WEBVTT - What If AI Simply Ruins Your Job Instead of Taking It? 0:00:02.720 --> 0:00:07.200 Bloomberg Audio Studios, podcasts, radio news. 0:00:16.960 --> 0:00:19.720 I'm Stephanie Flanders, head of Government and Economics at Bloomberg, 0:00:19.760 --> 0:00:23.120 and this is Trumpnomics, the podcast that looks broadly at 0:00:23.160 --> 0:00:26.320 the economic world of Donald Trump, how he's shaking up 0:00:26.320 --> 0:00:28.800 the global economy and what on earth is going to 0:00:28.840 --> 0:00:31.360 happen next. And we do talk a lot about AI 0:00:31.640 --> 0:00:34.440 on this podcast. In fact, on or off this show, 0:00:34.479 --> 0:00:37.240 I'm finding most conversations end up there sooner or later. 0:00:37.680 --> 0:00:39.920 More often than not, the focus is on how many 0:00:40.000 --> 0:00:43.800 jobs AI will replace and how quickly, also who the 0:00:43.880 --> 0:00:47.840 winners and losers will be. But there's another worry about 0:00:47.880 --> 0:00:50.040 AI that I'm hearing more and more and is vividly 0:00:50.080 --> 0:00:53.159 described in a new book by the always brilliant FT 0:00:53.360 --> 0:00:57.000 columnist Sarah O'Connor. It's the fear that not so much 0:00:57.040 --> 0:00:59.480 that AI will put humans out of work, but it 0:00:59.520 --> 0:01:03.400 will make people's jobs worse than they were before, even 0:01:03.440 --> 0:01:06.080 when they were pretty bad jobs to begin with. What 0:01:06.200 --> 0:01:08.920 if by making our jobs much easier they actually take 0:01:08.959 --> 0:01:12.559 away everything that made them interesting or rewarding. The book 0:01:12.680 --> 0:01:15.880 I'm talking about is we Are Not Machines. The Fight 0:01:16.000 --> 0:01:18.920 for the Future of Work, and Sarah is with me, Sarah. 0:01:18.959 --> 0:01:20.800 Welcome to Trump and Norman. Thank you for having me, 0:01:26.440 --> 0:01:29.040 and thank you for your columns which I'm often sending 0:01:29.200 --> 0:01:31.480 to the team, And thank you for your book, which 0:01:31.520 --> 0:01:34.199 I think does introduce, as often you do with your columns, 0:01:34.200 --> 0:01:38.440 a whole different element. Your title isn't exactly employment correspondent, 0:01:38.480 --> 0:01:40.520 but I wonder whether it should be more kind of 0:01:40.640 --> 0:01:43.440 humanity correspondent these days. That was kind of the sense 0:01:43.480 --> 0:01:45.080 I got from the book. You were focused on the 0:01:45.160 --> 0:01:49.520 human part of jobs and what might happen and in 0:01:49.560 --> 0:01:52.920 some cases, as you describe, has happened when robots take over. 0:01:53.360 --> 0:01:54.960 Yeah, that's right. I think the reason I wrote the 0:01:55.000 --> 0:01:57.000 book was that I was becoming a bit frustrated with 0:01:57.280 --> 0:01:59.440 the sort of the lack of humanity somehow in this 0:01:59.520 --> 0:02:01.840 discussion that even though it is all about us, and 0:02:01.920 --> 0:02:04.880 indeed we humans are the ones who've invented AI, it 0:02:04.920 --> 0:02:07.000 seemed as if the technology was being portrayed as the 0:02:07.000 --> 0:02:09.880 protagonist here, and that much of the debate was being 0:02:09.960 --> 0:02:12.799 driven either by the chief executives of those big technology 0:02:12.800 --> 0:02:15.679 companies and they're big predictions, the utopian ones and the 0:02:15.760 --> 0:02:19.359 dystopian ones, or by economists who have been doing lots 0:02:19.400 --> 0:02:22.240 of very careful studies, as you say, trying to map, 0:02:22.320 --> 0:02:25.680 you know, what are the particular abilities of these new 0:02:25.720 --> 0:02:29.359 AI systems. And if we break apart every single job 0:02:29.400 --> 0:02:32.080 in the economy into its constituent tasks, then can we 0:02:32.120 --> 0:02:34.760 work out which jobs are most exposed. I think that's 0:02:34.800 --> 0:02:37.359 fine as far as it goes, but it doesn't really 0:02:37.440 --> 0:02:39.680 get to the heart of things, which is how might 0:02:39.720 --> 0:02:42.800 it actually feel to be in the position of someone 0:02:42.840 --> 0:02:46.000 who is suddenly presented with this new technology? How does 0:02:46.040 --> 0:02:48.760 it actually change work on the ground. And I'm the 0:02:48.800 --> 0:02:50.919 sort of journalist who likes to get out of the office, 0:02:51.360 --> 0:02:53.320 get my notebook, get my company boots on, and go 0:02:53.360 --> 0:02:55.880 and meet people. So that's what I decided to do. 0:02:55.960 --> 0:02:58.440 So rather than the spreadsheet approach, I wanted to take 0:02:58.440 --> 0:03:03.000 the shoe leather approach. Here went to meet software developers, translators. 