WEBVTT - The Key to Making AI a Benefit, Not a Hazard 0:00:10.400 --> 0:00:13.000 Hello, and welcome to Stefanowie's the podcast that brings you 0:00:13.080 --> 0:00:16.200 the global economy, and we're dedicating this week's episode to 0:00:16.280 --> 0:00:19.119 a powerful book on a subject most of us have 0:00:19.160 --> 0:00:22.920 been thinking a lot about recently, how new technology can 0:00:23.040 --> 0:00:26.360 change the world. Chat GPT came out the end of 0:00:26.440 --> 0:00:29.440 last year, had a million users in the space of 0:00:29.520 --> 0:00:33.080 five days, and that was pretty cool. Then in March, 0:00:33.440 --> 0:00:36.840 chat GPT four came out. Now that was a whole 0:00:36.840 --> 0:00:42.680 lot better, but also alarming. Journalists, lawyers, accountants, teachers could 0:00:42.720 --> 0:00:45.120 all see how it could not only help them do 0:00:45.240 --> 0:00:49.200 their jobs but make them redundant. Not sometime in the future, 0:00:49.240 --> 0:00:53.199 but next month. Should we worry then about where this 0:00:53.240 --> 0:00:57.320 will lead? Well, the standard version of economic history says 0:00:57.720 --> 0:01:00.520 not really. It tells a story where I and again 0:01:00.600 --> 0:01:04.160 people fear technology will make the world worse, think of 0:01:04.200 --> 0:01:09.000 those Luddites smashing up machines. But in the end it's better. Yeah, 0:01:09.120 --> 0:01:12.600 there are adjustment costs, a bunch of people lose their jobs, maybe, 0:01:13.040 --> 0:01:17.080 but overall, the majority of people get new opportunities, more 0:01:17.120 --> 0:01:20.840 rewarding jobs, better lives. As the version of history I 0:01:20.959 --> 0:01:23.160 was taught in graduate school and have heard from fellow 0:01:23.200 --> 0:01:26.440 economists many many times since, and we've heard it again 0:01:26.640 --> 0:01:30.560 often in response to AI. To be scared of this 0:01:30.720 --> 0:01:34.600 new technology, to believe it will hurt workers, they say, 0:01:35.040 --> 0:01:38.240 you have to believe this technological revolution will be different 0:01:38.319 --> 0:01:41.039 from all those that have gone before. Now you might 0:01:41.040 --> 0:01:45.600 think that's true. This time is different. Chat GPT feels different, 0:01:46.480 --> 0:01:49.320 but it sets a high bar for being scared. After all, 0:01:49.360 --> 0:01:52.720 the horseless carriage also felt pretty different at the time. 0:01:53.640 --> 0:01:56.920 But two very distinguished economists have taken a fresh look 0:01:56.960 --> 0:02:01.800 at that history and decided the basic reassuring story about 0:02:01.800 --> 0:02:04.560 the past impact of technology on jobs and the quality 0:02:04.560 --> 0:02:10.559 of life of working people is wrong or seriously incomplete. 0:02:10.760 --> 0:02:13.919 So if their argument is right, we should be worried 0:02:13.919 --> 0:02:17.160 about the way AI will transform our economy and society 0:02:17.280 --> 0:02:19.560 because we are not now in a position to get 0:02:19.560 --> 0:02:22.720 the best out of it. Quite the opposite. That book 0:02:22.800 --> 0:02:26.840 is Power and Progress, Our thousand years Struggle over Technology 0:02:26.840 --> 0:02:31.440 and Prosperity, and the authors are darn Smoglu and Simon Johnson, 0:02:31.560 --> 0:02:35.680 both professors at the Massachusetts Institute of Technology. Simon's a 0:02:35.720 --> 0:02:38.960 former chief economist at the International Monetary Fund, and Darren 0:02:39.040 --> 0:02:41.959 won the John Bates Prize for the best US economist 0:02:42.080 --> 0:02:45.360 under forty a few years ago. He also wrote possibly 0:02:45.360 --> 0:02:48.640 the most widely read book of economic history in recent times, 0:02:49.040 --> 0:02:52.240 Why Nations Fail. And he's here with me now, Darren, 0:02:52.480 --> 0:02:54.400 Thank you very much. I'm really pleased we can have 0:02:54.480 --> 0:02:56.160 this conversation on Stephanomics. 