1 00:00:15,410 --> 00:00:22,370 Speaker 1: Pushkin. Tim here with an exciting announcement about the Cautionary Club, 2 00:00:22,690 --> 00:00:26,850 Speaker 1: our Patreon community for Cautionary Tales listeners. So many of 3 00:00:26,850 --> 00:00:30,210 Speaker 1: you joined our live table read earlier this year that 4 00:00:30,330 --> 00:00:33,610 Speaker 1: we're doing it again. Join me and I production team 5 00:00:33,770 --> 00:00:37,690 Speaker 1: for a live reading of an unreleased episode about the 6 00:00:37,730 --> 00:00:43,410 Speaker 1: people who almost invented the iPhone fifteen years early and 7 00:00:43,450 --> 00:00:46,770 Speaker 1: the surprising reasons they failed. This will be a chance 8 00:00:46,850 --> 00:00:49,010 Speaker 1: for you to see how the stories we tell are 9 00:00:49,010 --> 00:00:53,490 Speaker 1: developed in real time and ask your burning questions about 10 00:00:53,490 --> 00:00:57,010 Speaker 1: Cautionary Tales. It's on the eighth of July at five 11 00:00:57,050 --> 00:01:01,090 Speaker 1: pm UK or noon Eastern. If you join the Patreon 12 00:01:01,170 --> 00:01:03,770 Speaker 1: before then, or if you're already a member, you'll get 13 00:01:03,770 --> 00:01:09,810 Speaker 1: your exclusive invitation. Head to Patreon dot com slash Cautionary 14 00:01:09,850 --> 00:01:13,450 Speaker 1: Club that Patreon p A t R e O N 15 00:01:13,770 --> 00:01:25,490 Speaker 1: dot com slash Cautionary Club. Missus Goodair is enjoying a 16 00:01:25,570 --> 00:01:29,690 Speaker 1: leisurely breakfast while she reads the morning paper. It's April 17 00:01:29,930 --> 00:01:34,410 Speaker 1: eighteen twelve, and the headlines are dominated by Wellington's military 18 00:01:34,410 --> 00:01:40,010 Speaker 1: successes in Spain. A strange noise jolts her from her reading. 19 00:01:40,770 --> 00:01:44,650 Speaker 1: It sounds like the iron gates to her home rattling. 20 00:01:46,650 --> 00:01:49,730 Speaker 1: She rushes upstairs to get a better view and sees 21 00:01:49,770 --> 00:01:53,410 Speaker 1: the mob for it is a mob. More clearly, they're 22 00:01:53,410 --> 00:01:58,050 Speaker 1: throwing stones, making lude gestures, and shouting for her to 23 00:01:58,130 --> 00:02:02,250 Speaker 1: open the gates. Two men appear to be wearing dresses, 24 00:02:02,690 --> 00:02:08,450 Speaker 1: proclaiming loudly to be General Lud's wives. She knows exactly 25 00:02:08,450 --> 00:02:12,050 Speaker 1: what this is about. It's about what's in her husband's 26 00:02:12,130 --> 00:02:16,570 Speaker 1: mill next door. So why are they at her house? 27 00:02:18,010 --> 00:02:22,450 Speaker 1: Her husband, John Goodair, is out of town. She hovers 28 00:02:22,450 --> 00:02:26,770 Speaker 1: at the window, unsure of what to do. The mob 29 00:02:26,850 --> 00:02:30,450 Speaker 1: leaders grow impatient. They shout to their comrades that it's 30 00:02:30,490 --> 00:02:33,770 Speaker 1: time to move on and begin marching through the streets 31 00:02:33,810 --> 00:02:37,970 Speaker 1: of Stockport. As they weave through the town, they smash 32 00:02:38,050 --> 00:02:42,890 Speaker 1: windows and break down factory doors. By noon, the mob 33 00:02:42,930 --> 00:02:48,970 Speaker 1: has swelled over two thousand men. Soon they run out 34 00:02:48,970 --> 00:02:52,530 Speaker 1: of targets, so they turned back to where they began, 35 00:02:53,490 --> 00:02:58,250 Speaker 1: John Goodair's enormous mill, the place many of them have 36 00:02:58,330 --> 00:03:02,410 Speaker 1: spent their working lives. Once a symbol of pride in 37 00:03:02,450 --> 00:03:06,650 Speaker 1: their craft, now a symbol of everything they feel is 38 00:03:06,770 --> 00:03:11,850 Speaker 1: wrong with the industrialized world. They break down the doors, 39 00:03:12,490 --> 00:03:16,970 Speaker 1: smash through the windows, and destroy all they can inside 40 00:03:17,690 --> 00:03:22,810 Speaker 1: every last frame, every last spindle. The mob has grown 41 00:03:23,130 --> 00:03:26,250 Speaker 1: so large that even the iron gates to the Good 42 00:03:26,250 --> 00:03:31,970 Speaker 1: Air Home can't hold them back. Missus Goodair has already 43 00:03:31,970 --> 00:03:38,050 Speaker 1: gone a wise move. The Luddites burn the house to 44 00:03:38,130 --> 00:03:43,810 Speaker 1: the ground. I'm Tim Harford and you're listening to Cautionary 45 00:03:43,850 --> 00:04:08,770 Speaker 1: Tales live at the Bristol Festival of Economics. Loyal listeners 46 00:04:08,770 --> 00:04:11,690 Speaker 1: will remember this is not the first time the Luddites 47 00:04:11,770 --> 00:04:15,090 Speaker 1: have featured on this podcast. Two years ago I told 48 00:04:15,130 --> 00:04:18,730 Speaker 1: you about the failure of the Luddite Revolution, the weavers 49 00:04:18,770 --> 00:04:24,010 Speaker 1: who were incensed that automation was devaluing their craft in general, 50 00:04:24,090 --> 00:04:29,570 Speaker 1: Lud's rage against the machines. Today, the Ludites seem more 51 00:04:29,610 --> 00:04:34,730 Speaker 1: relevant than ever as the so called Ai revolution threatens 52 00:04:34,770 --> 00:04:38,890 Speaker 1: to overturn the world of work. So should we be 53 00:04:39,130 --> 00:04:43,850 Speaker 1: smashing up large language models or is resisting as futile 54 00:04:43,850 --> 00:04:46,410 Speaker 1: as it was for the Luddites? Who's going to be 55 00:04:46,490 --> 00:04:51,730 Speaker 1: most affected? And have economists been asking the wrong questions 56 00:04:51,890 --> 00:04:57,050 Speaker 1: all along? I'm joined by a very special guest, Sarah O'Connor. 57 00:04:57,250 --> 00:05:00,810 Speaker 1: She's a Financial Times journalist with a specialism in technology 58 00:05:00,890 --> 00:05:04,730 Speaker 1: and work, and she's written a book called we are 59 00:05:04,770 --> 00:05:09,490 Speaker 1: not machines the fight for the future of work. She 60 00:05:09,650 --> 00:05:12,610 Speaker 1: is the perfect person to guide us through what's coming next, 61 00:05:13,050 --> 00:05:15,690 Speaker 1: and she's got plenty of cautionary tales to tell us. 62 00:05:16,450 --> 00:05:18,890 Speaker 2: Welcome Sarah, Hello Tim, Thank you for having me. 63 00:05:19,010 --> 00:05:22,010 Speaker 1: Oh it's a pleasure. We've got to get something out 64 00:05:22,010 --> 00:05:25,770 Speaker 1: of the way first. So, Sarah O'Connor, have you seen 65 00:05:25,770 --> 00:05:31,570 Speaker 1: the Terminator movies yet? No, you're gonna have to watch them. 66 00:05:31,810 --> 00:05:33,490 Speaker 2: It's a point of principle. 67 00:05:33,090 --> 00:05:34,810 Speaker 1: That, yeah, it now feels like stubbornness. 68 00:05:34,970 --> 00:05:37,250 Speaker 2: Yes, so you might have noticed that my name bears 69 00:05:37,250 --> 00:05:40,730 Speaker 2: a faint resemblance to the heroine of the Terminator films. 70 00:05:40,810 --> 00:05:43,290 Speaker 2: Growing up, occasionally people made Terminator jokes to me, and 71 00:05:43,290 --> 00:05:45,810 Speaker 2: as a result, I never watched them on principle because 72 00:05:45,850 --> 00:05:48,290 Speaker 2: it annoyed me. And so about a decade ago, I 73 00:05:48,370 --> 00:05:53,210 Speaker 2: went accidentally sort of globally viral when I very simply 74 00:05:53,290 --> 00:05:56,370 Speaker 2: tweeted a news story that i'd seen come out, which 75 00:05:56,450 --> 00:05:58,810 Speaker 2: was that a robot had killed a worker in a 76 00:05:58,850 --> 00:06:02,730 Speaker 2: Volkswagen factory, and didn't realize that this would cause the 77 00:06:02,730 --> 00:06:06,650 Speaker 2: Internet to break. So yeah, so people just found it 78 00:06:06,770 --> 00:06:10,810 Speaker 2: very hilarious that someone called Sarah O'Connor had written about robots, 79 00:06:10,850 --> 00:06:13,970 Speaker 2: and for about a decade people have occasionally sent me, 80 00:06:14,570 --> 00:06:18,330 Speaker 2: you know, terminator gifts posters of the Terminator films with 81 00:06:18,370 --> 00:06:21,730 Speaker 2: my face photoshopped in, and as a result, no, I 82 00:06:21,730 --> 00:06:22,650 Speaker 2: have not watched those. 83 00:06:22,810 --> 00:06:27,850 Speaker 1: Okay, So the book is it's fantastic, we are not machines. 84 00:06:28,330 --> 00:06:30,690 Speaker 1: What's the headline? What are you basically driving out with 85 00:06:30,730 --> 00:06:31,210 Speaker 1: this book? 86 00:06:31,410 --> 00:06:34,130 Speaker 2: I think what I'm trying to say in the book 87 00:06:34,210 --> 00:06:37,810 Speaker 2: is that, you know, we're all sort of drowning in 88 00:06:37,810 --> 00:06:43,050 Speaker 2: information predictions about AI and what it might do to 89 00:06:43,090 --> 00:06:44,810 Speaker 2: the world of work, and a lot of those are 90 00:06:44,850 --> 00:06:49,010 Speaker 2: coming from either economists or the big tech executives who 91 00:06:49,090 --> 00:06:54,370 Speaker 2: have created large language models and are marketing them. 92 00:06:53,490 --> 00:06:55,970 Speaker 1: And both these sources are unreliable for different reasons. 