0:03:03.160 --> 0:03:05.960 I went down a deep mine in Sweden which is 0:03:06.000 --> 0:03:09.160 now falling with kind of autonomous vehicles, a brand new 0:03:09.200 --> 0:03:11.480 Amazon warehouse which is also full of robots, and I 0:03:11.520 --> 0:03:13.720 just wanted to find out how is this actually going 0:03:13.760 --> 0:03:15.799 and to talk to the people that it's actually happening too. 0:03:16.639 --> 0:03:19.240 And the big picture was you say that you have 0:03:19.400 --> 0:03:22.760 ended up more optimistic than at the start. I don't 0:03:22.760 --> 0:03:25.359 know whether everyone reading the book will feel more optimistic, 0:03:25.680 --> 0:03:27.840 but what were you gloomy about at the beginning and 0:03:27.880 --> 0:03:29.560 how did you change your view? 0:03:29.840 --> 0:03:31.360 I feel like I've been on a real journey with 0:03:31.400 --> 0:03:32.840 this one. I mean I used to be a real 0:03:32.919 --> 0:03:36.000 techno optimist before I even started the book. I basically 0:03:36.080 --> 0:03:38.360 because I've worked at the ETT for a long time. 0:03:38.400 --> 0:03:40.320 I've written a lot about the world of work over 0:03:40.360 --> 0:03:43.120 the last decade decade and a half, and I felt 0:03:43.160 --> 0:03:44.880 like over that time, I'd talked to a lot of 0:03:44.920 --> 0:03:47.640 people who were in jobs which weren't really that good, 0:03:47.720 --> 0:03:51.320 you know, quite boring, quite repetitive, quite physically arduous. Sometimes 0:03:51.640 --> 0:03:55.000 I wrote about the first generation of Amazon warehouses, where 0:03:55.240 --> 0:03:57.800 people were walking around for ten or fifteen miles a day, 0:03:58.320 --> 0:04:01.480 paced by algorithms, tracked by algorithms. They're getting blisters on 0:04:01.520 --> 0:04:03.720 their feet. And I used to think, like, you know, 0:04:03.840 --> 0:04:06.800 this is exactly why we need more automation. You know, 0:04:06.840 --> 0:04:10.400 we should be automating away bad jobs. We should be 0:04:10.440 --> 0:04:13.200 automating away the kind of the dull, the dirty, the 0:04:13.280 --> 0:04:17.080 dangerous stuff, and so you know, bring them on and then, 0:04:17.600 --> 0:04:21.200 of course, the robots did arrive, and so did generative AI, 0:04:21.320 --> 0:04:23.480 and that started to impact all kinds of white collar 0:04:23.680 --> 0:04:26.480 jobs as well. What I started to realize from talking 0:04:26.560 --> 0:04:29.680 to people was that it wasn't necessarily happening the way 0:04:29.680 --> 0:04:33.040 I thought it was, as you said in the introduction, 0:04:33.240 --> 0:04:35.839 rather than being a kind of force for liberation in 0:04:35.920 --> 0:04:38.599 some workplaces and in some professions, it was sort of 0:04:38.640 --> 0:04:41.920 crunching people into systems that were now kind of paced 0:04:41.960 --> 0:04:44.800 by machines, and also where people were kind of having 0:04:44.800 --> 0:04:47.920 to plug the inadequacies of the machines because they're not 0:04:48.000 --> 0:04:51.400 necessarily capable of doing everything that a person can do, 0:04:51.839 --> 0:04:54.359 but they can do certain things. And then you're starting 0:04:54.360 --> 0:04:57.280 to see jobs being redesigned around those sorts of strengths 0:04:57.320 --> 0:04:59.680 and weaknesses in which the humans are now sort of 0:04:59.760 --> 0:05:02.320 in but they're not necessarily in the loop in a 0:05:02.360 --> 0:05:06.040 way that's particularly enjoyable and sometimes was sort of more intense. 