving in the age 0:42:32.920 --> 0:42:35.920 of like data and data representation and so on, that 0:42:36.000 --> 0:42:39.319 needs to be embedded in the curriculum of everybody you 0:42:39.360 --> 0:42:41.840 know that is I would say, actually quite horizontally, but 0:42:41.920 --> 0:42:44.759 certainly in the context of medicine and scientists and so on, 0:42:44.880 --> 0:42:48.760 for sure, And to the extent that that gets ingrained, 0:42:48.920 --> 0:42:51.000 that will give us a lens that no matter what 0:42:51.840 --> 0:42:55.040 specialty they go within medicine, they will say, actually, the 0:42:55.120 --> 0:42:57.600 way I want to be able to tackle improving the 0:42:57.680 --> 0:43:00.160 quality of care, the way to do that is in 0:43:00.200 --> 0:43:02.920 addition to all the elements that we have practiced in 0:43:02.960 --> 0:43:05.839 the field of medicine. Is this new lens? And are 0:43:05.840 --> 0:43:08.080 we representing the data the right way? Do we have 0:43:08.160 --> 0:43:10.880 the right tools to be able to represent that knowledge? 0:43:11.000 --> 0:43:14.080 Am I incorporating that in my own so with my 0:43:14.120 --> 0:43:16.719 own knowledge in a way that gives me better outcomes? Right? 0:43:16.880 --> 0:43:20.200 Do I have the rigor of benchmarking too? And quality 0:43:20.480 --> 0:43:23.000 of the results? So that is what needs to be incorporated. 0:43:23.120 --> 0:43:28.839 How in a perfect world, if I asked you your 0:43:28.880 --> 0:43:33.360 team to rewrite curriculum for American medical schools, how dramatic 0:43:33.520 --> 0:43:36.760 a revision is that? Are we tinkering with ten percent 0:43:36.760 --> 0:43:38.960 of the curriculum or we tinkering with fifty percent of it? 0:43:41.200 --> 0:43:45.400 I think they would be a subset of classes. That 0:43:45.480 --> 0:43:47.920 is about the method the methodology, what has changed, like 0:43:48.400 --> 0:43:52.120 have these lens of it to understand and then within 0:43:52.320 --> 0:43:56.880 each class that methodology will represent something that is embedded 0:43:56.960 --> 0:44:01.880 in it. Right, Well, it will be substantive but not 0:44:02.080 --> 0:44:05.480 but doesn't mean replacing the specialization and the context and 0:44:05.520 --> 0:44:08.440 the knowledge of each domain. But I do think everybody 0:44:08.520 --> 0:44:11.920 should have sort of a basic knowledge of the horizontal, right, 0:44:12.000 --> 0:44:14.560 what is it, how does it work? What tools you have, 0:44:14.960 --> 0:44:16.840 what is the technology, and like you know what are 0:44:16.880 --> 0:44:19.600 the dos and don'ts around that? And then every area 0:44:19.680 --> 0:44:21.640 you say, and you know that thing that you learn, 0:44:21.760 --> 0:44:24.759 this is how it applies to anatomy, and this is 0:44:24.760 --> 0:44:27.000 how you know how it applies to you know, radiology 0:44:27.000 --> 0:44:29.080 if you studying that, or or this is how you 0:44:29.120 --> 0:44:31.360 apply you know, in the context of discovery right of 0:44:31.440 --> 0:44:33.360 self structure and this is how we can use it, 0:44:33.480 --> 0:44:36.480 or protein folding and this is how it does so 0:44:36.520 --> 0:44:39.840 that way you'll see a connecting tissue through throughout the 0:44:39.880 --> 0:44:40.319 whole thing. 0:44:40.520 --> 0:44:43.440 Yeah, I mean I would add to that because I 0:44:43.480 --> 0:44:47.840 was sticking it the sch that it's also this incredible 0:44:47.840 --> 0:44:50.960 opportunity to do what doctors are supposed to do but 0:44:51.040 --> 0:44:53.600 don't have time to do now, which is they're so 0:44:53.719 --> 0:44:58.799 consumed with figuring out what's wrong with you that they 0:44:58.800 --> 0:45:01.560 have little, little time to talk about the implications of 0:45:01.600 --> 0:45:04.600 the diagnosisness and what we really want to if we 0:45:04.640 --> 0:45:08.080 