93 00:06:56,450 --> 00:06:59,490 Speaker 2: Yeah, the tech guys have something to sell, right, and 94 00:06:59,530 --> 00:07:04,970 Speaker 2: so they have their own motivations. And then you know, economists, 95 00:07:05,050 --> 00:07:07,170 Speaker 2: I think are trying their best to figure out what's 96 00:07:07,170 --> 00:07:10,250 Speaker 2: going on. But the pay which they're doing that I 97 00:07:10,250 --> 00:07:12,850 Speaker 2: think is very feels very abstract to me as someone 98 00:07:12,930 --> 00:07:15,170 Speaker 2: you know, I'm a reporter at heart, So my favorite 99 00:07:15,210 --> 00:07:17,250 Speaker 2: thing to do is to like get my notebook, put 100 00:07:17,250 --> 00:07:19,970 Speaker 2: my boots on and like go and meet people and 101 00:07:20,490 --> 00:07:22,810 Speaker 2: stand outside the factory gates and talk to people about 102 00:07:22,810 --> 00:07:25,730 Speaker 2: what's actually happening. And so over the past few years, 103 00:07:25,730 --> 00:07:28,250 Speaker 2: I've just felt very frustrated that everything that I'm hearing 104 00:07:28,290 --> 00:07:30,970 Speaker 2: about this is coming from people who are either looking 105 00:07:30,970 --> 00:07:34,970 Speaker 2: at spreadsheets and models or who have a product to sell, 106 00:07:35,370 --> 00:07:37,050 Speaker 2: and I just wanted to hear less about what these 107 00:07:37,130 --> 00:07:39,210 Speaker 2: men say is going to happen, and more about what's 108 00:07:39,290 --> 00:07:42,410 Speaker 2: already actually happening from the people that it's actually happening too. 109 00:07:42,810 --> 00:07:45,370 Speaker 2: So in the book, I basically went to find people 110 00:07:45,610 --> 00:07:48,570 Speaker 2: and places and workplaces that are on the front lines 111 00:07:48,610 --> 00:07:52,690 Speaker 2: of what's already happening. So people who are encountering autonomous 112 00:07:52,770 --> 00:07:56,370 Speaker 2: robots in their workplaces, or are working with or for 113 00:07:57,090 --> 00:08:00,170 Speaker 2: or around large luggage models, and they're starting to change 114 00:08:00,410 --> 00:08:02,530 Speaker 2: the way they work. And when I did that, what 115 00:08:02,610 --> 00:08:05,370 Speaker 2: I realized was that you, no, I wasn't really hearing 116 00:08:05,530 --> 00:08:09,330 Speaker 2: this very utopian story that this is liberating me, this 117 00:08:09,570 --> 00:08:13,450 Speaker 2: is freeing me from dull, dirty, dangerous work and making 118 00:08:13,810 --> 00:08:16,290 Speaker 2: everything fantastic. But nor was I really hearing a lot 119 00:08:16,330 --> 00:08:19,010 Speaker 2: of stories that were like my job has been completely 120 00:08:19,090 --> 00:08:21,850 Speaker 2: wiped out. What I was hearing instead were stories about 121 00:08:22,410 --> 00:08:25,890 Speaker 2: quite profound changes to the nature of people's jobs and 122 00:08:25,930 --> 00:08:28,690 Speaker 2: how they felt about their jobs and whether they enjoyed 123 00:08:28,730 --> 00:08:30,770 Speaker 2: their jobs or not. And some of those were quite 124 00:08:31,170 --> 00:08:32,250 Speaker 2: cautionary tales. 125 00:08:32,450 --> 00:08:32,690 Speaker 3: You know. 126 00:08:32,890 --> 00:08:35,730 Speaker 2: There were a lot of people who were saying, actually, 127 00:08:35,810 --> 00:08:38,370 Speaker 2: what's happening to me is almost the opposite of this 128 00:08:38,490 --> 00:08:41,570 Speaker 2: idea that technology is going to give us space to 129 00:08:41,610 --> 00:08:44,170 Speaker 2: be more human. People were finding themselves sort of crunched 130 00:08:44,210 --> 00:08:47,890 Speaker 2: into systems that were paced by machines run by machines, 131 00:08:47,930 --> 00:08:52,530 Speaker 2: in which judgment was being overtaken by machines, in which 132 00:08:52,570 --> 00:08:55,970 Speaker 2: creativity was being sort of compressed or contained by machines. 133 00:08:56,250 --> 00:08:58,370 Speaker 2: And I thought, actually, if that's what's beginning to happen 134 00:08:58,410 --> 00:09:00,570 Speaker 2: to people, then everyone should know about that, and we 135 00:09:00,570 --> 00:09:03,450 Speaker 2: should try and talk about why that's happening and whether 136 00:09:03,490 --> 00:09:06,170 Speaker 2: there is a way to avoid that, because fundamentally, I 137 00:09:06,210 --> 00:09:07,490 Speaker 2: don't think that's what anybody wants. 138 00:09:07,650 --> 00:09:09,490 Speaker 1: Before we sort of take a trip to to the 139 00:09:09,490 --> 00:09:12,370 Speaker 1: cutting edge, I just I wanted to go back a 140 00:09:12,410 --> 00:09:15,090 Speaker 1: couple of hundred years we began with the ludd heightes. 141 00:09:15,170 --> 00:09:17,810 Speaker 1: Are they still relevant? Do people misunderstand the story of 142 00:09:17,850 --> 00:09:18,410 Speaker 1: the loudtes? 143 00:09:18,530 --> 00:09:18,690 Speaker 3: Yeah? 144 00:09:18,690 --> 00:09:20,650 Speaker 2: I think they are still relevant. I mean, Luddite is 145 00:09:20,770 --> 00:09:23,490 Speaker 2: just now used as a sort of a flippant term 146 00:09:23,530 --> 00:09:26,730 Speaker 2: for somebody who's sort of reflexively anti technology and just 147 00:09:26,730 --> 00:09:29,290 Speaker 2: doesn't like any new tech. But actually the ludd Heites, 148 00:09:29,410 --> 00:09:32,850 Speaker 2: they weren't really fighting against technology full stop. They were 149 00:09:33,010 --> 00:09:36,810 Speaker 2: fighting against the way in which machines were being put 150 00:09:36,850 --> 00:09:39,810 Speaker 2: to a new use, which was to cut them out 151 00:09:39,810 --> 00:09:43,610 Speaker 2: of the equation. These sort of knitting and wide frame machines. 152 00:09:43,650 --> 00:09:47,010 Speaker 2: They weren't displacing work entirely. What they did was they 153 00:09:47,050 --> 00:09:50,490 Speaker 2: allowed people who had no skill and no experience to 154 00:09:50,570 --> 00:09:54,090 Speaker 2: make a less good quality product. And in that sense, 155 00:09:54,130 --> 00:09:56,290 Speaker 2: as we might come onto it, actually has an awful 156 00:09:56,330 --> 00:09:59,690 Speaker 2: lot of similarities with some of the technologies that we're 157 00:09:59,730 --> 00:10:00,330 Speaker 2: seeing today. 158 00:10:00,490 --> 00:10:03,290 Speaker 1: Instead of knitting a stocking, you knit a massive bit 159 00:10:03,330 --> 00:10:05,930 Speaker 1: of cloth that you can make up punch stockings out of, 160 00:10:05,970 --> 00:10:07,730 Speaker 1: and you just cut them into strips and then you 161 00:10:07,730 --> 00:10:09,850 Speaker 1: make the stockings. The stockings the terrible they fall apart. 162 00:10:10,090 --> 00:10:13,890 Speaker 1: So you've suddenly got interesting, well paid jobs being replaced 163 00:10:13,890 --> 00:10:19,010 Speaker 1: by boring, badly paid jobs. You've got expensive, high quality 164 00:10:19,610 --> 00:10:22,130 Speaker 1: work being replaced by cheap, crappy work. 165 00:10:22,330 --> 00:10:25,450 Speaker 2: Even if they were lower quality. They were also a 166 00:10:25,530 --> 00:10:27,250 Speaker 2: much lower price, and that I think that was a 167 00:10:27,250 --> 00:10:29,210 Speaker 2: deal that a lot of consumers were willing to make. 168 00:10:29,690 --> 00:10:33,330 Speaker 1: So let's talk about warehouses. The picture I have in 169 00:10:33,370 --> 00:10:35,290 Speaker 1: my mind, or the picture I had in my mind 170 00:10:35,370 --> 00:10:38,290 Speaker 1: before I read the book, was, in fact, I realized 171 00:10:38,330 --> 00:10:42,210 Speaker 1: a picture painted for me by you about ten years ago. 172 00:10:42,690 --> 00:10:45,770 Speaker 1: One of the things that some workers in these had 173 00:10:45,930 --> 00:10:49,570 Speaker 1: was they had this earpiece, the Jennifer unit, basically just 174 00:10:49,610 --> 00:10:51,650 Speaker 1: a voice in their ear telling them where to go 175 00:10:51,690 --> 00:10:54,450 Speaker 1: in the warehouse, what to pull off the shelves. You 176 00:10:54,490 --> 00:10:56,890 Speaker 1: don't need them to think about the best way to 177 00:10:56,930 --> 00:10:59,770 Speaker 1: get around the warehouse or to remember anything, just tobay 178 00:10:59,770 --> 00:11:01,930 Speaker 1: the voice in your head. So this sort of creation 179 00:11:02,010 --> 00:11:05,610 Speaker 1: of these flesh robots. So I was reading this article 180 00:11:05,650 --> 00:11:07,570 Speaker 1: by you about ten years ago and thinking that that 181 00:11:07,690 --> 00:11:10,770 Speaker 1: sounds bad. But of course that's ten years ago. So 182 00:11:11,250 --> 00:11:15,890 Speaker 1: how accurate is that image of what is now happening? 183 00:11:15,890 --> 00:11:16,570 Speaker 1: What's changed? 184 00:11:17,090 --> 00:11:19,290 Speaker 2: Yeah, so quite a lot has changed. And actually I 185 00:11:19,330 --> 00:11:22,770 Speaker 2: think that article from ten years ago and that sort 186 00:11:22,810 --> 00:11:24,850 Speaker 2: of reporting that I've done, And just to clarify that 187 00:11:24,970 --> 00:11:27,450 Speaker 2: Jennifer unit isn't something that's used in Amazon warehouses, but 188 00:11:27,490 --> 00:11:30,450 Speaker 2: it is used in some other warehouses really informed the 189 00:11:30,490 --> 00:11:32,810 Speaker 2: way I used to think about technology. So, yeah, I 190 00:11:32,850 --> 00:11:34,450 Speaker 2: had met and interviewed a lot of people over the 191 00:11:34,490 --> 00:11:36,330 Speaker 2: years who worked in jobs in which they were basically 192 00:11:36,410 --> 00:11:40,850 Speaker 2: expected to work like robots. I used to think, well, 193 00:11:41,090 --> 00:11:43,330 Speaker 2: you know, bring on the down robots. Then you know, like, 194 00:11:43,370 --> 00:11:45,650 Speaker 2: let's get real robots in to do these jobs, like 195 00:11:45,690 --> 00:11:48,530 Speaker 2: this is a waste of kind of human potential, and 196 00:11:48,890 --> 00:11:51,490 Speaker 2: this is a great example of where automation would be brilliant. 