0:02:56.280 --> 0:02:58.040 I am so happy to be here, and you give 0:02:58.120 --> 0:03:00.440 such a wonderful introduction. I don't think I have much 0:03:00.480 --> 0:03:01.400 to add, No, no. 0:03:01.320 --> 0:03:03.560 I think we have plenty to discuss, and I'm sorry 0:03:03.600 --> 0:03:05.640 about the long introduction, but I wanted to make clear 0:03:05.680 --> 0:03:08.480 why I thought the book was so important, and we 0:03:08.560 --> 0:03:11.280 do have I think a lot to unpack about the 0:03:11.320 --> 0:03:14.400 past and the present and what you hope might be 0:03:14.639 --> 0:03:16.880 a better future. But maybe we can start at that 0:03:17.040 --> 0:03:20.840 endpoint with what worries you right now about the impact 0:03:20.880 --> 0:03:24.919 that AI and similar technologies might have on the world. 0:03:25.080 --> 0:03:27.520 Well, I think you put it so well. We need 0:03:27.560 --> 0:03:31.720 to be concerned but not scared. I think this is 0:03:31.720 --> 0:03:37.000 a turning point. There are many transformative choices we have 0:03:37.080 --> 0:03:39.520 to make about the future of work, future of inequality, 0:03:39.600 --> 0:03:42.560 future of democracy. And the two worst positions we can 0:03:42.600 --> 0:03:47.320 take are to say everything's going to be fine, just 0:03:47.480 --> 0:03:52.640 let experts worry about it or be scared about killer robots. 0:03:52.720 --> 0:03:56.920 Both of them pacify us and push us not to 0:03:57.000 --> 0:04:01.160 take up our responsibility of trying to shape the technology, 0:04:01.200 --> 0:04:04.440 trying to get involved about decisions about the future of 0:04:04.440 --> 0:04:05.240 this technology. 0:04:05.720 --> 0:04:07.600 And I guess there is a lot there, but one 0:04:07.640 --> 0:04:09.200 piece of it that I was struck by because I 0:04:09.200 --> 0:04:12.440 feel like that was definitely many many of my economic 0:04:12.560 --> 0:04:15.240 history lessons had this sort of underlying premise. And you 0:04:15.280 --> 0:04:18.480 talk about the productivity bandwagon, so maybe you should explain that. 0:04:18.839 --> 0:04:22.520 Let me actually take a step back, Simon, and I 0:04:22.640 --> 0:04:26.599 are definitely not arguing that we haven't tremendously benefited from 0:04:26.640 --> 0:04:30.000 industrial technology and the scientific advances. Today we are so 0:04:30.080 --> 0:04:32.960 much more comfortable, so much more prosperous, so much healthier 0:04:33.000 --> 0:04:35.000 than people who lived three hundred years ago, and that's 0:04:35.040 --> 0:04:38.760 thanks to industrial technology and use of scientific knowledge in 0:04:39.000 --> 0:04:43.359 further improving technology in every aspect of our lives. What 0:04:43.440 --> 0:04:46.480 we are questioning that that was an automatic process. So 0:04:46.560 --> 0:04:49.640 the techno optimism that you so eloquently described at the 0:04:49.680 --> 0:04:53.120 beginning is that we don't need to take any drastic 0:04:53.200 --> 0:04:58.120 actions or become involved in shaping the future of technology 0:04:58.160 --> 0:05:01.440 because there is a very powerful in axorable, automatic process 0:05:01.480 --> 0:05:03.839 that's going to bring all sorts of good things to people. 0:05:03.880 --> 0:05:05.400 And at the center of it is what we call 0:05:05.440 --> 0:05:08.600 the productivity bandwagon. Because most of us earn our livings 0:05:09.000 --> 0:05:12.680 by supplying our labor. So the process via which technology 0:05:12.720 --> 0:05:14.880 is going to improve brings shared prosperity has to go 0:05:14.920 --> 0:05:17.560 through the labor market, which means wages have to improve. 0:05:17.800 --> 0:05:23.279 And the productivity bandwagon says that if productivity grows, if 0:05:23.360 --> 0:05:27.840 technological capabilities improve, that's going to create a very powerful 0:05:27.839 --> 0:05:32.279 force towards employers wanting more labor, and that raises wages 0:05:32.320 --> 0:05:37.040 and employment. If the productivity bandwagon breaks down or doesn't 0:05:37.040 --> 0:05:40.440 have many people on it, then shared prosperity would become 0:05:40.440 --> 0:05:43.560 a dream. And the real fear is that that's exactly 0:05:43.600 --> 0:05:45.880 what could happen with AI, and we've seen some of 0:05:45.920 --> 0:05:48.880 that happen with digital technologies over the last forty years anyway, 0:05:49.400 --> 0:05:54.200 and history says productivity bandwagon can work, but only if 0:05:54.240 --> 0:05:57.160 we create the right institutions and the right direction of technology. 