0:05:06.720 --> 0:05:08.720 So that made me quite depressed about the future. In 0:05:08.760 --> 0:05:11.600 the end, the reason that I felt more optimistic was 0:05:11.640 --> 0:05:14.240 because I realized, through the course of talking to all 0:05:14.240 --> 0:05:17.919 these people that the idea that this is an inevitability, 0:05:17.920 --> 0:05:20.120 that this is just sort of coming at us and 0:05:20.160 --> 0:05:22.400 that some people will sink and some people will swim, 0:05:22.520 --> 0:05:24.520 or as you say, you know, they'll be winners and losers. 0:05:24.920 --> 0:05:27.120 I realized that that's not actually the case. None of 0:05:27.120 --> 0:05:30.039 the people I met could be sort of slotted into 0:05:30.080 --> 0:05:33.320 a winner or a loser category and an economist's spreadsheet. 0:05:33.520 --> 0:05:37.200 They were all responding to this new technology in various ways. 0:05:37.200 --> 0:05:39.400 Some were trying to take advantage of what was useful 0:05:39.400 --> 0:05:42.400 about it. Others were sort of pivoting away from risks. 0:05:42.839 --> 0:05:46.400 Some people had just decided to leave their profession altogether 0:05:46.480 --> 0:05:49.279 and to create something new for themselves. And it made 0:05:49.279 --> 0:05:51.919 me think that actually we all have more agency in 0:05:51.960 --> 0:05:55.280 this than we're sometimes giving ourselves credit for, and that 0:05:55.360 --> 0:05:59.000 fundamentally people are unbelievably adaptable and they will find their 0:05:59.000 --> 0:06:01.880 ways through, but that we need to all kind of 0:06:01.880 --> 0:06:04.599 be aware that this is a story in which we're 0:06:04.600 --> 0:06:06.560 the protagonists rather than the technologies. 0:06:06.839 --> 0:06:08.600 There's so much there, and I want to get quite 0:06:08.640 --> 0:06:10.200 a lot of it. There's a piece of it that 0:06:10.279 --> 0:06:13.040 I thought that you start talking about because you focus 0:06:13.080 --> 0:06:17.280 on what happened to translation and sort of the creativity 0:06:17.320 --> 0:06:21.080 involved in translating subtitles, which I think if anyone's actually 0:06:21.200 --> 0:06:23.680 been a translator or knows a translator, they know there 0:06:23.760 --> 0:06:26.200 is a real art to it. There's a sort of 0:06:26.279 --> 0:06:30.360 classic model, which is the sort of white collar version 0:06:30.920 --> 0:06:33.400 of what you were describing. You know, you have the 0:06:33.440 --> 0:06:36.000 sort of drudgery, and the drudgery is taken away and 0:06:36.040 --> 0:06:38.960 you're just making someone's job easier. The version of that 0:06:39.000 --> 0:06:40.880 in white collar work, which also applies to kind of 0:06:40.960 --> 0:06:44.200 editors and journalists, is the AI does eighty percent of 0:06:44.240 --> 0:06:47.839 the job and then the human gets to spend less 0:06:47.880 --> 0:06:50.960 time just sort of tweaking it, making a bit better. 0:06:51.120 --> 0:06:55.159 And I suspect a lot of newsrooms and anywhere that 0:06:55.240 --> 0:06:57.760 is involved in translation and other things. That's been the 0:06:57.800 --> 0:07:00.800 way they've introduced AI, and they've said, isn't it nice 0:07:00.839 --> 0:07:02.760 for these people? They don't have to do like the 0:07:02.800 --> 0:07:06.600 first nine yards, They just perfect it and you kind 0:07:06.640 --> 0:07:10.640 of capture that that actually takes away most of what 0:07:10.880 --> 0:07:13.240 was rewarding about that job. It really made me think 0:07:13.240 --> 0:07:13.680 about it. 0:07:14.320 --> 0:07:15.960 Yeah, I mean, one of the things that were so 0:07:16.040 --> 0:07:18.960 lovely about doing the reporting for the book is talking 0:07:19.000 --> 0:07:21.640 to people in depth about what they actually do. And 0:07:21.680 --> 0:07:24.320 it's only when you do that that you realize how 0:07:24.400 --> 0:07:27.760 much kind of care and craft and complicated sort of 0:07:27.800 --> 0:07:30.160 knowledge and skill goes into all kinds of things that 0:07:30.200 --> 0:07:32.560 we often take for granted. So I quite often watch 0:07:32.640 --> 0:07:34.760 TV with the subtitles on it, even in English, just 0:07:34.800 --> 0:07:37.760 because I don't know why my attention span is obviously collapsing. 