197 00:11:51,530 --> 00:11:53,410 Speaker 2: And so I used to be quite a techno optimist 198 00:11:53,450 --> 00:11:55,570 Speaker 2: in the sense that I thought there were huge numbers 199 00:11:55,570 --> 00:11:58,570 Speaker 2: of roles that actually could and should be automated. So 200 00:11:58,610 --> 00:12:02,890 Speaker 2: what's happened now in Amazon warehouses, at least in the 201 00:12:02,890 --> 00:12:06,050 Speaker 2: one that I went to visit, is that the robots 202 00:12:06,090 --> 00:12:09,170 Speaker 2: have arrived, but not really in the way that I had. 203 00:12:09,410 --> 00:12:11,970 Speaker 2: So I'd basically imagined like a one for one swap, right, 204 00:12:12,010 --> 00:12:14,450 Speaker 2: so instead of a person walking around, you'd have like 205 00:12:14,450 --> 00:12:17,290 Speaker 2: a kind of human humanoid robot doing it. 206 00:12:17,730 --> 00:12:20,170 Speaker 1: But it's never actually like that, is it. It's never 207 00:12:20,250 --> 00:12:24,250 Speaker 1: ever like that, Like, a robot accountant is basically Microsoft exell. 208 00:12:24,290 --> 00:12:27,010 Speaker 1: A robot accountant is not C three post sitting in 209 00:12:27,010 --> 00:12:29,050 Speaker 1: your chair top. 210 00:12:28,970 --> 00:12:31,570 Speaker 2: No, exactly, and it's not how it works in Amazon either. 211 00:12:31,650 --> 00:12:35,050 Speaker 2: So I went to visit a new Amazon warehouse which 212 00:12:35,370 --> 00:12:38,490 Speaker 2: has their sort of latest automation technology. And the way 213 00:12:38,530 --> 00:12:41,010 Speaker 2: it works now is that rather than you, as a 214 00:12:41,050 --> 00:12:43,610 Speaker 2: worker walking around for like maybe ten miles a day 215 00:12:43,730 --> 00:12:46,970 Speaker 2: just weaving between all of these shelves and picking things off, 216 00:12:47,490 --> 00:12:50,250 Speaker 2: now on each floor there's a big fence and around 217 00:12:50,290 --> 00:12:54,570 Speaker 2: the perimeter of that fence, the workers stand stationary in 218 00:12:54,610 --> 00:12:56,530 Speaker 2: one place and inside the fence of the robots. But 219 00:12:56,530 --> 00:12:58,930 Speaker 2: the robots are basically like giant rumbers, and what they 220 00:12:58,970 --> 00:13:01,210 Speaker 2: do is they drive around inside this perim defence and 221 00:13:01,210 --> 00:13:04,130 Speaker 2: they pick up the shelves and they bring the shelves 222 00:13:04,170 --> 00:13:06,810 Speaker 2: to the workers. And so you now, as a worker, 223 00:13:06,850 --> 00:13:08,770 Speaker 2: you stand in one place and a robot brings you 224 00:13:08,810 --> 00:13:12,250 Speaker 2: a shirt, and then a light illuminates which part of 225 00:13:12,250 --> 00:13:14,250 Speaker 2: the shelf you need to look at, and a screen 226 00:13:14,330 --> 00:13:15,810 Speaker 2: gives you a picture of what the thing is you 227 00:13:15,850 --> 00:13:17,770 Speaker 2: need to pick off. And I tried this out. I 228 00:13:17,770 --> 00:13:19,610 Speaker 2: think I had to pick off like a mobile phone 229 00:13:19,650 --> 00:13:21,290 Speaker 2: case with a dolphin on it, so you know, you 230 00:13:21,330 --> 00:13:24,130 Speaker 2: pick it out, and then another light illuminates which box 231 00:13:24,170 --> 00:13:25,570 Speaker 2: that you need to put it into and you push 232 00:13:25,570 --> 00:13:28,370 Speaker 2: a big button and then that robot takes that shelf away, 233 00:13:28,410 --> 00:13:30,250 Speaker 2: and the next robot's already queuing up for you with 234 00:13:30,290 --> 00:13:31,930 Speaker 2: the next shelf, and you do the same thing. 235 00:13:32,130 --> 00:13:33,370 Speaker 1: How long did you do it? For? 236 00:13:33,650 --> 00:13:34,290 Speaker 2: Two minutes? 237 00:13:34,530 --> 00:13:36,490 Speaker 1: Okay? How long? How long do they do it for? 238 00:13:36,610 --> 00:13:39,570 Speaker 2: Ten hours a day, forty hours a week with two 239 00:13:39,650 --> 00:13:42,170 Speaker 2: thirty minute breaks? And so this is not what I 240 00:13:42,250 --> 00:13:45,370 Speaker 2: had imagined when I sort of cheered on the robots 241 00:13:45,370 --> 00:13:47,530 Speaker 2: and thought this would be great. You know, in some 242 00:13:47,570 --> 00:13:50,050 Speaker 2: ways I think that job is better. It is less 243 00:13:50,090 --> 00:13:52,930 Speaker 2: physically demanding than walking around for ten hours a day. 244 00:13:53,330 --> 00:13:55,930 Speaker 2: Amazon says that it's safer so they have, you know, 245 00:13:56,090 --> 00:13:59,250 Speaker 2: fewer injuries and accidents, and I think that's probably true. 246 00:13:59,770 --> 00:14:02,210 Speaker 2: But if anything, I think it's also become more boring 247 00:14:02,530 --> 00:14:05,970 Speaker 2: and more monotonous. And also it's still not that easy 248 00:14:06,530 --> 00:14:08,410 Speaker 2: on the human body to stand up for ten hours 249 00:14:08,450 --> 00:14:11,810 Speaker 2: a day, and because they're now much more productive, right 250 00:14:11,850 --> 00:14:14,770 Speaker 2: because you're not walking around from one place to the next, 251 00:14:14,770 --> 00:14:17,370 Speaker 2: and so the products are coming at you, so you're bending, 252 00:14:17,410 --> 00:14:20,250 Speaker 2: you're lifting, you're twisting over and over and over again. 253 00:14:20,290 --> 00:14:22,410 Speaker 2: So I think it's still quite quite a physically demanding 254 00:14:22,530 --> 00:14:25,690 Speaker 2: job and arguably like a bit more monotonous and a 255 00:14:25,690 --> 00:14:26,530 Speaker 2: bit lonelier. 256 00:14:27,010 --> 00:14:28,850 Speaker 1: Presumably there are people who did the old job and 257 00:14:28,890 --> 00:14:30,450 Speaker 1: now do the new job, So what do they think. 258 00:14:31,050 --> 00:14:34,970 Speaker 2: Well, I interviewed one worker who had volunteer to transfer 259 00:14:35,050 --> 00:14:39,370 Speaker 2: from a manual, old fashioned warehouse to a new one, 260 00:14:39,530 --> 00:14:41,530 Speaker 2: and he wanted to go back. He said, actually, I 261 00:14:41,530 --> 00:14:44,090 Speaker 2: don't like working with the robots. It's too intense, it's 262 00:14:44,130 --> 00:14:46,810 Speaker 2: too exhausting, it's really lonely. I don't get to talk 263 00:14:46,850 --> 00:14:49,730 Speaker 2: to anyone anymore. And he yeah, he's now transferred back 264 00:14:49,770 --> 00:14:51,450 Speaker 2: to his old warehouse. 265 00:14:51,730 --> 00:14:56,050 Speaker 1: By way of contrast, you also visited a mine in Sweden, 266 00:14:56,330 --> 00:14:59,090 Speaker 1: So how is automation taking off there? 267 00:14:59,890 --> 00:15:02,730 Speaker 2: Yeah, so this mine in Sweden was right up near 268 00:15:02,770 --> 00:15:06,290 Speaker 2: the fringe of the Arctic Circle and their mine copper 269 00:15:06,330 --> 00:15:10,050 Speaker 2: there and various other minerals, and it is one of 270 00:15:10,090 --> 00:15:12,970 Speaker 2: the most sort of technologically advanced mines in the world, 271 00:15:13,010 --> 00:15:14,930 Speaker 2: or so they say. So it used to be that 272 00:15:14,970 --> 00:15:19,610 Speaker 2: miners would drive vehicles around below. Grounds are very deep down. 273 00:15:19,930 --> 00:15:22,650 Speaker 2: It's horrible down there. It's like really dark, it's really 274 00:15:22,770 --> 00:15:26,730 Speaker 2: humid and claustrophobic. And now the vehicles are autonomous, so 275 00:15:26,770 --> 00:15:30,530 Speaker 2: they drive themselves, and the miners sit in a control room. 276 00:15:30,850 --> 00:15:32,850 Speaker 2: They can take over the vehicles if they need to, 277 00:15:32,930 --> 00:15:34,970 Speaker 2: but a lot of the time they're just they're sitting there, 278 00:15:35,010 --> 00:15:37,250 Speaker 2: They're watching what's happening on big screens. They're in a 279 00:15:37,250 --> 00:15:40,850 Speaker 2: comfy chair, and they're listening to Spotify, they're listening to 280 00:15:41,490 --> 00:15:43,650 Speaker 2: the ice hockey. And the ones that I spoke to 281 00:15:44,090 --> 00:15:47,250 Speaker 2: were quite satisfied with this change in their working conditions. 282 00:15:47,490 --> 00:15:50,330 Speaker 1: Is the fact that the Swedish mineworkers are happy and 283 00:15:50,370 --> 00:15:54,650 Speaker 1: the Amazon workers are at best ambivalent about the way 284 00:15:54,690 --> 00:15:57,250 Speaker 1: their job works. What explains the difference. 