0:05:57.320 --> 0:06:00.679 So what matters most for making an outcome better worse 0:06:00.760 --> 0:06:01.360 for workers? 0:06:01.520 --> 0:06:04.800 Perfect That's exactly the right question. And what Simon and 0:06:04.839 --> 0:06:08.159 I argue is that there are two pillars to it, 0:06:08.200 --> 0:06:12.720 and you can see them very clearly in most historical episodes. First, 0:06:12.760 --> 0:06:15.880 you need a direction of technology that doesn't just automate work. 0:06:15.880 --> 0:06:18.080 Automation is always going to be with us, but it 0:06:18.120 --> 0:06:20.480 doesn't just automate work, but at the same time creates 0:06:20.520 --> 0:06:24.200 new tasks, new capabilities, new things in which human labor 0:06:24.279 --> 0:06:27.400 can be productively used. And second, you need an institutional 0:06:27.400 --> 0:06:31.839 framework in which there are forces such as worker voice 0:06:31.880 --> 0:06:35.520 and worker power that induces employers to share some of 0:06:35.560 --> 0:06:38.560 the gains with workers. If either of those two things 0:06:38.560 --> 0:06:42.680 are broken down, then we're in trouble. If both of 0:06:42.720 --> 0:06:45.640 them break down, that's really damaging. And that's the age 0:06:45.680 --> 0:06:48.159 we are living in. There is no worker voice. AI 0:06:48.320 --> 0:06:50.800 is being used to sideline workers even more in the 0:06:50.839 --> 0:06:54.480 production process, and there isn't a democratic process that's actually 0:06:54.720 --> 0:06:58.479 contributing to it, to a sharing of prosperity, or to 0:06:59.040 --> 0:07:01.440 reshaping the direction of technology. 0:07:01.000 --> 0:07:02.720 You're talking about where we are now, which I think 0:07:02.720 --> 0:07:05.560 we should definitely get to. But for those who sort 0:07:05.600 --> 0:07:08.640 of feel like, oh, this sounds like people who are 0:07:08.720 --> 0:07:14.960 just seeing this new technology and fearing the worst, you know, 0:07:15.000 --> 0:07:17.360 I think at least we should show how it's rooted 0:07:17.400 --> 0:07:20.679 in that understanding of history that you mentioned, and maybe 0:07:21.080 --> 0:07:23.840 draw some contrasts when you're looking back, for example, at 0:07:23.840 --> 0:07:25.400 the nineteenth century. I mean, there's a lot of a 0:07:25.400 --> 0:07:28.560 book which is looking at what happened in the UK 0:07:28.720 --> 0:07:31.120 and the Industrial Revolution, So maybe say a little bit 0:07:31.120 --> 0:07:35.080 about the sort of contrasting impact of the different technologies 0:07:35.120 --> 0:07:35.720 that came in there. 0:07:35.720 --> 0:07:38.720 Thank you them for bringing that up absolutely. I mentioned 0:07:38.760 --> 0:07:43.200 already that we are so fortunate to have had the 0:07:43.240 --> 0:07:48.840 industrial technological improvement that started somewhere in the UK in 0:07:48.880 --> 0:07:52.160 the middle of the eighteenth century. We are so fortunate, 0:07:52.160 --> 0:07:54.760 but the people who live through it weren't. The first 0:07:54.800 --> 0:07:58.560 eighty ninety years of the British Industrial Revolution was dreadful 0:07:58.600 --> 0:08:05.080 for people. Incomes stagnated, working hours, expanded working conditions, Work 0:08:05.280 --> 0:08:09.000 worsened in factories which much greater discipline, much less autonomy. 0:08:09.400 --> 0:08:15.480 People were filled into unhealthy cities in which their life 0:08:15.520 --> 0:08:20.520 expectancy dropped, and there was no worker voice, no democratic process. 0:08:21.200 --> 0:08:25.480 The whole thing was just a very difficult time for 0:08:25.640 --> 0:08:30.880 most working people. But it didn't remain that way in 0:08:30.920 --> 0:08:33.520 the second half of the nineteenth century. You already see 0:08:33.679 --> 0:08:38.439 higher wages, much greater use of technology for improving conditions, 0:08:38.480 --> 0:08:44.000 both as public infrastructure, health improves and factories improved. And 0:08:44.040 --> 0:08:46.960 why why did that happen? Was that automatic? Again, our 0:08:47.000 --> 0:08:49.160 reading of history with a lot of evidence says no, 0:08:49.280 --> 0:08:53.320 that wasn't automatic. There was a complete transformation of British 0:08:53.320 --> 0:09:00.800 institutions with democracy, public sector involvement in cleaning up the cities, 0:09:01.080 --> 0:09:05.320 education and other public infrastructure, and very transformatively trade unions. 