0:07:38.120 --> 0:07:40.480 But when I talk to these translators of subtitles, they 0:07:40.520 --> 0:07:44.880 talked about what a remarkably creative and enjoyable job it is, 0:07:45.120 --> 0:07:48.800 particularly if you're watching a TV show in English and 0:07:48.840 --> 0:07:50.920 you have to translate it into say, check, which is 0:07:50.920 --> 0:07:53.400 what one of my interviews have to do. You have 0:07:53.440 --> 0:07:56.880 to make so many decisions, you know, like in English, 0:07:56.880 --> 0:07:59.560 there's just one way of saying you, but actually in 0:07:59.560 --> 0:08:02.400 lots of other languages, particularly European languages, there's the formal 0:08:02.440 --> 0:08:04.520 you and there's the informal you, And so you, as 0:08:04.520 --> 0:08:07.760 the translator, have to decide in any given scene, would 0:08:07.800 --> 0:08:10.760 these people be talking in the informal or the formal 0:08:10.880 --> 0:08:13.920 and sometimes that might change as their relationship sort of shifts. 0:08:14.240 --> 0:08:16.360 And those are the sorts of decisions that you have 0:08:16.440 --> 0:08:18.480 to make. And then there's like how do you translate 0:08:18.520 --> 0:08:21.080 a joke that might play really well on the West 0:08:21.120 --> 0:08:24.280 coast of America, so that it still is funny for 0:08:24.320 --> 0:08:25.920 people in the Czech Republic, you know, and all of 0:08:25.960 --> 0:08:28.640 those things are actually a really great creative challenge. 0:08:28.880 --> 0:08:32.080 But you have that great example from Asterisks of Obelisks's 0:08:32.200 --> 0:08:37.240 dog is called Dogmatics, but in French it's Ed Fichs, 0:08:37.320 --> 0:08:39.800 which I just is a fantastic example exactly. 0:08:39.920 --> 0:08:42.840 Yeah, so Dave Fichs was the original name of the dog, 0:08:42.920 --> 0:08:45.880 and then the English translators came up with Dogmatics, which 0:08:45.920 --> 0:08:48.240 is even a better pun. Right. It was so fun 0:08:48.280 --> 0:08:50.240 to talk to those translators about what they love about 0:08:50.240 --> 0:08:52.760 their job. But yeah, what's changing for them isn't necessarily 0:08:52.760 --> 0:08:56.240 that they're being completely cut out of the equation, because 0:08:57.000 --> 0:08:59.520 these subtitles are still not kind of perfect if they're 0:08:59.520 --> 0:09:01.640 done by miss but as you say, they're being left 0:09:01.760 --> 0:09:05.160 to They call it now machine translation post editing, so 0:09:05.160 --> 0:09:07.640 you're given the machine translation that you're expected to tidy 0:09:07.720 --> 0:09:09.679 it up to check that it's accurate. And what they 0:09:09.720 --> 0:09:13.000 said was that actually this is no longer creative, and 0:09:13.040 --> 0:09:15.880 also it's cognitively quite challenging actually because you're sort of 0:09:15.960 --> 0:09:19.400 checking one thing against the other. But it's lost the 0:09:19.480 --> 0:09:21.720 sort of the joy and the meaning and the pleasure. 0:09:22.120 --> 0:09:24.840 And also they're being paid half the price, and so 0:09:24.880 --> 0:09:26.800 they have to do it twice the pace, and so 0:09:26.840 --> 0:09:29.720 there's been a sort of intensification of their work as well, 0:09:29.760 --> 0:09:31.520 and so some of them said that it felt a 0:09:31.559 --> 0:09:33.960 bit more like being on a production line compared to 0:09:34.400 --> 0:09:36.360 what it used to be. I mean, I would say 0:09:36.400 --> 0:09:38.160 to people who are starting to feel quite depressed. I 0:09:38.160 --> 0:09:42.439 did also go and visit some much more interesting, optimistic 0:09:42.679 --> 0:09:46.079 workplaces where I think AI and other kinds of automation 0:09:46.120 --> 0:09:49.160 technologies are sort of living up to that promise that 0:09:49.200 --> 0:09:51.480 we all had, which was that it would free us 0:09:51.559 --> 0:09:53.760 up to be more human. And then the key question, 0:09:53.800 --> 0:09:56.080 I think is like, what do you do with that. 0:09:56.160 --> 0:09:59.160 I didn't really want to write a book that concludes, well, 0:09:59.640 --> 0:10:03.000 there's a opportunities and dangers. You know, there's winners and losers. 