285 00:15:57,610 --> 00:15:59,930 Speaker 2: So I think there's a few differences. One difference is that, 286 00:16:00,530 --> 00:16:02,690 Speaker 2: you know, the reason that in Amazon they don't have 287 00:16:02,770 --> 00:16:05,570 Speaker 2: humanoids walking around doing the whole thing is that that 288 00:16:05,730 --> 00:16:09,090 Speaker 2: technology is not ready yet and so in some ways 289 00:16:09,210 --> 00:16:12,530 Speaker 2: are still kind of plugged into a system which is 290 00:16:12,970 --> 00:16:17,250 Speaker 2: like part robot, part human, but is increasingly being paced 291 00:16:17,330 --> 00:16:20,370 Speaker 2: by the capabilities of the robots, Whereas I guess for 292 00:16:20,450 --> 00:16:23,690 Speaker 2: the miners, they do not need to be in the 293 00:16:23,690 --> 00:16:26,890 Speaker 2: cab anymore, and so that means that there's the potential 294 00:16:26,890 --> 00:16:30,410 Speaker 2: for a much kind of greater improvement in their health 295 00:16:30,410 --> 00:16:32,210 Speaker 2: and safety. But as well as that, I think there 296 00:16:32,290 --> 00:16:35,370 Speaker 2: is a governance thing. So I don't know how much 297 00:16:35,410 --> 00:16:38,770 Speaker 2: everyone knows about Sweden, but they have a very distinctive 298 00:16:38,890 --> 00:16:42,690 Speaker 2: kind of labor market whereby trade unions are very powerful, 299 00:16:43,330 --> 00:16:46,250 Speaker 2: and there is a rule that any new technology has 300 00:16:46,290 --> 00:16:49,290 Speaker 2: to be bargained collectively about before it happens. And so 301 00:16:50,090 --> 00:16:52,330 Speaker 2: before they introduced this new technology, they had to sit 302 00:16:52,370 --> 00:16:54,210 Speaker 2: down with the workers and they had to talk about 303 00:16:54,210 --> 00:16:56,570 Speaker 2: it and how it would go and what the workers' 304 00:16:56,610 --> 00:16:59,890 Speaker 2: concerns were and what the company wanted to get out 305 00:16:59,890 --> 00:17:03,330 Speaker 2: of it. And you know, more broadly in Scandinavian countries, 306 00:17:03,370 --> 00:17:06,250 Speaker 2: what you find is that when people answer surveys about 307 00:17:06,250 --> 00:17:09,210 Speaker 2: how they feel about technology, people are like much more 308 00:17:09,210 --> 00:17:11,690 Speaker 2: positive than they are in the UK or in the US, 309 00:17:11,810 --> 00:17:13,330 Speaker 2: and I think that's because they have a sense that 310 00:17:13,370 --> 00:17:15,410 Speaker 2: they will have a seat at the table when those 311 00:17:15,410 --> 00:17:16,330 Speaker 2: decisions get made. 312 00:17:16,610 --> 00:17:20,450 Speaker 1: Thank you, Sarah. You're listening to a special cautionary conversation 313 00:17:20,610 --> 00:17:24,370 Speaker 1: recorded live at the Bristol Festival of Economics. And after 314 00:17:24,410 --> 00:17:28,530 Speaker 1: the break, my guests Sarah O'Connor will be telling a 315 00:17:28,690 --> 00:17:42,770 Speaker 1: cautionary tale about language translators at artificial intelligence. Stay with us, 316 00:17:45,570 --> 00:17:49,050 Speaker 1: we're back this is the Bristol Festival of Economics. I 317 00:17:49,090 --> 00:17:53,010 Speaker 1: am Tim Harford and my special guest is Financial Times 318 00:17:53,090 --> 00:17:59,930 Speaker 1: journalist and author Sarah O'Connor. So, Sarah, perhaps the canaries 319 00:17:59,970 --> 00:18:03,850 Speaker 1: in the coal mine are the translators. When I thought 320 00:18:03,890 --> 00:18:07,210 Speaker 1: about translators, I just thought, oh, you guys are just 321 00:18:07,250 --> 00:18:09,490 Speaker 1: all You've lost all your jobs because of Google Translate. 322 00:18:09,570 --> 00:18:13,290 Speaker 1: It's all gone. But actually your the story you tell 323 00:18:13,290 --> 00:18:15,010 Speaker 1: in the book is is a lot more interesting and 324 00:18:15,050 --> 00:18:15,890 Speaker 1: more complex than that. 325 00:18:16,170 --> 00:18:20,210 Speaker 2: Yeah. So I interviewed some translators for my book. One 326 00:18:20,210 --> 00:18:22,490 Speaker 2: of them is a guy called Peter. He lives in 327 00:18:22,490 --> 00:18:26,370 Speaker 2: the Czech Republic and he translates subtitles for TV shows, 328 00:18:26,930 --> 00:18:28,770 Speaker 2: which is a great job. I mean, he told me 329 00:18:28,810 --> 00:18:31,210 Speaker 2: lots of things about why it's more difficult than you 330 00:18:31,290 --> 00:18:34,450 Speaker 2: might imagine to do that. So, for example, you know, 331 00:18:34,490 --> 00:18:36,530 Speaker 2: in English, we just have one way of saying you, 332 00:18:36,730 --> 00:18:39,610 Speaker 2: but in lots of languages, including Zech, there's like a 333 00:18:39,690 --> 00:18:42,010 Speaker 2: formal tense of you and an informal tense of you. 334 00:18:42,130 --> 00:18:45,570 Speaker 2: And so when you're writing the translated subtitles, you have 335 00:18:45,650 --> 00:18:48,850 Speaker 2: to think like, how well do these characters know each other? 336 00:18:48,890 --> 00:18:51,010 Speaker 2: Like what tense would they be speaking in? And that 337 00:18:51,090 --> 00:18:53,170 Speaker 2: might change, you know, through the course of an episode 338 00:18:53,250 --> 00:18:55,090 Speaker 2: or through the course of a series, and so there 339 00:18:55,130 --> 00:18:57,250 Speaker 2: are lots of kind of interesting judgment calls that you 340 00:18:57,730 --> 00:19:00,370 Speaker 2: have to make. And then, you know, translating jokes is 341 00:19:00,410 --> 00:19:04,130 Speaker 2: another thing that's incredibly difficult because so many jokes are 342 00:19:04,170 --> 00:19:07,090 Speaker 2: like culture specific or they're like plays on language. 343 00:19:07,170 --> 00:19:09,650 Speaker 1: We're in a bookshop where the guests of war systems 344 00:19:09,690 --> 00:19:12,690 Speaker 1: in Bristol. There must be some asterisks and obeliqs somewhere 345 00:19:12,690 --> 00:19:13,290 Speaker 1: in the bookshop. 346 00:19:13,370 --> 00:19:14,730 Speaker 2: Yeah, they're the perfect example. 347 00:19:14,810 --> 00:19:15,010 Speaker 1: Right. 348 00:19:15,090 --> 00:19:20,090 Speaker 2: So Obelix's dog in the original French is called ede feaks, 349 00:19:20,210 --> 00:19:24,290 Speaker 2: which means obsession, and the brilliant English translation is dogmatics. 350 00:19:24,850 --> 00:19:27,210 Speaker 2: You know. So like, there's so much sort of creativity 351 00:19:27,250 --> 00:19:30,050 Speaker 2: and humor involved in being a really good translator, and 352 00:19:30,090 --> 00:19:31,930 Speaker 2: so like, this is a great job for people that 353 00:19:32,130 --> 00:19:34,490 Speaker 2: enjoy doing that sort of thing. But what's happened to 354 00:19:34,570 --> 00:19:37,010 Speaker 2: him and to people like him, is that he has 355 00:19:37,050 --> 00:19:40,410 Speaker 2: not lost his job entirely. Like, it's actually quite hard 356 00:19:40,450 --> 00:19:43,290 Speaker 2: to get a machine to translate things with that level 357 00:19:43,330 --> 00:19:47,650 Speaker 2: of cultural knowledge and understanding. So, just to explain slightly 358 00:19:47,690 --> 00:19:50,130 Speaker 2: the way a lot of translation works is a lot 359 00:19:50,130 --> 00:19:53,610 Speaker 2: of translators of freelancers, and there are agencies that will 360 00:19:53,610 --> 00:19:57,850 Speaker 2: take work from you know, a TV studio and then 361 00:19:57,890 --> 00:20:00,290 Speaker 2: parcel it out to freelance translators. And what those agencies 362 00:20:00,330 --> 00:20:03,930 Speaker 2: have started to do is to take something, get it 363 00:20:03,970 --> 00:20:06,890 Speaker 2: translated by a large language model or a different kind 364 00:20:06,930 --> 00:20:10,050 Speaker 2: of machine translation service, and then it to the translator 365 00:20:10,090 --> 00:20:11,970 Speaker 2: and say, hey, can you just check that this is 366 00:20:12,050 --> 00:20:14,210 Speaker 2: right and maybe like just finesse it a little bit, 367 00:20:14,250 --> 00:20:16,890 Speaker 2: polish it, make it sound a bit more human. And 368 00:20:16,930 --> 00:20:18,930 Speaker 2: this has become the kind of bete noir of lots 369 00:20:18,970 --> 00:20:23,090 Speaker 2: of translators, because what they say is that it's a 370 00:20:23,650 --> 00:20:26,290 Speaker 2: you know, you're expected to do this much faster and 371 00:20:26,330 --> 00:20:28,530 Speaker 2: be you're expected to do it for much less money. 372 00:20:29,010 --> 00:20:31,210 Speaker 2: But in fact, if you care about quality, it's very 373 00:20:31,210 --> 00:20:35,330 Speaker 2: difficult to actually check a translation is accurate, and then 374 00:20:35,370 --> 00:20:38,010 Speaker 2: even things like trying to make it sound a bit 375 00:20:38,050 --> 00:20:40,330 Speaker 2: better like that's actually what the translators told me was, 376 00:20:40,370 --> 00:20:42,930 Speaker 2: it's actually quite hard to do when there's already like 377 00:20:42,930 --> 00:20:45,450 Speaker 2: an answer in front of you, and it feels much 378 00:20:45,530 --> 00:20:49,050 Speaker 2: less creative and more kind of cumbersome and less enjoyable. 