0:09:05.360 --> 0:09:08.439 You know, being a unionist was illegal in the ant 0:09:08.720 --> 0:09:11.480 Kingdom and that started changing in the second half of 0:09:11.520 --> 0:09:14.920 the nineteenth century, and that worker voice, worker negotiation were critical. 0:09:15.200 --> 0:09:18.559 As part of that process, the direction of technology changed significantly. 0:09:19.800 --> 0:09:23.600 What brought part of that misery was the automation focus 0:09:23.800 --> 0:09:28.680 and the very high discipline modern factory system. All of 0:09:28.720 --> 0:09:32.440 that started improving. No longer you could allow child labor 0:09:32.679 --> 0:09:36.360 or you know, twelve hour days and mind shaft for people, 0:09:36.520 --> 0:09:39.320 for children as young as five. All of these things 0:09:39.320 --> 0:09:41.800 were institutional in nature as well as technological. 0:09:42.000 --> 0:09:44.280 And actually I was struck by the example. I mean, 0:09:44.320 --> 0:09:46.840 we going up in Britain. You feel like you hear 0:09:46.880 --> 0:09:49.040 about the industrial revolution all the time, but I think 0:09:49.360 --> 0:09:52.319 and indeed about child labor and some of the worst aspects. 0:09:52.320 --> 0:09:54.719 But I think what was in your book you sort 0:09:54.720 --> 0:09:59.080 of bring home what a deterioration in circumstances it was 0:09:59.120 --> 0:10:02.520 that actually six seven year olds hadn't been doing twelve 0:10:02.600 --> 0:10:06.240 hours work, certainly not in the dark underground before, and 0:10:06.280 --> 0:10:08.880 so there was this a period where people's lives were 0:10:08.960 --> 0:10:12.040 actively worse. I think is worth reminding. 0:10:11.640 --> 0:10:15.000 People absolutely absolutely, and people were very exercised about it. 0:10:15.040 --> 0:10:17.840 I mean, at some point it reached such alarming proportion 0:10:18.000 --> 0:10:22.600 that middle class Brits, you know, said disc cannot go on. 0:10:23.840 --> 0:10:27.280 But all wishful thinking would not have done anything unless 0:10:27.280 --> 0:10:29.520 we changed the institutions and the direction of technology, and 0:10:29.559 --> 0:10:31.520 that's what Britain managed in the second half of the 0:10:31.559 --> 0:10:32.400 nineteenth century. 0:10:34.600 --> 0:10:37.480 So there's quite a few things there, because there's the 0:10:37.559 --> 0:10:40.440 nature of the technology and whether it tends to just 0:10:40.520 --> 0:10:44.400 replace workers or actually also produce more demand more other 0:10:44.520 --> 0:10:51.719 kinds of jobs for workers. There's also the institutions surrounding it. 0:10:52.679 --> 0:10:56.640 But part of that is about the companies that are 0:10:56.679 --> 0:10:59.560 producing the technology, that are driving technology, and how powerful 0:10:59.600 --> 0:11:02.599 they are relative to other parts of society. And I 0:11:02.600 --> 0:11:05.199 guess what the example of that is in the Gilded Age, 0:11:04.840 --> 0:11:08.040 in the US. So how does that feed into it 0:11:08.080 --> 0:11:10.199 the sort of market power of companies. 0:11:10.320 --> 0:11:12.040 I think the market power is one of the very 0:11:12.040 --> 0:11:17.040 important elements as well, because new technologies, especially those that 0:11:17.080 --> 0:11:19.800 make better use of labor, come out of the competitive process. 