can freedom of some of the burden of what is 0:45:08.080 --> 0:45:10.279 actually quite a prosaic question of what's wrong with you 0:45:10.719 --> 0:45:14.040 and leave the hard human thing of let make should 0:45:14.040 --> 0:45:17.759 you be scared or hopeful? Should you you know? What 0:45:17.800 --> 0:45:18.560 do you need to do? 0:45:18.640 --> 0:45:18.879 Or what? 0:45:19.160 --> 0:45:20.640 Let me put this in the context of all the 0:45:20.640 --> 0:45:23.640 patients I've seen, that conversation, which is the most important one, 0:45:23.680 --> 0:45:26.279 is the one that seems to me so like, if 0:45:26.280 --> 0:45:29.600 I had to, I would add, if we're reimagining the 0:45:29.760 --> 0:45:33.920 curriculum of med school, I'd like, with whatever this, by 0:45:33.920 --> 0:45:35.719 the way, very little time. Maybe we have to add 0:45:35.719 --> 0:45:38.560 two more years to med school, but like. 0:45:38.440 --> 0:45:39.200 A whole. 0:45:40.960 --> 0:45:44.959 But the whole thing about bringing back the human side 0:45:45.040 --> 0:45:49.080 of you know, now, if I can give you ten 0:45:49.080 --> 0:45:52.160 more minutes, how do you use that ten more minutes in. 0:45:52.120 --> 0:45:55.520 That in that reconceptualization that you just did is what 0:45:55.560 --> 0:45:57.440 we should be doing around that, Because I think the 0:45:57.480 --> 0:46:00.759 debate as to like, well, I'm I need doctors or 0:46:00.800 --> 0:46:03.359 not it's actually a not very useful debate. But rather 0:46:03.440 --> 0:46:06.080 this other question is how is your time being spent? 0:46:06.239 --> 0:46:09.120 What problems are you getting stuck? I mean I generalize 0:46:09.160 --> 0:46:11.719 this by like the obvious observation that if you look 0:46:11.719 --> 0:46:14.480 around in your professions, in our daily lives, we have 0:46:14.520 --> 0:46:16.719 not run out of problems to solve. So as an 0:46:16.760 --> 0:46:18.880 example of that is, hey, if I'm spending all my 0:46:18.920 --> 0:46:20.840 time trying to do diagnosis, and I could do that 0:46:20.880 --> 0:46:23.600 ten times faster, and it allows me actually to go, 0:46:24.680 --> 0:46:26.560 you know, and take care of the patients and all 0:46:26.600 --> 0:46:28.439 the next steps of what we have to do about it. 0:46:28.560 --> 0:46:30.880 That's probably a trade off that a lot of doctors 0:46:30.920 --> 0:46:33.480 would take, right. And then you say, well, you know, 0:46:33.520 --> 0:46:35.279 to what degree that does it allow me to do that? 0:46:35.320 --> 0:46:37.319 And I can do these other things and these other 0:46:37.360 --> 0:46:40.880 things are critically important for my profession around that. So 0:46:41.200 --> 0:46:44.480 when you actually become less abstract and like we get 0:46:44.840 --> 0:46:47.440 past the futile conversation of like, oh, there's no more 0:46:47.520 --> 0:46:49.000 jobs and I'm going to take it all of it, 0:46:49.040 --> 0:46:52.120 which is kind of nonsense, is you go back to say, 0:46:52.160 --> 0:46:56.239 in practice in your context, right, for you, what does 0:46:56.280 --> 0:46:58.560 it mean? How do you work? What can you do 0:46:58.560 --> 0:47:01.560 differently around that. Actually, that's a much richer conversation, and 0:47:01.640 --> 0:47:03.840 very often we would find ourselves that there's a portion 0:47:03.920 --> 0:47:05.799 of the work we do that we say I would 0:47:05.880 --> 0:47:07.919 rather do less of that. This is this other part 0:47:08.160 --> 0:47:10.920 I like a lot, And if it is possible that 0:47:10.920 --> 0:47:13.440 technology could help us make that trade off, I'll take 0:47:13.440 --> 0:47:17.799 it in a heartbeat. Now, poorly implemented technology can also 0:47:17.840 --> 0:47:19.920 create another problem. You say, Hey, this was supposed to 0:47:19.960 --> 0:47:23.319 solve me things, but the way it's being implemented is 