379 00:20:49,170 --> 00:20:52,210 Speaker 2: And so a job that that was sort of creative 380 00:20:52,250 --> 00:20:58,730 Speaker 2: and challenging and interesting, has become faster, harder, more monotonous, 381 00:20:58,770 --> 00:21:03,330 Speaker 2: feels more mechanical, and fundamentally kind of less satisfying. And 382 00:21:03,370 --> 00:21:05,050 Speaker 2: I think that is a real cautionary tale. And when 383 00:21:05,050 --> 00:21:06,650 Speaker 2: I was speaking to them, it made me think a 384 00:21:06,690 --> 00:21:09,890 Speaker 2: lot of the luod ites, because what the translators will 385 00:21:09,890 --> 00:21:12,530 Speaker 2: tell you is that the end product is less good 386 00:21:12,610 --> 00:21:15,530 Speaker 2: quality than There are even some studies that have checked this. 387 00:21:15,730 --> 00:21:18,130 Speaker 2: So there's a big study that looked at a bunch 388 00:21:18,170 --> 00:21:22,210 Speaker 2: of subtitles from TV shows before and after the introduction 389 00:21:22,330 --> 00:21:26,090 Speaker 2: of machine translation post editing, and the quality has kind 390 00:21:26,090 --> 00:21:27,010 Speaker 2: of deteriorated. 391 00:21:27,370 --> 00:21:28,370 Speaker 1: It's the stockings again. 392 00:21:28,490 --> 00:21:29,930 Speaker 2: It's the stockings again, exactly. 393 00:21:30,210 --> 00:21:33,170 Speaker 1: And I guess partly obviously we care about the quality 394 00:21:33,170 --> 00:21:36,930 Speaker 1: of jobs. That's important. But I guess one of the 395 00:21:36,970 --> 00:21:40,850 Speaker 1: interesting questions from the point of view of the consumer, 396 00:21:41,370 --> 00:21:45,210 Speaker 1: because the whole idea here is even if individuals get 397 00:21:45,250 --> 00:21:50,890 Speaker 1: worse jobs, the consumer gets more choice, more quality, lower prices. 398 00:21:51,810 --> 00:21:56,130 Speaker 1: I always do sort of wander is the product actually 399 00:21:56,170 --> 00:22:00,930 Speaker 1: better and will the market deliver kind of the right 400 00:22:01,010 --> 00:22:02,970 Speaker 1: trade offs or are we actually just going to make 401 00:22:03,250 --> 00:22:06,130 Speaker 1: all kinds of mistakes and we end up with crappy 402 00:22:06,130 --> 00:22:09,890 Speaker 1: products that we don't want. I'm just wondering whether I 403 00:22:09,890 --> 00:22:12,610 Speaker 1: know it don't sound like an economist here, but I'm 404 00:22:12,610 --> 00:22:14,890 Speaker 1: wondering whether we're all just going to get stockings that 405 00:22:14,930 --> 00:22:19,770 Speaker 1: fall apart and translations that we that are joyless. But 406 00:22:19,850 --> 00:22:24,290 Speaker 1: some manager somewhere was convinced by some AI salesman somewhere 407 00:22:24,450 --> 00:22:26,090 Speaker 1: that it would be that it would be fine. 408 00:22:26,690 --> 00:22:28,050 Speaker 2: Yeah, I mean, I think this is one of the 409 00:22:28,090 --> 00:22:30,130 Speaker 2: things that I realized working on the book is that 410 00:22:30,650 --> 00:22:33,490 Speaker 2: you know, economists, when they're sort of doing their modeling 411 00:22:33,530 --> 00:22:38,770 Speaker 2: about which jobs might be displaced or degraded or changed 412 00:22:38,810 --> 00:22:41,250 Speaker 2: by technology, they look at like how good is the 413 00:22:41,690 --> 00:22:44,490 Speaker 2: technology and compare that to how good is the human? 414 00:22:44,890 --> 00:22:47,130 Speaker 2: But actually, like can a robot do my job as 415 00:22:47,130 --> 00:22:49,650 Speaker 2: well as me? It's not always a particularly useful question, 416 00:22:49,850 --> 00:22:53,250 Speaker 2: Like can someone persuade my boss that a robot can 417 00:22:53,290 --> 00:22:55,250 Speaker 2: do my job as well as me? Is a more 418 00:22:55,290 --> 00:22:59,690 Speaker 2: relevant question? Or would my customers or consumers be content 419 00:22:59,810 --> 00:23:02,730 Speaker 2: to have someone do my job less well than me 420 00:23:03,130 --> 00:23:04,450 Speaker 2: but for a much cheaper price. 421 00:23:04,770 --> 00:23:08,650 Speaker 1: With the case of the translators, is this fundamentally about 422 00:23:08,650 --> 00:23:11,610 Speaker 1: the technology? This just happens to be a thing that 423 00:23:11,690 --> 00:23:13,770 Speaker 1: computers can do. At a certain speed with a certain 424 00:23:13,850 --> 00:23:19,330 Speaker 1: quality that kind of is maximally disruptive to the job 425 00:23:19,370 --> 00:23:22,450 Speaker 1: of being a translator. Or is it actually more to 426 00:23:22,530 --> 00:23:25,210 Speaker 1: do with the fact, for example, that all these people 427 00:23:25,210 --> 00:23:27,650 Speaker 1: are freelancers. Is it actually to do with economic power 428 00:23:27,650 --> 00:23:29,010 Speaker 1: and not to do with technology at all. 429 00:23:29,370 --> 00:23:31,290 Speaker 2: I think it's to do with both. But I think 430 00:23:31,330 --> 00:23:33,970 Speaker 2: you make a good point, which is that often this 431 00:23:34,530 --> 00:23:38,450 Speaker 2: question about what will AI do to the world of 432 00:23:38,490 --> 00:23:40,890 Speaker 2: work as seen as a technical question, but it's not 433 00:23:41,010 --> 00:23:43,570 Speaker 2: just a question of the technical side. It's also a 434 00:23:43,650 --> 00:23:49,850 Speaker 2: question of consumer choice and institutions and culture and bargaining 435 00:23:49,930 --> 00:23:55,050 Speaker 2: power and economic power. So yeah, I mean, partly the 436 00:23:55,130 --> 00:23:58,410 Speaker 2: reason it's been very disruptive to translators is that the 437 00:23:58,530 --> 00:24:02,410 Speaker 2: translating is like them basically all they do, like that 438 00:24:02,490 --> 00:24:05,810 Speaker 2: one task of like translating something. If you're a freelance translator, 439 00:24:05,850 --> 00:24:09,850 Speaker 2: that is pretty much your whole job, whereas for a journalist, 440 00:24:10,130 --> 00:24:12,890 Speaker 2: writing an article is actually quite a small part of 441 00:24:12,890 --> 00:24:16,050 Speaker 2: my job. My job also involves doing a lot of reading, 442 00:24:16,290 --> 00:24:19,490 Speaker 2: going out to interview people, doing stuff like this, and 443 00:24:19,570 --> 00:24:26,090 Speaker 2: so it's more disruptive in some particular professions, depending on 444 00:24:27,210 --> 00:24:31,410 Speaker 2: how much of your tasks are automatical. And then also, yeah, 445 00:24:31,450 --> 00:24:35,210 Speaker 2: like your your market power, right, If you're a freelancer, 446 00:24:35,250 --> 00:24:38,010 Speaker 2: then that's a very different thing to being a well 447 00:24:38,050 --> 00:24:42,010 Speaker 2: protected miner in a unionized mind company in Sweden. 448 00:24:42,450 --> 00:24:46,890 Speaker 1: And can we talk about coders, because superficially programmers software 449 00:24:46,930 --> 00:24:52,930 Speaker 1: coders seems quite like translators. It turns out that for 450 00:24:52,970 --> 00:24:55,130 Speaker 1: some reason, large language models seem to be pretty good 451 00:24:55,130 --> 00:24:57,090 Speaker 1: at coding. At least that's what I'm told. I wouldn't know, 452 00:24:57,170 --> 00:25:01,130 Speaker 1: but people say it's pretty good. So coders in the 453 00:25:01,130 --> 00:25:04,330 Speaker 1: same positions as translators, suddenly I've got all this terrible 454 00:25:04,410 --> 00:25:06,250 Speaker 1: code and kind of they want me to do twice 455 00:25:06,250 --> 00:25:07,970 Speaker 1: as much work for half the salary. Do they feel 456 00:25:07,970 --> 00:25:09,010 Speaker 1: the same way as the translators. 457 00:25:09,330 --> 00:25:12,090 Speaker 2: Most of the ones I interviewed do not feel like that. 458 00:25:13,130 --> 00:25:16,090 Speaker 2: I think coding is an example of a job that 459 00:25:16,130 --> 00:25:18,970 Speaker 2: in many ways actually has probably been enriched and made 460 00:25:19,170 --> 00:25:23,650 Speaker 2: more fun and more productive by this new technology. And 461 00:25:23,690 --> 00:25:26,370 Speaker 2: I think the reason that it is playing out different 462 00:25:26,450 --> 00:25:29,290 Speaker 2: is partly, you know this thing about bargaining power and 463 00:25:29,330 --> 00:25:32,650 Speaker 2: where you sit in the sort of economic chain, but also, 464 00:25:33,090 --> 00:25:35,530 Speaker 2: if you're a computer programmer, like you're what you really 465 00:25:35,570 --> 00:25:38,290 Speaker 2: care about and love is not literally sitting there and 466 00:25:38,330 --> 00:25:41,170 Speaker 2: writing out lines of code. What you love doing is 467 00:25:41,210 --> 00:25:45,810 Speaker 2: like solving problems. And so what the tools do. They 468 00:25:45,890 --> 00:25:48,770 Speaker 2: might automate some of the kind of the actual literal coding, 469 00:25:48,850 --> 00:25:52,810 Speaker 2: but they're not automating that kind of really fun creative 470 00:25:52,890 --> 00:25:53,850 Speaker 2: problem solving. 471 00:25:54,290 --> 00:25:54,530 Speaker 3: Bit. 472 00:25:54,850 --> 00:25:57,610 Speaker 2: It just allows you to like try things out or 473 00:25:58,090 --> 00:25:58,850 Speaker 2: do stuff faster. 474 00:25:59,090 --> 00:26:00,770 Speaker 1: This is the problem. This is what they keep telling us. 475 00:26:00,810 --> 00:26:02,690 Speaker 1: It's going to do for all the jobs. It's going 476 00:26:02,730 --> 00:26:04,490 Speaker 1: to do the boring stuff so you can do the 477 00:26:04,490 --> 00:26:05,090 Speaker 1: fun stuff. 