0:11:19.960 --> 0:11:23.320 A more diverse approach to innovation is an important part 0:11:23.360 --> 0:11:26.200 of it. Now, large companies have always been with us. 0:11:26.280 --> 0:11:28.760 I don't think we're going to be able to reverse that, 0:11:28.800 --> 0:11:32.520 and I don't think we should. Ford Motor Company started 0:11:32.559 --> 0:11:35.400 small but became one of the most important employers in 0:11:35.400 --> 0:11:37.600 the United States, and it was at the forefront of 0:11:38.080 --> 0:11:40.120 automating work. But it was also at the forefront of 0:11:40.160 --> 0:11:45.760 creating new tasks, much better working conditions for workers with 0:11:46.080 --> 0:11:49.960 higher wages, and accommodating workers into the production process so 0:11:50.000 --> 0:11:54.800 that they could actually reduce a turnover. But all of 0:11:54.840 --> 0:11:58.040 that becomes much more likely when we have countervailing powers, 0:11:58.200 --> 0:12:02.520 and countervailing powers have to have several sources. For large companies, 0:12:02.559 --> 0:12:06.200 you need competition. If they become so secure that nobody 0:12:06.200 --> 0:12:08.840 can replace them, that's not going to be good. You 0:12:08.880 --> 0:12:12.320 need countervailing powers in the form of worker voice, worker involvement. 0:12:12.720 --> 0:12:15.800 Trade unions provided that, labor movement provided that in the past, 0:12:15.840 --> 0:12:17.920 What will provide it in the future That remains to 0:12:17.920 --> 0:12:20.479 be seen, and you need the government regulation in there. 0:12:20.679 --> 0:12:24.000 You know, if companies can do whatever they want to 0:12:24.200 --> 0:12:29.080 their customers, to the environment, two workers, that's not going 0:12:29.080 --> 0:12:32.800 to lead to good outcomes. So a regulatory framework is 0:12:32.840 --> 0:12:36.280 also quite critical. Overall. I think a good way of 0:12:36.320 --> 0:12:40.120 thinking about this is democratic control of technology. Technology is 0:12:40.120 --> 0:12:43.560 something that affects us all. To say that one or 0:12:43.559 --> 0:12:46.440 two genius in Silicon Valley have to be responsible for 0:12:46.480 --> 0:12:48.320 the future of technology, I mean we all have to 0:12:48.600 --> 0:12:50.560 take whatever it's dished out to us, that's not the 0:12:50.640 --> 0:12:54.120 right perspective. And the democratic control comes from companies being 0:12:54.160 --> 0:12:57.200 at the forefront of technological progress. But those companies are 0:12:57.920 --> 0:13:01.800 threatened by rivals, countable to their workers, and they're accountable 0:13:01.840 --> 0:13:03.560 to society through democratic means. 0:13:03.679 --> 0:13:07.559 Because we do tend to think of invention and technology 0:13:07.880 --> 0:13:11.760 as a sort of as a being out being outside 0:13:11.800 --> 0:13:14.160 the system, that there's you know, people are sitting around 0:13:14.240 --> 0:13:16.880 in their labs or where you know, wherever you can, 0:13:17.080 --> 0:13:20.640 wherever you picture them, or in their garages, coming up 0:13:20.640 --> 0:13:23.440 with their ideas, and there's only it's only after a 0:13:23.480 --> 0:13:27.240 certain point that they're they're interacting with the broader world, 0:13:27.280 --> 0:13:29.840 that there is this sort of natural process of invention 0:13:29.960 --> 0:13:33.360 that happens. It doesn't feel like that process has ever 0:13:33.440 --> 0:13:39.480 been really organized or run by government or with a 0:13:39.559 --> 0:13:41.880 kind of democratic infact. So can you democratic? 0:13:41.960 --> 0:13:45.760 Interesting? You're right at some degree, But there is a 0:13:45.840 --> 0:13:50.880 broader ecosystem. First of all, a lot of innovation is 0:13:50.920 --> 0:13:53.920 coordinated by large companies. Today. If you look at the 0:13:55.040 --> 0:13:58.920 in the United States, most RND is by large and 0:13:58.960 --> 0:14:04.560 publicly traded companies. But second, even the innovation that takes 0:14:04.559 --> 0:14:10.000 place in universities, in people's garages, in small companies, it's 0:14:10.040 --> 0:14:13.360 influenced by the market system where people think profits are, 0:14:13.360 --> 0:14:16.200 and it's influenced by what we call a vision what 0:14:16.520 --> 0:14:20.360 is the best use of our scientific knowledge. And I 0:14:20.400 --> 0:14:25.160 think we have created an incorrect direction of technology because 0:14:25.200 --> 0:14:28.920 of both reasons. We have provided the wrong market incentives 0:14:29.640 --> 0:14:32.280 to digital technologies and we've provided the wrong vision. And 0:14:32.280 --> 0:14:36.600 they've both met in saying digital technologies should be designed 0:14:36.600 --> 0:14:39.720 by geniuses to be imposed on people, and they should 