0:47:23.360 --> 0:47:26.280 not helping me, right, it's making my life more more miserable, 0:47:26.400 --> 0:47:29.000 or so on, or I've lost connection in how I 0:47:29.120 --> 0:47:32.480 used to work, et cetera. So that is why design 0:47:33.160 --> 0:47:36.279 is so important. That is why I also workflow is 0:47:36.320 --> 0:47:39.080 so important in being able to solve these problems. But 0:47:39.920 --> 0:47:43.440 it begins by, you know, going from the intergalactic to 0:47:43.520 --> 0:47:45.920 the reality of it, of that faculty member in the 0:47:45.920 --> 0:47:48.880 Liberal Arts college or you know, or a you know, 0:47:49.000 --> 0:47:51.680 a practitioner in medicine in a hospital and what it 0:47:51.800 --> 0:47:52.799 means for them. Right. 0:47:53.480 --> 0:47:57.680 Yeah. What struck me AREIO throughout our conversation is how 0:47:57.800 --> 0:48:02.960 much of this revolution and is non technical? Is to say, 0:48:03.280 --> 0:48:05.200 you guys are doing the technical thing here, but the 0:48:05.239 --> 0:48:08.200 real the revolution is going to require a whole range 0:48:08.200 --> 0:48:11.920 of people doing things that have nothing to do with software, 0:48:12.160 --> 0:48:15.640 that have to do with working out new new human arrangements. 0:48:16.080 --> 0:48:18.719 Talking about that, I mean, does keep coming back to 0:48:18.760 --> 0:48:21.480 the Hollywood strike thing that you have to have a 0:48:21.520 --> 0:48:28.240 conversation about our values. Is creators of of of of movies? 0:48:28.280 --> 0:48:31.000 How are we going to divide up the exactly credit 0:48:31.040 --> 0:48:35.439 and the like. That's a that's a conversation about philosophy, 0:48:35.520 --> 0:48:37.399 and you know it is. 0:48:37.360 --> 0:48:40.680 And is it's in the grand tradition of why you know, 0:48:42.160 --> 0:48:46.440 a liberal education is so important in the broadest possible sense, Right, 0:48:46.840 --> 0:48:50.640 there's no common conception of the good, right, that is 0:48:50.640 --> 0:48:54.880 always a contested dialogue that happens within our society. And 0:48:55.000 --> 0:48:57.279 technology is going to fit in that context too, right. 0:48:57.400 --> 0:48:59.560 So that's why I personally, as a philosophy I'm not 0:48:59.600 --> 0:49:03.360 a technological determinists, right, And I don't like when colleagues 0:49:03.360 --> 0:49:06.040 in my profession right starts saying like, well, this is 0:49:06.080 --> 0:49:08.640 the way the technology is going to be, and by consequence, 0:49:08.880 --> 0:49:10.840 this is how society is going to be. I'm like, 0:49:10.920 --> 0:49:14.000 that's a highly contested goal. And if you want to 0:49:14.120 --> 0:49:16.640 enter into realm of politics or the real other ones, 0:49:16.680 --> 0:49:19.279 go and stand up on a stool and discuss with it. 0:49:19.280 --> 0:49:21.440 That's what society wants. You will find that it's a 0:49:21.520 --> 0:49:25.680 huge diversity of opinions and perspective and that's what makes 0:49:25.719 --> 0:49:28.319 you know, you know, in a democracy, the richness of 0:49:28.320 --> 0:49:30.520 our society. And in the end, that is going to 0:49:30.560 --> 0:49:33.520 be the centerpiece of the conversation what do we want? 0:49:34.400 --> 0:49:37.000 You know, who gets what? And so on? And that 0:49:37.320 --> 0:49:39.560 is actually I don't think it's anything negative. That's acid 0:49:39.640 --> 0:49:42.560 should be because in the end is anchor of who 0:49:42.560 --> 0:49:46.080 we want as humans, you know, you know, as friends, family, citizens, 0:49:46.280 --> 0:49:49.080 and we have many overlapping sets of responsibilities, right and 0:49:49.120 --> 0:49:52.239 as a technology creator, my only responsibility is not just 0:49:52.239 --> 0:49:55.080 as a scientist and a technology creator. I'm also a 0:49:55.080 --> 0:49:57.120 member of family. I'm a citizen, and