478 00:26:05,090 --> 00:26:07,730 Speaker 2: And this is why they're so enthusiastic and think it's brilliant, 479 00:26:07,730 --> 00:26:09,850 Speaker 2: because for their jobs it is. That's pretty brilliant. 480 00:26:09,930 --> 00:26:15,050 Speaker 1: What is it then, that determines whether your job is 481 00:26:15,170 --> 00:26:19,370 Speaker 1: like translating or whether your job is like coding, And 482 00:26:19,810 --> 00:26:21,210 Speaker 1: is it under our control or not? 483 00:26:21,850 --> 00:26:24,970 Speaker 2: Yeah. So I think one of the things that I've realized, 484 00:26:25,290 --> 00:26:27,450 Speaker 2: and I now really flinch when I hear it, and 485 00:26:27,490 --> 00:26:30,050 Speaker 2: I flinch when I think that I've probably said it 486 00:26:30,090 --> 00:26:32,490 Speaker 2: sometimes as well in the past, is that I think 487 00:26:32,530 --> 00:26:36,170 Speaker 2: we're really using some of the wrong metaphors when we 488 00:26:36,210 --> 00:26:38,530 Speaker 2: talk about this. So I don't know if you've noticed this, 489 00:26:38,650 --> 00:26:41,730 Speaker 2: but I find that like technology executives in particular, but 490 00:26:41,770 --> 00:26:45,490 Speaker 2: also sometimes economists. We'll talk about like there's a tsunami 491 00:26:45,570 --> 00:26:48,770 Speaker 2: of change coming or there is a new wave of 492 00:26:49,210 --> 00:26:52,610 Speaker 2: technological change. I think it's misleading because it makes it 493 00:26:52,650 --> 00:26:55,050 Speaker 2: sound like it's akin to a natural phenomenon, like it's 494 00:26:55,050 --> 00:26:57,770 Speaker 2: something that's just happening, that just came out of nowhere, 495 00:26:57,770 --> 00:26:59,890 Speaker 2: and it's going to happen to us, and it invites 496 00:26:59,970 --> 00:27:02,490 Speaker 2: us to think that at best, all we can do 497 00:27:02,570 --> 00:27:05,130 Speaker 2: is a prepare for it and be get ready to 498 00:27:05,170 --> 00:27:08,130 Speaker 2: mop up after it. And you know, these metaphors I 499 00:27:08,130 --> 00:27:12,250 Speaker 2: think are really useful to technology executives in their conversations 500 00:27:12,250 --> 00:27:15,610 Speaker 2: with policymakers because you know, nobody wants to look like 501 00:27:15,650 --> 00:27:18,050 Speaker 2: the fool who thinks you can hold back the tide, right, 502 00:27:18,090 --> 00:27:21,410 Speaker 2: and so it invites you to think of this as 503 00:27:21,450 --> 00:27:25,010 Speaker 2: something that is happening anyway, and you know that there's 504 00:27:25,090 --> 00:27:27,370 Speaker 2: very little that you can fundamentally do about it. But 505 00:27:27,410 --> 00:27:30,370 Speaker 2: that's not true at all. That's never been true. Technology 506 00:27:30,410 --> 00:27:33,330 Speaker 2: is stuff that is made by people and implemented by people, 507 00:27:33,890 --> 00:27:36,370 Speaker 2: and there is a huge variety, as we've just discussed, 508 00:27:36,410 --> 00:27:38,090 Speaker 2: in the ways in which it can play out in 509 00:27:38,130 --> 00:27:41,490 Speaker 2: different occupations and in different countries and in different parts 510 00:27:41,530 --> 00:27:43,050 Speaker 2: of the labor market, and a lot of that is 511 00:27:43,050 --> 00:27:46,170 Speaker 2: dictated by the choices that people make and the things 512 00:27:46,170 --> 00:27:47,130 Speaker 2: they choose to value. 513 00:27:47,410 --> 00:27:49,730 Speaker 1: So you can do something about this because you can 514 00:27:49,730 --> 00:27:51,970 Speaker 1: write columns and the Financial Time as telling Sam Altman 515 00:27:52,010 --> 00:27:55,370 Speaker 1: that he's being a naughty boy. What can ordinary people 516 00:27:55,410 --> 00:27:59,010 Speaker 1: who do not have newspaper columns and who are not 517 00:27:59,170 --> 00:28:03,370 Speaker 1: running AI companies, just regular people, what can they do 518 00:28:04,170 --> 00:28:10,450 Speaker 1: to help direct this change in a more productive, empowering way. 519 00:28:10,930 --> 00:28:14,690 Speaker 2: I mean, I think people do and are and will 520 00:28:15,370 --> 00:28:19,850 Speaker 2: direct and determine this next progression of technological change, because 521 00:28:19,850 --> 00:28:22,650 Speaker 2: that's what's that's what's always happened. I mean, some people 522 00:28:22,730 --> 00:28:25,290 Speaker 2: obviously have more power than others, particularly in terms of 523 00:28:25,330 --> 00:28:27,690 Speaker 2: like what kind of working conditions they may or may 524 00:28:27,730 --> 00:28:31,010 Speaker 2: not have to accept. But you know, everyone that I 525 00:28:31,090 --> 00:28:34,890 Speaker 2: interviewed in my book is responding to what's happening. And 526 00:28:35,250 --> 00:28:38,530 Speaker 2: you know, whether that is that they are going on 527 00:28:38,610 --> 00:28:41,370 Speaker 2: strike or forming a union, or whether it's that they 528 00:28:41,570 --> 00:28:45,530 Speaker 2: are changing their profession or figuring out some new way 529 00:28:45,770 --> 00:28:49,050 Speaker 2: of doing what they do in a different way. Also 530 00:28:49,090 --> 00:28:51,010 Speaker 2: as consumers, you know a lot of the questions we've 531 00:28:51,050 --> 00:28:53,250 Speaker 2: been talking about is like, well, what will the market accept? 532 00:28:53,530 --> 00:28:55,690 Speaker 2: We are the market, you know, so it's kind of 533 00:28:55,810 --> 00:28:59,210 Speaker 2: up to us to decide whether we like AI music 534 00:28:59,370 --> 00:29:00,210 Speaker 2: or real music. 535 00:29:00,330 --> 00:29:01,810 Speaker 1: You know, so flat that the one's ever said on 536 00:29:01,890 --> 00:29:04,250 Speaker 1: the market before you are the market. The nicest thing 537 00:29:04,290 --> 00:29:05,010 Speaker 1: anyone's ever said. 538 00:29:05,890 --> 00:29:07,130 Speaker 2: How to complement an economist? 539 00:29:07,210 --> 00:29:11,090 Speaker 1: Absolutely so you heard it. Fight back, stand up show, show, 540 00:29:11,530 --> 00:29:14,410 Speaker 1: show the man, why you're better than the machine. Sarah 541 00:29:14,450 --> 00:29:19,330 Speaker 1: and I are going to talk about large language models 542 00:29:20,170 --> 00:29:34,850 Speaker 1: after the break. We're back. I'm Tim Harford. We are 543 00:29:34,970 --> 00:29:38,490 Speaker 1: live at the Bristol Festival of Economics, and this is 544 00:29:38,530 --> 00:29:46,650 Speaker 1: a cautionary conversation with expert in artificial intelligence, expert in workplaces, 545 00:29:46,930 --> 00:29:51,370 Speaker 1: author of We Are Not Machines, Sarah O'Connor, who is 546 00:29:51,370 --> 00:29:55,690 Speaker 1: also my colleague at the Financial Times. The title of 547 00:29:55,690 --> 00:29:58,610 Speaker 1: the book is We Are Not Machines. You've given us 548 00:29:58,610 --> 00:30:05,370 Speaker 1: some really striking examples of demands of made of workers 549 00:30:05,730 --> 00:30:09,410 Speaker 1: to be more machine like, being under constant surveillance, to 550 00:30:09,450 --> 00:30:13,210 Speaker 1: move in a very kind of weird and unnatural way. 551 00:30:14,970 --> 00:30:17,290 Speaker 1: I wanted to bring it back to the experience a 552 00:30:17,330 --> 00:30:21,250 Speaker 1: lot of white collar workers have of sort of sitting 553 00:30:21,290 --> 00:30:23,570 Speaker 1: in a computer every day and suddenly discovering there's this 554 00:30:23,570 --> 00:30:28,050 Speaker 1: thing called chat gpt. Do you think that chat gpt 555 00:30:28,370 --> 00:30:31,490 Speaker 1: is this making us more machine like and we don't 556 00:30:31,810 --> 00:30:32,170 Speaker 1: know it. 557 00:30:32,610 --> 00:30:34,610 Speaker 2: So this is something that a lot of the translators 558 00:30:34,650 --> 00:30:38,170 Speaker 2: that I interviewed brought up. I think because they they're linguist, 559 00:30:38,170 --> 00:30:41,570 Speaker 2: so they have like an ear for language and how 560 00:30:41,610 --> 00:30:43,730 Speaker 2: it's changing. But a number of them said to me 561 00:30:43,770 --> 00:30:47,210 Speaker 2: independently that they had noticed that the sorts of language 562 00:30:47,250 --> 00:30:52,090 Speaker 2: that we are using is becoming narrower and flatter and 563 00:30:52,250 --> 00:30:55,370 Speaker 2: more homogeneous. And I wonder if that is partly because 564 00:30:56,010 --> 00:30:59,170 Speaker 2: we are using chat GPT more to write our LinkedIn 565 00:30:59,250 --> 00:31:02,690 Speaker 2: posts and our emails, and maybe that even seeps into 566 00:31:02,730 --> 00:31:05,890 Speaker 2: the way that we communicate with one another. So I 567 00:31:05,970 --> 00:31:08,650 Speaker 2: think there are certain risks there. But also, you know, 568 00:31:08,730 --> 00:31:12,770 Speaker 2: I think large angers models do clearly have the capacity 569 00:31:12,810 --> 00:31:14,650 Speaker 2: to do that thing that we talked about at the start, 570 00:31:14,690 --> 00:31:16,850 Speaker 2: which is like take away some of the boring stuff. 