0:10:03.000 --> 0:10:04.480 I've read a lot of books like that, as I'm 0:10:04.480 --> 0:10:06.680 sure you have as well. And it's not that that's 0:10:06.720 --> 0:10:07.600 not true, but I think. 0:10:07.480 --> 0:10:09.880 You have to talk about uncertainty as well. There's the 0:10:09.920 --> 0:10:11.920 winners and losers, and it's all very answering. 0:10:11.559 --> 0:10:14.480 And it's all terribly uncertain exactly. But that's not very helpful, 0:10:14.520 --> 0:10:16.520 is it to say there'll be winners and losers. The 0:10:16.600 --> 0:10:19.120 kind of more difficult question is the next one, which is, well, 0:10:19.160 --> 0:10:22.120 what makes the difference? Why are some people winning and 0:10:22.160 --> 0:10:25.240 why are some people losing? Why is this exact same 0:10:25.360 --> 0:10:28.920 set of technologies playing out so differently in different people's work. 0:10:29.000 --> 0:10:31.000 And I think unless you try and challenge yourself to 0:10:31.040 --> 0:10:33.320 ask that question, we're not really going to get much 0:10:33.360 --> 0:10:35.240 further in terms of shaping this the way we would 0:10:35.280 --> 0:10:36.000 quite like it to go. 0:10:36.400 --> 0:10:40.000 And I think also you get a little bit into 0:10:40.480 --> 0:10:43.800 even when someone might be a winner in economic terms 0:10:44.240 --> 0:10:46.720 because they're in a well prayed profession where they have 0:10:46.800 --> 0:10:49.320 quite a lot of autonomy over the use you mentioned that, 0:10:49.360 --> 0:10:51.200 you know, that's part of the thing is do you 0:10:51.320 --> 0:10:56.160 have control over how it's used. You're voluntarily adopting these 0:10:56.200 --> 0:10:59.400 tools and they are making you much more productive, often 0:10:59.520 --> 0:11:01.760 in sort of coding or in any of those kind 0:11:01.840 --> 0:11:04.480 of areas. But as you highlight, and you know people 0:11:04.559 --> 0:11:08.839 have written about in real time that intensification thing, people 0:11:08.920 --> 0:11:11.480 end up feeling they're working much harder. And of course 0:11:11.480 --> 0:11:14.840 we've seen that in a kind of macro sense for 0:11:14.880 --> 0:11:19.520 the last hundred years. Famously, Canes had said productivity is 0:11:19.559 --> 0:11:21.640 going to be increased, you know, quintuple in the next 0:11:21.760 --> 0:11:23.760 hundred years, and so we'll all have lots of leisure time. 0:11:23.840 --> 0:11:26.480 Of course, it turns out, yes, the people who've become 0:11:26.559 --> 0:11:29.199 much more productive, many of them are also quite well off, 0:11:29.520 --> 0:11:32.040 but they feel like they're working harder than never and 0:11:32.080 --> 0:11:34.840 they're worried about how many hours sleep. So I think 0:11:34.880 --> 0:11:38.560 even that definition of winners and losers, once you start 0:11:38.600 --> 0:11:41.400 thinking about the quality of life and the quality of 0:11:41.480 --> 0:11:44.400 your job, it becomes a bit harder to tell which 0:11:44.440 --> 0:11:44.800 is which. 0:11:45.000 --> 0:11:47.800 Definitely, it's very much the case that you can win 0:11:47.920 --> 0:11:50.600 something and lose something at the same time, and I 0:11:50.640 --> 0:11:54.480 think that we sometimes don't notice necessarily what we're losing 0:11:54.600 --> 0:11:57.160 until it's too late, or until we sort of look 0:11:57.240 --> 0:12:00.480 back and think, huh, something about the quality of my 0:12:00.679 --> 0:12:02.600 work doesn't feel quite the same anymore, But I can't 0:12:02.640 --> 0:12:05.680 put my finger on exactly what it is. Software developers 0:12:05.720 --> 0:12:08.640 are really an interesting example because in many ways it's 0:12:08.640 --> 0:12:11.719 sort