0:14:39.760 --> 0:14:43.000 be used for automation, for surveillance, for data collection, for 0:14:43.240 --> 0:14:47.000 reducing labour's involvement in the production process, for creating some 0:14:47.040 --> 0:14:50.480 sort of amorphous autonomous machine intelligence, and all of those 0:14:50.520 --> 0:14:53.240 are related, and they're the wrong direction. What we call 0:14:53.320 --> 0:14:55.520 for in the book is that we should strive for 0:14:55.640 --> 0:14:58.960 machine usefulness, not machine intelligence. Machines are valuable to us 0:14:59.040 --> 0:15:03.280 because they enable us to do useful things. The calculator Wikipedia, 0:15:03.320 --> 0:15:07.040 those are amazing inventions because they expand what we can do. 0:15:07.600 --> 0:15:10.680 The amorphous notion of AI that is so good that 0:15:10.720 --> 0:15:13.440 it can do everything that humans do, actually in practice 0:15:13.480 --> 0:15:17.320 not so well. But that vision, which guides a lot 0:15:17.360 --> 0:15:18.640 of research, is the wrong one. 0:15:19.480 --> 0:15:22.400 And actually, when you talk about vision, you had a 0:15:22.440 --> 0:15:25.200 fascinating phrase which for me was quite resonant in a 0:15:25.320 --> 0:15:29.280 kind of broader way, which is vision oligarchy. Tell us 0:15:29.280 --> 0:15:30.240 a bit more about that. 0:15:30.400 --> 0:15:32.000 You know, at the end of the day, I described 0:15:32.000 --> 0:15:40.000 a vision which is this machine intelligence created by a 0:15:40.080 --> 0:15:44.640 few very smart engineers and scientists that's going to transform 0:15:44.680 --> 0:15:49.800 everybody's lives. That is a very powerful vision. But where 0:15:49.800 --> 0:15:54.080 did it come from? Well, it came from Turing to 0:15:54.120 --> 0:15:58.640 some degree, but in a very different context. But it 0:15:58.760 --> 0:16:02.760 got operationalized by a number of very like minded people 0:16:02.840 --> 0:16:06.360 in Silicon Valley who've pushed this vision and have achieved 0:16:06.360 --> 0:16:11.040 some degree of commercial success early on and now are 0:16:11.240 --> 0:16:15.880 influencing the rest of society through their oversized role in 0:16:15.960 --> 0:16:20.160 the media, in all public debates, in policy, and of 0:16:20.200 --> 0:16:23.880 course their amazing wealth. And that's what we mean by 0:16:23.960 --> 0:16:27.160 division oligarchy. That's a small group of people who have 0:16:27.680 --> 0:16:30.560 captured the vision of what we can do with technology 0:16:30.600 --> 0:16:32.080 and what we should do with technology. 0:16:32.720 --> 0:16:36.320 And I guess moving on to how we think about 0:16:36.360 --> 0:16:42.600 the more recent waves of technology, you're quite damning about 0:16:42.640 --> 0:16:46.600 the impact that recent automation that AI has had. I 0:16:46.600 --> 0:16:49.720 guess some would say it's just too soon to tell 0:16:50.120 --> 0:16:52.600 how some of these technologies are really going to affect 0:16:52.600 --> 0:16:54.760 the nature of jobs and the workplace. 0:16:54.880 --> 0:16:58.000 That's right, You're right, it's too soon to tell. But 0:16:58.080 --> 0:17:00.880 there is a problem in there. It's too soon to tell. 0:17:00.920 --> 0:17:04.400 Why are we rushing to automate work so quickly? What's 0:17:04.440 --> 0:17:08.600 the rush? So? I have no doubt that automation will 0:17:08.640 --> 0:17:10.760 be part of our future. It has to be part 0:17:10.800 --> 0:17:13.840 of our future. There will be things that machines can 0:17:13.880 --> 0:17:17.000 do better than us, and nothing wrong with that, but 0:17:19.920 --> 0:17:22.440 a we should do that only when they are truly 0:17:22.480 --> 0:17:28.359 better than humans, and in humane way, rearranging work in 0:17:28.400 --> 0:17:32.480 a manner that's consistent with human priorities. And second, we 0:17:32.520 --> 0:17:35.520 should at the same time create better jobs, better tasks 0:17:35.560 --> 0:17:39.560 for humans unique skills. That's the problem. We are rushing 0:17:39.560 --> 0:17:44.320 to automate work, even when it's not so productive customer service. 