571 00:31:16,850 --> 00:31:18,810 Speaker 2: And I think there are plenty of examples of people 572 00:31:19,290 --> 00:31:22,010 Speaker 2: doing that, whether it's like, oh, now I don't have 573 00:31:22,090 --> 00:31:24,650 Speaker 2: to take any minutes for this meeting, or now I 574 00:31:24,650 --> 00:31:28,450 Speaker 2: don't have to spend two hours googling three relevant research 575 00:31:28,530 --> 00:31:31,850 Speaker 2: articles but I do think it's also it's very easy 576 00:31:31,930 --> 00:31:34,890 Speaker 2: to start also using it for things that maybe in 577 00:31:34,930 --> 00:31:36,530 Speaker 2: the past we would have put a bit of human 578 00:31:36,530 --> 00:31:38,930 Speaker 2: care and attention into. Like, you know, I think a 579 00:31:38,930 --> 00:31:40,890 Speaker 2: lot of people are using chat gupt now to write 580 00:31:40,930 --> 00:31:44,050 Speaker 2: wedding speeches or speeches at funerals and that sort of 581 00:31:44,050 --> 00:31:47,450 Speaker 2: thing which you know, you might once have thought were 582 00:31:47,490 --> 00:31:49,530 Speaker 2: among the more human tasks that we did. 583 00:31:49,930 --> 00:31:53,290 Speaker 1: Yeah, although I have heard a few wedding speeches, chatchupt 584 00:31:53,330 --> 00:31:57,330 Speaker 1: would have done better than some of them. So you know, yeah, 585 00:31:58,690 --> 00:32:02,970 Speaker 1: rising tide lifts sundboats. Are you are you frightened or 586 00:32:02,970 --> 00:32:04,130 Speaker 1: are you hopeful for the future. 587 00:32:04,450 --> 00:32:06,970 Speaker 2: I'm a lot more hopeful now than I was when 588 00:32:06,970 --> 00:32:09,970 Speaker 2: I started writing the book, actually, and I think that's 589 00:32:10,210 --> 00:32:13,930 Speaker 2: probably because of the people that I met along the way. 590 00:32:15,130 --> 00:32:18,810 Speaker 2: So like another of these sort of metaphors or like 591 00:32:18,810 --> 00:32:21,770 Speaker 2: cliches that we often hear is you know, people will say, well, 592 00:32:21,810 --> 00:32:24,330 Speaker 2: you know, inevitably there's going to be winners and losers. 593 00:32:24,850 --> 00:32:28,250 Speaker 2: You know, there's there's dangers and opportunities, and that sort 594 00:32:28,250 --> 00:32:30,170 Speaker 2: of invites us to think, like, oh good, I'd better 595 00:32:30,290 --> 00:32:31,730 Speaker 2: just wait and find out if I'm going to be 596 00:32:31,770 --> 00:32:34,170 Speaker 2: a winner. Or a loser, and like you know, fingers crossed, 597 00:32:34,530 --> 00:32:36,170 Speaker 2: I'll be a winner. You just kind of got your 598 00:32:36,210 --> 00:32:38,490 Speaker 2: lottery ticket and then you wait and see. But like 599 00:32:38,570 --> 00:32:41,330 Speaker 2: that's not actually what people are doing. Like people actually 600 00:32:41,450 --> 00:32:45,250 Speaker 2: are maneuvering themselves and trying to figure out what they 601 00:32:45,250 --> 00:32:47,810 Speaker 2: can do to make sure that things work out well 602 00:32:47,850 --> 00:32:51,650 Speaker 2: for them, Like everyone is shaping and determining this. So 603 00:32:51,690 --> 00:32:54,730 Speaker 2: in that sense, I ended up more optimistic because I 604 00:32:54,810 --> 00:32:57,810 Speaker 2: think that we have a lot more agency to figure 605 00:32:57,850 --> 00:32:59,970 Speaker 2: out how this goes and to make sure it goes 606 00:33:00,010 --> 00:33:02,010 Speaker 2: in a way that we want. Then we're sometimes led 607 00:33:02,010 --> 00:33:02,410 Speaker 2: to believe. 608 00:33:03,450 --> 00:33:06,890 Speaker 1: Thank you so much, Sarah. Let's take some questions. So 609 00:33:06,930 --> 00:33:11,930 Speaker 1: we've got a question from a Cautionary Club member, Emily. 610 00:33:12,090 --> 00:33:14,410 Speaker 1: By the way, anybody wants to join the Cautionary Club, 611 00:33:15,290 --> 00:33:17,570 Speaker 1: it's very wonderful. You'd get the podcast ad free, you 612 00:33:17,570 --> 00:33:20,690 Speaker 1: get bonus episodes all very good, and you get your 613 00:33:20,770 --> 00:33:23,730 Speaker 1: question read out live at the Bristol Festival of Economics. 614 00:33:24,170 --> 00:33:27,930 Speaker 1: So Emily says, there's a research instituite called me et 615 00:33:28,290 --> 00:33:31,970 Speaker 1: R which found that both the expected and the perceived 616 00:33:32,010 --> 00:33:36,490 Speaker 1: efficiency improvements from using AI in software R and D 617 00:33:37,210 --> 00:33:42,530 Speaker 1: were substantially overstated. She asks, do you think businesses are 618 00:33:42,610 --> 00:33:46,690 Speaker 1: likely to scale back AI investment until the efficiency gains 619 00:33:46,730 --> 00:33:51,010 Speaker 1: promised are demonstrated, or have they sunk so much money 620 00:33:51,010 --> 00:33:53,970 Speaker 1: in at this point that they feel they have to 621 00:33:54,010 --> 00:33:54,930 Speaker 1: justify the expense. 622 00:33:55,450 --> 00:33:57,690 Speaker 2: I think it remains to be seen. I think that 623 00:33:57,810 --> 00:34:01,530 Speaker 2: matr study was really interesting because what it sort of 624 00:34:01,570 --> 00:34:04,210 Speaker 2: highlights to me is that, you know, we were talking 625 00:34:04,210 --> 00:34:08,210 Speaker 2: about this distinction between tasks and jobs, and that actually 626 00:34:08,410 --> 00:34:10,690 Speaker 2: your whole job might disappear, but some tasks within it 627 00:34:10,770 --> 00:34:12,970 Speaker 2: might change. But when you think about something at the 628 00:34:13,090 --> 00:34:16,370 Speaker 2: level of an entire organization, you've also got to think 629 00:34:16,370 --> 00:34:21,090 Speaker 2: about jobs and systems, and so you know, a computer 630 00:34:21,170 --> 00:34:24,730 Speaker 2: programmer on his or her own might have become much 631 00:34:24,730 --> 00:34:28,770 Speaker 2: more productive if they can use a coding AI A system, 632 00:34:28,770 --> 00:34:30,810 Speaker 2: but actually, if there are still bottlenecks in other parts 633 00:34:30,850 --> 00:34:33,530 Speaker 2: of that system, you might not really see the benefit 634 00:34:33,650 --> 00:34:36,250 Speaker 2: of it. You know, you might just create lots and 635 00:34:36,250 --> 00:34:38,610 Speaker 2: lots and lots of code, but then you still you 636 00:34:38,690 --> 00:34:41,530 Speaker 2: still don't have enough people to review it, for example, 637 00:34:41,730 --> 00:34:43,690 Speaker 2: or you know, whatever it might be. And so I 638 00:34:43,730 --> 00:34:46,250 Speaker 2: think what a lot of companies are finding is that 639 00:34:46,730 --> 00:34:51,090 Speaker 2: suddenly being able to do certain tasks within jobs much 640 00:34:51,090 --> 00:34:53,690 Speaker 2: more and much faster, doesn't necessarily mean that you're going 641 00:34:53,730 --> 00:34:56,890 Speaker 2: to see the same kind of productivity outcome filter through 642 00:34:57,330 --> 00:35:00,010 Speaker 2: to the whole organization. The other thing I would say 643 00:35:00,050 --> 00:35:02,210 Speaker 2: is that we're still like in the very early stages 644 00:35:02,250 --> 00:35:05,970 Speaker 2: of companies trying to figure out how to use these 645 00:35:06,010 --> 00:35:09,690 Speaker 2: new tools effectively. And what I think we might start 646 00:35:09,730 --> 00:35:15,570 Speaker 2: to see is companies kind of redesigning workflows around what 647 00:35:15,650 --> 00:35:17,810 Speaker 2: the machines can and can't do. I mean, in a way, 648 00:35:17,850 --> 00:35:20,450 Speaker 2: that's exactly what Amazon has done, right. I mean, it's 649 00:35:20,450 --> 00:35:23,730 Speaker 2: completely changed its workflow in order to make the most 650 00:35:23,730 --> 00:35:25,810 Speaker 2: of the machines that it has available to it at 651 00:35:25,850 --> 00:35:29,330 Speaker 2: the moment, and plugged humans in to do the bits 652 00:35:29,330 --> 00:35:31,450 Speaker 2: that the machines can't do. But I think it's still 653 00:35:31,490 --> 00:35:34,690 Speaker 2: a really open question, you know, particularly with large language models. 654 00:35:34,730 --> 00:35:37,530 Speaker 2: I mean, the kind of the hallucination problem I think 655 00:35:37,610 --> 00:35:40,490 Speaker 2: is quite could be quite existential for a lot of 656 00:35:40,850 --> 00:35:44,530 Speaker 2: sort of high risk professions. You know, certainly, you know, 657 00:35:44,570 --> 00:35:46,490 Speaker 2: we're not going to be using large language models that 658 00:35:46,690 --> 00:35:49,010 Speaker 2: the ft to write our articles because we just can't 659 00:35:49,250 --> 00:35:50,890 Speaker 2: possibly trust them not to hallucinate. 