of the opposite of those translators. If anything, I 0:12:11.720 --> 0:12:14.720 think those jobs are becoming more human in the sense 0:12:14.760 --> 0:12:17.640 that there's less sitting there typing out code by hand. 0:12:17.640 --> 0:12:20.400 I mean no, no software developers that I know are 0:12:20.400 --> 0:12:23.040 writing code by hand anymore, and so what is now 0:12:23.080 --> 0:12:25.680 demanded of them is actually more kind of managerial skills, 0:12:26.000 --> 0:12:28.199 those sort of more people skills. They've got to coordinate 0:12:28.360 --> 0:12:31.240 between different AI agents and different humans, They've got to 0:12:31.280 --> 0:12:33.440 think about the big picture a bit more, and so 0:12:33.520 --> 0:12:36.120 it's starting to put a premium on those more thoughtful 0:12:36.320 --> 0:12:39.520 judgment taste, all of those sorts of things, which does 0:12:39.559 --> 0:12:41.760 sound more enjoyable, and I think a lot of software 0:12:41.760 --> 0:12:44.000 developers are enjoying it and are kind of certainly a 0:12:44.040 --> 0:12:47.280 bit intoxicated by it, But as you say, they're not 0:12:47.520 --> 0:12:51.559 sort of banking that productivity and going. You know, it's 0:12:51.600 --> 0:12:53.640 interesting that both the sort of the utopian and the 0:12:53.720 --> 0:12:57.880 dystopian narratives about AI envisit just doing less work. Right 0:12:57.960 --> 0:13:00.520 in the utopian story, we're all sitting in our hammocks 0:13:00.520 --> 0:13:02.600 and working two hours a day in writing poetry to 0:13:02.640 --> 0:13:05.120 each other. And in the dystopian story, there's like ten 0:13:05.200 --> 0:13:09.520 people who've become multi billionaires and everyone else's mass unemployed. 0:13:10.120 --> 0:13:12.880 But actually, as you say, what is seems to be 0:13:12.880 --> 0:13:15.120 happening is that we're just finding more and more stuff 0:13:15.160 --> 0:13:16.839 to do with our time, which is often the way 0:13:16.840 --> 0:13:18.000 it goes. Well. 0:13:18.080 --> 0:13:20.000 Another thing I felt was sort of in the back 0:13:20.040 --> 0:13:21.760 of my mind as I was reading the book, and 0:13:21.800 --> 0:13:23.880 it's you know, one of those tweets that went viral 0:13:23.920 --> 0:13:25.600 a couple of years ago along the lines of I 0:13:25.640 --> 0:13:27.920 wanted AI to do my laundry and dishes so I 0:13:27.920 --> 0:13:29.960 could write a novel, but it seems to be working 0:13:29.960 --> 0:13:32.040 the other way around. There is that sort of concern. 0:13:32.400 --> 0:13:35.800 I think your point is the novels are not very good. 0:13:36.480 --> 0:13:39.320 You get a lot of management consultants now saying because 0:13:39.360 --> 0:13:43.679 it's so easy to do AI written reports and research, 0:13:44.120 --> 0:13:47.600 the premium will be on interaction with humans and being 0:13:47.640 --> 0:13:50.520 able to prove that you've had humans involved using their 0:13:50.600 --> 0:13:53.439 judgments and other things. I mean, that would be great 0:13:53.679 --> 0:13:57.440 if it's true, but at least some of the examples 0:13:57.480 --> 0:14:00.839 in your book, it suggests that what's produced by AI 0:14:01.040 --> 0:14:05.160 may be objectively less good than what humans do, but 0:14:05.720 --> 0:14:07.880 it'll still end up replacing the humans. 0:14:08.960 --> 0:14:11.520 I think it's definitely the case that you can't just 0:14:11.720 --> 0:14:14.880 take what the AI can do, or what any kind 0:14:14.880 --> 0:14:17.880 of automation technology can do, compare it to what a 0:14:17.920 --> 0:14:20.440 human does, and then say, well, This is how the 0:14:20.520 --> 0:14:23.040 job will change, because so much of it depends, as 0:14:23.040 --> 0:14:26.680 you say, on what will managers be satisfied with, what 0:14:26.760 --> 0:14:30.480 will consumers be satisfied with. So it's very possible that 0:14:30.880 --> 0:14:34.400 we might accept subtitles that are objectively worse but far 0:14:34.520 --> 0:14:37.720 far cheaper. Now. Translation is actually a really good example, 0:14:37.760 --> 0:14:40.440 because it's really hard to know if you're on the 0:14:40.440 --> 0:14:43.160 receiving end of a translation whether the quality has got 0:14:43.160 --> 0:14:44.800 worse or not. The whole point is that you don't 0:14:44.800 --> 0:14:47.480 know what the original language was. And there have been 0:14:47.480 --> 0:14:50.720 some studies of the quality of subtitles, for example, over time, 0:14:51.080 --> 0:14:53.800 and since this new kind of machine translation post editing 0:14:53.880 --> 0:14:56.920 system has come in, the quality has dropped quite substantially. 