0:17:45.560 --> 0:17:48.359 It's done by AI in many places, and nobody's happy 0:17:48.400 --> 0:17:52.000 with it. We've displaced workers and we are faced with 0:17:52.040 --> 0:17:54.560 these menus that are supposed to be smart. They never work. 0:17:57.119 --> 0:18:01.359 Productivity gains from that are minimal, perhaps negative. But we're 0:18:01.440 --> 0:18:03.400 rushing to do it. And at the same time we're 0:18:03.400 --> 0:18:05.840 not creating any new tasks in new jobs and new capabilities. 0:18:05.840 --> 0:18:08.600 And we can do that. AI enables us or large 0:18:08.680 --> 0:18:13.080 language models, they have the capacity to help us. As 0:18:13.119 --> 0:18:15.919 you said in your introduction, we're not doing that. 0:18:19.960 --> 0:18:22.840 I think that what you've just said describes a lot 0:18:22.880 --> 0:18:24.960 of the technologies that we've lived with over the last 0:18:25.040 --> 0:18:29.720 few years. But it does feel like CHATGPT and that 0:18:30.040 --> 0:18:36.359 much more interactive technology is different to interact with and 0:18:36.480 --> 0:18:39.800 certainly seems to be learning faster than many of these 0:18:39.840 --> 0:18:43.880 other technologies. People who are making an effort to make 0:18:43.920 --> 0:18:48.240 it part of their lives, whether it's professors or lawyers 0:18:48.359 --> 0:18:54.080 or people working in human resources are finding very quickly 0:18:54.200 --> 0:18:56.639 that it can change the way they do their work. 0:18:56.920 --> 0:18:59.800 So is there something a bit different about the generative AI. 0:19:00.040 --> 0:19:05.320 Well, I would say first, it is impressive, but part 0:19:05.359 --> 0:19:09.360 of the reason why it's impressive is because that's how 0:19:09.359 --> 0:19:14.560 it's been marketed. So what people are impressed by chat 0:19:14.600 --> 0:19:18.440 GIPT is it gives authoritative answers. It can write sonnets 0:19:18.480 --> 0:19:24.600 and poetry. It feels like these are things machines shouldn't 0:19:24.640 --> 0:19:27.200 be able to do, and that's what we're impressed by. 0:19:28.400 --> 0:19:32.480 Don't get me wrong. I do completely believe that large 0:19:32.560 --> 0:19:35.720 language models and generative AI can be used in ways 0:19:35.760 --> 0:19:39.280 that are very positive for humans, and some people have 0:19:39.400 --> 0:19:42.640 found ways of doing that with chat GPT, But chat 0:19:42.680 --> 0:19:47.399 GPT's architecture is not optimal for that. What we want, 0:19:47.920 --> 0:19:53.520 if we believe my pitch for machine usefulness, is that 0:19:53.640 --> 0:19:57.800 these programs should make us better in our jobs, in 0:19:57.840 --> 0:20:03.040 our lives, in our cognition. It doesn't work. If chachipet 0:20:03.160 --> 0:20:06.640 gives you an authoritative answer without explaining to you why 0:20:06.640 --> 0:20:08.680 you should believe it, you either believe it, which is 0:20:08.800 --> 0:20:12.680 not good, or you completely dismiss it. If I were 0:20:12.760 --> 0:20:14.399 to give you an argument, you would ask me, why 0:20:14.440 --> 0:20:16.960 are you saying that, what's your evidence? Where does that 0:20:17.000 --> 0:20:20.080 come from? Give me the provenance of that. You can't 0:20:20.080 --> 0:20:24.120 do that with chatchipt. It's not designed that way. If 0:20:24.160 --> 0:20:27.720 you ask it to provide references, it will make up some. 0:20:28.600 --> 0:20:33.400 It never really processes the reliability of information. It is 0:20:33.480 --> 0:20:36.320 not designed so that it can interact with you in 0:20:36.359 --> 0:20:39.320 a way that filters the vast amount of information that 0:20:39.359 --> 0:20:42.160 you have available. But you don't know which one is reliable. 0:20:42.280 --> 0:20:44.760 So there are many things that we could design these 0:20:44.840 --> 0:20:48.359 machines or these models differently that could be more useful 0:20:48.359 --> 0:20:51.160 to us, but as not the direction the industry is going. 0:20:51.240 --> 0:20:53.360 Many employers are excited by it not to make their 0:20:53.400 --> 0:20:55.679 labor more productive, but they want to eliminate labor. 