660 00:35:51,130 --> 00:35:52,810 Speaker 1: Yeah, I mean, I always find when I'm using large 661 00:35:52,850 --> 00:35:55,850 Speaker 1: language models The first question is will anybody else ever 662 00:35:55,930 --> 00:35:58,530 Speaker 1: see any of this? And if the answer is yes, 663 00:35:58,570 --> 00:36:01,850 Speaker 1: then it's unusable. Like if it's just for me, then 664 00:36:02,170 --> 00:36:05,010 Speaker 1: sometimes it comes up with useful ideas, but I absolutely 665 00:36:05,010 --> 00:36:07,970 Speaker 1: cannot trust it at all to produce anything that anybody 666 00:36:08,010 --> 00:36:13,410 Speaker 1: else will ever see. So anyway, more questions. There were 667 00:36:13,410 --> 00:36:16,450 Speaker 1: loads of questions, can we get the microphone? Somebody looking 668 00:36:16,690 --> 00:36:18,970 Speaker 1: quite young and fresh face there. I want a question 669 00:36:19,010 --> 00:36:20,850 Speaker 1: from Hi. 670 00:36:22,010 --> 00:36:25,410 Speaker 4: I have a question, which industry do you think is 671 00:36:25,730 --> 00:36:32,610 Speaker 4: most at risk of AI replacing the job or having 672 00:36:32,770 --> 00:36:37,610 Speaker 4: the biggest amount of the job being taken away? 673 00:36:38,690 --> 00:36:42,210 Speaker 1: So, Sarah, which industry is most at risk of just 674 00:36:42,250 --> 00:36:44,610 Speaker 1: having these jobs completely removed? 675 00:36:44,890 --> 00:36:46,530 Speaker 2: I mean, I hate to say it, but. 676 00:36:46,890 --> 00:36:47,930 Speaker 1: It's the translators. 677 00:36:48,210 --> 00:36:50,050 Speaker 2: I know I was going to say it's I think 678 00:36:50,170 --> 00:36:53,050 Speaker 2: sort of creative industries more generally. I mean not that 679 00:36:53,130 --> 00:36:56,770 Speaker 2: I think that, you know, people will stop wanting to 680 00:36:56,810 --> 00:37:02,210 Speaker 2: pay for human created stuff, but I think that there 681 00:37:02,290 --> 00:37:08,410 Speaker 2: are lots of people who work in you know, they're 682 00:37:08,450 --> 00:37:12,210 Speaker 2: not necessarily originating the ideas, but they work in the 683 00:37:12,290 --> 00:37:15,330 Speaker 2: kind of in that long process between someone coming up 684 00:37:15,330 --> 00:37:18,490 Speaker 2: with an idea and the thing being made. Whether that's 685 00:37:18,610 --> 00:37:24,530 Speaker 2: like working in special effects in Hollywood, or music production, 686 00:37:25,090 --> 00:37:29,290 Speaker 2: or storyboarding in an advertising agency or whatever it might 687 00:37:29,330 --> 00:37:32,530 Speaker 2: be your copywriting. I think there are lots and lots 688 00:37:32,530 --> 00:37:36,370 Speaker 2: of jobs there which you know, are good jobs and 689 00:37:36,410 --> 00:37:38,570 Speaker 2: people like them, and so I think it's a real 690 00:37:38,610 --> 00:37:42,450 Speaker 2: shame if they become disrupted. But I can imagine that 691 00:37:42,450 --> 00:37:45,410 Speaker 2: those are the sorts of roles that could be quite 692 00:37:45,890 --> 00:37:48,010 Speaker 2: quite vulnerable. But I might be wrong, and I hope 693 00:37:48,050 --> 00:37:48,650 Speaker 2: I am. 694 00:37:49,050 --> 00:37:51,410 Speaker 1: So you heard it here first, human creativity is dead. 695 00:37:52,890 --> 00:37:54,370 Speaker 2: It's definitely what I said. Thanks to him. 696 00:37:54,450 --> 00:37:58,090 Speaker 1: This from the woman who calls herself an optimist, So 697 00:37:58,170 --> 00:38:00,410 Speaker 1: thank you for that. We've got time for one more question. 698 00:38:01,410 --> 00:38:05,490 Speaker 3: You said that AI is no advanced enough to translate 699 00:38:05,570 --> 00:38:09,250 Speaker 3: the tags accurately enough, but it's really been around or 700 00:38:09,850 --> 00:38:12,970 Speaker 3: two years, like even by twenty SAIDs is going to 701 00:38:12,970 --> 00:38:16,610 Speaker 3: be enough advance to replace some of the jobs. Surely 702 00:38:16,730 --> 00:38:20,290 Speaker 3: it will lead to like structural unemployment and the Regon 703 00:38:20,650 --> 00:38:22,090 Speaker 3: government will there happen. 704 00:38:22,370 --> 00:38:25,130 Speaker 1: Thank you very much. So basically I think two ideas 705 00:38:25,130 --> 00:38:27,490 Speaker 1: nested in that question. So one is this point that 706 00:38:28,170 --> 00:38:30,490 Speaker 1: we can point to what the AIS can't do yet, 707 00:38:30,730 --> 00:38:35,090 Speaker 1: but they change fast, So maybe a lot of the 708 00:38:35,130 --> 00:38:37,650 Speaker 1: things that we're saying AI can't the AI vision doesn't 709 00:38:37,690 --> 00:38:40,050 Speaker 1: work that well, AI translation doesn't work that well. Well, 710 00:38:40,370 --> 00:38:43,450 Speaker 1: maybe the answer is just wait. And then the second 711 00:38:43,490 --> 00:38:50,450 Speaker 1: question is, if this really is incredibly disruptive to labor markets, 712 00:38:50,610 --> 00:38:53,930 Speaker 1: are government's actually going to just do something drastic to 713 00:38:53,970 --> 00:38:54,650 Speaker 1: outlaw it? 714 00:38:55,770 --> 00:38:58,650 Speaker 2: So the first part of your question, I mean, we've 715 00:38:58,690 --> 00:39:01,530 Speaker 2: already seen how much better these things have gotten in 716 00:39:01,570 --> 00:39:04,130 Speaker 2: the last few years. I mean, when things like mid 717 00:39:04,170 --> 00:39:07,650 Speaker 2: Journey first came out, the image generation, like every person's 718 00:39:07,690 --> 00:39:10,730 Speaker 2: hand had like six or somenings right, and we thought 719 00:39:10,770 --> 00:39:14,450 Speaker 2: it was hilarious. And now they're not making those mistakes anymore. 720 00:39:14,490 --> 00:39:17,530 Speaker 2: So yes, definitely things are getting better. And in a way, 721 00:39:17,570 --> 00:39:19,930 Speaker 2: that's what's a bit difficult about writing about this topic 722 00:39:20,050 --> 00:39:22,130 Speaker 2: is it's such a kind of moving target. So you're like, 723 00:39:22,170 --> 00:39:25,410 Speaker 2: you're trying to write about something that just keeps keeps changing. 724 00:39:25,890 --> 00:39:28,170 Speaker 2: But as I was saying, like, I don't actually think 725 00:39:28,250 --> 00:39:31,810 Speaker 2: the capability of what they can do is the only 726 00:39:32,010 --> 00:39:35,090 Speaker 2: thing that matters, because it's also about like what kind 727 00:39:35,130 --> 00:39:37,090 Speaker 2: of trade offs are people willing to accept? And then 728 00:39:37,130 --> 00:39:40,610 Speaker 2: in terms of if we started to see really big 729 00:39:41,490 --> 00:39:46,170 Speaker 2: labor market dislocations, would governments step in and do something 730 00:39:46,210 --> 00:39:49,730 Speaker 2: drastic maybe. I mean, I don't think there's any precedent 731 00:39:50,450 --> 00:39:53,930 Speaker 2: for that that I can think of. I mean, in 732 00:39:53,970 --> 00:39:56,090 Speaker 2: the example of the ludd Ites, the government stepped in 733 00:39:56,490 --> 00:39:59,610 Speaker 2: on behalf of the owners of the machines and sent 734 00:39:59,650 --> 00:40:01,890 Speaker 2: the luodyes to Australia. So that was like the opposite. 735 00:40:03,330 --> 00:40:08,770 Speaker 1: Well, the President surely is is tariffs because I mean, effectively, 736 00:40:09,930 --> 00:40:12,090 Speaker 1: this is the point that economists like to make. But 737 00:40:12,210 --> 00:40:16,050 Speaker 1: trade with China, for example, it's effectively a technology. I mean, yeah, 738 00:40:16,210 --> 00:40:18,850 Speaker 1: it's China over there, but it could be just some 739 00:40:19,570 --> 00:40:22,530 Speaker 1: robots in a factory in Los Angeles next to the docks. 740 00:40:22,970 --> 00:40:27,290 Speaker 1: And you know, whether you're shipping grain into the robot 741 00:40:27,290 --> 00:40:30,130 Speaker 1: factory and they're turning it into television sets, or whether 742 00:40:30,130 --> 00:40:32,650 Speaker 1: you're shipping the grain across the Pacific and the Chinese 743 00:40:33,170 --> 00:40:35,090 Speaker 1: by the grain and then they say and they send 744 00:40:35,130 --> 00:40:38,610 Speaker 1: you television sets. I mean, it kind of doesn't matter. 745 00:40:38,770 --> 00:40:43,330 Speaker 1: So you could imagine taxing robot produced you know, robot 746 00:40:43,330 --> 00:40:45,890 Speaker 1: produce goods. I just don't know if people hate robots 747 00:40:45,930 --> 00:40:47,290 Speaker 1: as much as they hate foreigners. 748 00:40:49,450 --> 00:40:50,250 Speaker 2: Let's find out. 749 00:40:50,650 --> 00:40:53,010 Speaker 1: That is sadly all we have time for. Thank you 750 00:40:53,090 --> 00:40:55,130 Speaker 1: so much for coming. It's been an absolute pleasure to 751 00:40:55,130 --> 00:40:57,250 Speaker 1: be here in Bristol. Thank you so much to my 752 00:40:57,330 --> 00:41:01,290 Speaker 1: wonderful guest Sarah O'Connor. Her book We Are Not Machines 753 00:41:01,490 --> 00:41:03,930 Speaker 1: The Fight for the Future of Work is out now 754 00:41:04,410 --> 00:41:08,130 Speaker 1: do please pick up a copy, But before then, please 755 00:41:08,210 --> 00:41:16,170 Speaker 1: join me in thanking my guest, Sarah O'Connor. Horse Retales 756 00:41:16,210 --> 00:41:19,170 Speaker 1: is written by me Tim Harford, with Andrew Wright, Ryan 757 00:41:19,250 --> 00:41:23,170 Speaker 1: Dilly and Alice Fines. This live episode was produced by 758 00:41:23,210 --> 00:41:27,370 Speaker 1: Georgia Mills with help from Marilyn Russ. The sound design 759 00:41:27,410 --> 00:41:30,810 Speaker 1: and original music was the work of Pascal Wise. The 760 00:41:31,050 --> 00:41:35,690 Speaker 1: sound engineer for tonight's show was Tom Dunn with eduardo' galley. 761 00:41:36,170 --> 00:41:39,290 Speaker 1: Thank you to xaviie Levantis and the Bristol Festival of 762 00:41:39,290 --> 00:41:40,010 Speaker 1: Economics