0:14:57.160 --> 0:15:00.520 The language is less rich, less varied. Everything is becoming 0:15:00.560 --> 0:15:03.080 sort of a bit thinner. There's sort of odd punctuation, 0:15:03.320 --> 0:15:06.200 it's all a bit less human. But as you say, 0:15:06.320 --> 0:15:08.520 from the studios point of view, if that's much much 0:15:08.600 --> 0:15:10.280 cheaper and no one is really going to be in 0:15:10.280 --> 0:15:13.160 a position to know, then that might happen anyway, even 0:15:13.200 --> 0:15:18.120 if the humans are still superior. I'm hopeful that ultimately 0:15:18.200 --> 0:15:19.840 there will be a sort of pushback, and I think 0:15:19.840 --> 0:15:21.720 we're already starting to see that. There are a lot 0:15:21.800 --> 0:15:25.800 of people who are saying, I don't want music that 0:15:25.960 --> 0:15:28.120 is made by AI. I don't want to read a 0:15:28.160 --> 0:15:30.880 book that is written with the help of AI. And 0:15:30.880 --> 0:15:32.480 there are lots of people trying to come up with 0:15:32.520 --> 0:15:35.200 like labeling schemes, you know, made by humans, all of 0:15:35.200 --> 0:15:37.840 that sort of thing. I think the tricky thing is 0:15:37.840 --> 0:15:41.600 that the way AI integrates into people's creative process, it's 0:15:41.680 --> 0:15:43.560 quite subtle and quite messy, and there's a sort of 0:15:43.640 --> 0:15:47.520 huge gradient between how much did you use AI, whether 0:15:47.560 --> 0:15:50.320 you're a writer or a researcher, particularly, you know, if 0:15:50.320 --> 0:15:52.920 you're working in something like TV, if it's used in 0:15:52.920 --> 0:15:55.480 post production, does that count? Does that not count? And 0:15:55.520 --> 0:15:57.840 so I'm not sure that this labeling idea is going 0:15:57.920 --> 0:16:00.600 to be the thing, but it's definitely I think there 0:16:00.640 --> 0:16:04.040 is a demand from humans for stuff made by other humans, 0:16:04.040 --> 0:16:06.880 and so I'm hopeful that in some way the market 0:16:06.880 --> 0:16:09.360 and capitalism will figure out a way to associate that demand. 0:16:21.560 --> 0:16:23.960 I should say we're recording this on the sixteenth of June. 0:16:24.000 --> 0:16:26.400 I'm not entirely sure when you'll be hearing it, but 0:16:26.960 --> 0:16:29.440 one of the columns you wrote was asking why is 0:16:29.480 --> 0:16:32.800 it so controversial what the impact of the Industrial Revolution is, 0:16:32.800 --> 0:16:36.320 why after all these years, we suddenly fighting with renewed 0:16:36.400 --> 0:16:41.640 vigor about how the Industrial Revolution affected the economy in society. 0:16:41.720 --> 0:16:44.360 And the point you made there, I guess was people 0:16:44.360 --> 0:16:46.400 were looking in the wrong place when they were just 0:16:46.440 --> 0:16:48.560 looking at the numbers. And I think we've also had 0:16:48.560 --> 0:16:50.720 that actually on the show from people like Simon Johnson 0:16:50.760 --> 0:16:53.720 and Darrenis and Moglu. If you read the historians or 0:16:53.760 --> 0:16:57.080 the novelists of that nineteenth century, it was about how 0:16:57.080 --> 0:17:00.760 it changed people's lives as well as the jobs and 0:17:00.880 --> 0:17:02.800 people out of work exactly. 