0:20:56.160 --> 0:20:57.639 And it's interesting because I guess a lot of the 0:20:57.680 --> 0:21:00.520 commentators I was reading something by Ethan Molly other day, 0:21:00.640 --> 0:21:04.639 you know, who are excited about it, have tended to 0:21:04.640 --> 0:21:07.399 be the ones who are finding ways to make it 0:21:07.440 --> 0:21:10.679 more valuable for them. And it doesn't feel like a 0:21:10.680 --> 0:21:12.920 big leap, you know. It comes down to the way 0:21:13.000 --> 0:21:15.640 we interact with it, whether we trust its answers, whether 0:21:15.680 --> 0:21:18.480 we come back. So that doesn't seem like a big change. 0:21:19.520 --> 0:21:23.240 What you've described about changing the whole environment in which 0:21:23.320 --> 0:21:27.760 these technologies are implemented, the whole public attitude towards them. 0:21:28.080 --> 0:21:29.320 That's a pretty big change. 0:21:29.320 --> 0:21:33.359 Absolutely, absolutely, the power attitude regulation. I think these are 0:21:33.440 --> 0:21:37.360 big changes, and you're absolutely right there are people who 0:21:37.400 --> 0:21:40.040 are using chatchipitin a productive where there are some companies 0:21:40.080 --> 0:21:42.919 that have already used in productive ways. But I think 0:21:43.359 --> 0:21:47.439 the model attitude of the corporate world is not the 0:21:47.440 --> 0:21:50.200 healthy one. And that's partly because of the corporate world, 0:21:50.280 --> 0:21:52.320 but a lot also because of the way that the 0:21:52.359 --> 0:21:55.840 technology is structured right now and is marketed right now. 0:21:56.280 --> 0:21:58.960 And we should get into that because you describe certain 0:21:58.960 --> 0:22:01.639 things about it, the way it's been driven towards automation, 0:22:02.359 --> 0:22:07.560 the way that it's led down a path of being 0:22:07.680 --> 0:22:10.600 used for surveillance of individuals, and the impact that that's 0:22:10.600 --> 0:22:14.480 had on democracy. So tell us about that. How you 0:22:14.600 --> 0:22:18.959 think the technology itself has been pushed in a certain 0:22:19.000 --> 0:22:20.920 direction by its origins. 0:22:21.000 --> 0:22:25.000 Well, I think let's start with surveillance. The current field 0:22:25.080 --> 0:22:32.000 of AI is contintely intermingled with data collection, and it 0:22:32.080 --> 0:22:37.760 is hungry for data and employers are hungry for getting 0:22:37.760 --> 0:22:41.439 more information about their workers'. Governments, especially authoritarian governments, are 0:22:41.520 --> 0:22:45.880 hungry for getting more information about dissident activities. So there 0:22:45.960 --> 0:22:52.199 is a confluence of factors that is intensifying monitoring of 0:22:52.320 --> 0:22:54.800 surveillance of both citizens and workers. I think that's one 0:22:54.800 --> 0:22:57.040 of the things that we have to worry about, and 0:22:59.080 --> 0:23:02.080 generative AI is going to push more in that direction. 0:23:02.920 --> 0:23:10.600 Automation is related, but quite a distinct phenomenon. US corporations 0:23:10.640 --> 0:23:13.760 are under pressure competitively because of their shareholders, because of 0:23:13.760 --> 0:23:17.960 the vision of their managers to reduce labor costs. Nobody 0:23:17.960 --> 0:23:20.640 in the schooling system is, for example, talking about, let's 0:23:20.720 --> 0:23:24.040 hire more teachers, give them better tools, make them more skill, 0:23:24.080 --> 0:23:26.040 pay them higher wages so that they can do a 0:23:26.040 --> 0:23:29.400 better job of creating the human capital of the next generation. 0:23:29.440 --> 0:23:31.760 But that's what we need. We need much more individualized 0:23:31.760 --> 0:23:34.040 teaching in the education system. In the United States and 0:23:34.040 --> 0:23:37.760 the United Kingdom, a lot of low socioeconomic background children 0:23:37.840 --> 0:23:41.840 are having trouble getting the right type of education, the 0:23:41.920 --> 0:23:46.200 right type of skills from the schooling system. More individualized 0:23:46.280 --> 0:23:49.480 education targeted to their strengths and weaknesses could be a 0:23:49.560 --> 0:23:52.480 great boot. We can use AI for doing that, but 0:23:52.520 --> 0:23:55.160 nobody is doing that, because that means actually hiring more teachers. 0:23:55.520 --> 0:23:58.760 What schools are interested in hiring less teachers. What companies 0:23:58.800 --> 0:24:01.359