I'm many other 0:49:57.160 --> 0:49:59.239 things that I care about. And I think that that 0:49:59.440 --> 0:50:04.879 sometimes the debate of the technological determinists they start now 0:50:04.960 --> 0:50:10.080 budding into what is the realm of you know, justice 0:50:10.320 --> 0:50:13.920 and you know, in society and philosophy and democracy, And 0:50:13.960 --> 0:50:16.480 that's where they get the most uncomfortable because it's like 0:50:16.480 --> 0:50:19.560 I'm just telling you, like you know what's possible, and 0:50:19.640 --> 0:50:23.640 when there's pushback, it's like, yeah, but now we're talking 0:50:23.640 --> 0:50:27.480 about how we live and how we work and how 0:50:27.560 --> 0:50:31.160 much I get paid or not paid. So that technology 0:50:31.200 --> 0:50:34.480 is important. Technology shapes that conversation, but we're going to 0:50:34.560 --> 0:50:38.160 have the conversation with a different language, as it should be, 0:50:38.680 --> 0:50:41.160 and technologies need to get accustomed to if they want 0:50:41.160 --> 0:50:44.080 to participate in that world with the broad consequences. Hey, 0:50:44.280 --> 0:50:46.920 get a custom to deal with the complexity of that 0:50:47.000 --> 0:50:51.399 world of politics, society, institutions, unions, all that stuff. And 0:50:51.440 --> 0:50:53.440 you know, you can be like whiny about it. It's 0:50:53.440 --> 0:50:56.000 like they're not adopting my technology. That's what it takes 0:50:56.040 --> 0:50:57.360 to bring technology into the world. 0:50:58.280 --> 0:51:04.840 Yeah, well said, thank you Dario for this wonderful conversation. 0:51:05.040 --> 0:51:08.720 Thank you to all of you for coming and listening, 0:51:08.880 --> 0:51:10.360 and thank you. 0:51:10.640 --> 0:51:10.920 Thank you. 0:51:14.440 --> 0:51:17.560 Dario gild transformed how I think about the future of AI. 0:51:18.280 --> 0:51:20.560 He explained to me how huge of a leap it 0:51:20.760 --> 0:51:24.080 was when we went from chess playing models to language 0:51:24.160 --> 0:51:27.400 learning models, and he talked about how we still have 0:51:27.480 --> 0:51:30.120 a lot of room to grow. That's why it's important 0:51:30.440 --> 0:51:33.759 that we get things right. The future of AI is 0:51:33.840 --> 0:51:38.120 impossible to predict, but the technology has so much potential 0:51:38.239 --> 0:51:41.960 in every industry. Zooming into an academic or medical setting 0:51:42.200 --> 0:51:45.280 showed just how close we are to the widespread adoption 0:51:45.640 --> 0:51:49.279 of AI. Even Hollywood is being forced to figure this out. 0:51:49.920 --> 0:51:52.640 Institutions of all sorts will have to be at the 0:51:52.719 --> 0:51:56.360 forefront of integration in order to unlock the full power 0:51:56.400 --> 0:52:01.360 of AI thoughtfully and responsibly. Humans have the power and 0:52:01.400 --> 0:52:05.360 the responsibility to shape the tech for our world. I 0:52:05.600 --> 0:52:08.480 for one, I'm excited to see how things play out. 0:52:09.840 --> 0:52:13.960 Smart Talks with IBM is produced by Matt Romano, Joey Fishground, 0:52:14.200 --> 0:52:18.600 David jaw and Jacob Goldstein. We're edited by Lydia Jane Kott. 0:52:18.920 --> 0:52:23.400 Our engineers are Jason Gambrel, Sarah Bruguier, and Ben Holliday. 0:52:24.040 --> 0:52:29.360 Theme song by Gramoscope. Special thanks to Andy Kelly, Kathy Callahan, 0:52:29.640 --> 0:52:32.480 and the eight Bar and IBM teams, as well as 0:52:32.480 --> 0:52:36.239 the Pushkin marketing team. Smart Talks with IBM is a 0:52:36.239 --> 0:52:40.840 production of Pushkin Industries and Ruby Studio at iHeartMedia. To 0:52:40.880 --> 0:52:45.799 find more Pushkin podcasts, listen on the iHeartRadio app, Apple Podcasts, 0:52:46.080 --> 0:52:51.319 or wherever you listen to podcasts. I'm Malcolm Gladwell. This 0:52:51.440 --> 0:52:58.319 is a paid advertisement from IBM.