0:17:02.840 --> 0:17:04.679 And also if you look at the kind of fights 0:17:04.680 --> 0:17:07.480 that broke out politically and even on the streets and 0:17:07.520 --> 0:17:11.360 certainly inside workplaces during the Industrial Revolution, often they were 0:17:11.359 --> 0:17:14.680 not about real wages and whether real wages have ticked 0:17:14.720 --> 0:17:17.040 up point five percent or point seven percent. They were 0:17:17.040 --> 0:17:21.760 about things like craftsmanship, dignity, health and safety, the role 0:17:21.760 --> 0:17:23.679 of children in the workplace. You know, all of these 0:17:23.720 --> 0:17:26.200 things are actually much more fundamental to kind of who 0:17:26.240 --> 0:17:28.800 we are and how we relate to our work. And 0:17:28.840 --> 0:17:31.520 I think we're beginning to see all of those same 0:17:31.720 --> 0:17:33.960 debates now and none of them are really going to 0:17:34.000 --> 0:17:36.679 be captured in these questions about where the productivity has 0:17:36.720 --> 0:17:38.760 gone up by zero point seven or point nine in 0:17:38.760 --> 0:17:40.440 the US in the last quarter, So. 0:17:40.480 --> 0:17:44.280 You talking about the nineteenth century. It's funny because I 0:17:44.320 --> 0:17:45.800 did a story a long time ago at the BBC 0:17:46.520 --> 0:17:50.359 about how housekeepers in Las Vegas, all the hotels in 0:17:50.440 --> 0:17:53.639 Las Vegas, the people cleaning the rooms and the hotels 0:17:53.760 --> 0:17:56.360 are all unionized, which is kind of surprising, but that 0:17:56.440 --> 0:17:59.000 was from a very conscious effort by the service sector. 0:17:59.080 --> 0:18:02.920 Union leader had said to me, Look, these great jobs, 0:18:02.960 --> 0:18:06.480 the manufacturing jobs that people are now trying of fighting 0:18:06.520 --> 0:18:10.040 to keep in cars industry, for example, that have great 0:18:10.080 --> 0:18:12.280 benefits and people can have pensions and they can send 0:18:12.320 --> 0:18:15.280 their kids to college. They didn't just become out of 0:18:15.359 --> 0:18:17.800 thin air. Those jobs were terrible one hundred and twenty 0:18:17.880 --> 0:18:20.119 years ago, and it was unions that made them better. 0:18:20.760 --> 0:18:24.680 And of course, in these kind of very human centric jobs, 0:18:24.760 --> 0:18:27.560 we've tended to think that they were very hard to unionize. 0:18:27.560 --> 0:18:29.760 And this is one of the not very many examples 0:18:29.880 --> 0:18:32.199 the housekeepers in Las Vegas, is that part of the 0:18:32.240 --> 0:18:36.160 answer that these jobs that definitely won't be automated need 0:18:36.200 --> 0:18:38.960 to become sort of good blue collar jobs in the 0:18:39.000 --> 0:18:43.040 way that say, being a plumber or even working in 0:18:43.160 --> 0:18:45.560 a car factory are good jobs. 0:18:46.280 --> 0:18:48.800 Yeah, I think that would be a good outcome. I 0:18:48.800 --> 0:18:51.400 mean it's tricky, I think because if you look at 0:18:51.400 --> 0:18:53.560 something like care work, for example, and I spend a 0:18:53.600 --> 0:18:57.560 whole chapter with these amazing community nurses in the Netherlands. 0:18:58.680 --> 0:19:02.760 The reason that it's difficult to sort of raise wages 0:19:02.800 --> 0:19:06.200 in the care sector is partly because they're not unionized, 0:19:06.240 --> 0:19:08.080 but there's some other things going on there as well. 0:19:08.200 --> 0:19:10.600 One of them is that often the people who are 0:19:10.600 --> 0:19:13.520 paying for that are taxpayers in one form or another, 0:19:13.680 --> 0:19:18.960 and we have struggled to put aside enough money really 0:19:19.000 --> 0:19:22.439 to pay for good quality jobs in those sectors, partly 0:19:22.480 --> 0:19:25.720 because the economies in many places are aging and so 0:19:25.800 --> 0:19:27.800 the costs are just going up. There are lots of 0:19:27.800 --> 0:19:31.520 other kind of competing demands on taxpayers money. And then 0:19:31.560 --> 0:19:33.880 the other thing, of course, is that with manufacturing, it's 0:19:33.880 --> 0:19:36.560 definitely true that unions helped workers get a share of 0:19:36.600 --> 0:19:39.439 the pie, but also automation did enable productivity to go up, 0:19:39.480 --> 0:19:42.199 so there was a growing pie to be divided, and 0:19:42.440 --> 0:19:47.119