307. The world’s most influential economist: how we can deliver "worker-friendly AI"

0:00Robert Colvile:For adults with Crohn's disease or ulcerative colitis symptoms, every choice matters. Tremphaya offers self-injection or intravenous infusion from the start. Tremphaya is administered as injections under the skin or infusions through a vein every four weeks, followed by injections under the skin every four or eight weeks. If your doctor decides that you can self-inject Tremphaya, proper training is required. Tremphaya is a prescription medicine used to treat adults with moderately to severely active Crohn's disease and adults with moderately to severely active ulcerative colitis. Serious allergic reactions, increased risk of infections or lower ability to fight them, and liver problems may occur.

0:40Robert Colvile:Before treatment, get checked for infections and tuberculosis. Tell your doctor if you have an infection, flu-like symptoms, or need a vaccine. Explore what's possible. Ask your doctor about Tremphaya today. Call 1-800-526-7736 to learn more or visit Tremfierradio.com. Queen Carvania stood haloed by the morning sun. An army hung on her every word. My champions, I have sold my chariot on Carvana. It was a lovely SUV, an inexplicably queenly offer. They're even coming to the castle to collect it. Tonight, we feast! An offer you can feast on. Sell your car today on Carvana. Pickup fees may apply.

1:31Robert Colvile:This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required compatibility and availability varies 18 plus. Nobel Prize winner Darren Asimoglu says those who've run our countries in recent decades were wrong to believe that the digital and automation revolution that crushed so many jobs and suppressed wages was inevitable.

2:10Robert Colvile:He says all technology can be shaped to make it more friendly to workers. And his current obsession is to do this with artificial intelligence to prevent AI from eliminating millions and millions of jobs and make us all dependent on handouts from the state. So what is his worker-friendly AI and is it deliverable? In part two of my interview with the acclaimed economist, I find out. Here it is. Support for this episode comes from Octopus Energy and the founder and CEO Greg Jackson is with us now. So, Greg, talk to me about Kraken. What is it and why is it spinning off from Octopus Energy? When we built Octopus, we also built a software platform like an operating system like iOS or Android on your phone to enable energy companies, utilities.

3:04Robert Colvile:starting with Octopus but frankly anyone in the world to be much more efficient to use vast amounts of data to be able to become more innovative and to serve their customers better that piece of software is cracking we've used it not only for Octopus but companies in the UK like EDF and Eon have used it to improve their business too but you know what it turns out over time companies realized that they were licensing from a competitor and so we've had to spin it off so it can reach its full potential. Yeah, makes sense, Craig. Thank you. Right, we're going to go to the episode.

3:49Robert Colvile:Darren, in our earlier episode, you were saying that we do have the ability to shape how new technologies are deployed, which is counter to what, as many of our leaders have believed over the last 20, 30 years, and you are absolutely focused on how artificial intelligence can be shaped so that it does not do what many fear it will do, which is to eliminate vast numbers of jobs and whether, you know, as a period of disruption or permanently leads to massive unemployment. So talk me through what you think is necessary to avoid such a potential jobs cataclysm. So I think the first thing is we have to change our mindset.

4:39There are two synergistic ideas that have taken over the public discourse about AI and technology. One is that technology, whatever it means, comes in some inevitable form. That it has its own dynamics and it has its own direction that's already been shaped. And all that powerful, clever entrepreneurs do is pursue that particular predetermined path. The second is that as long as technology advances in some way, and again, what does that mean? There are so many different ways in which technology can advance. But as long as it advances, that's going to be for the good of society. Both of these propositions are wrong.

5:29Technology never has a predetermined direction, especially its inequality consequences are very much dependent on how we use it. And as the examples of technologies such as nuclear power applied to build bombs or perhaps AI that becomes out of control would illustrate, not all advances are good for society. So we have to have a clear-eyed understanding of what's on the menu of technologies and which parts of that menu are for the social good. The discussion in the United States especially, then exported to Europe and the UK, is polluted by this belief that the menu is exactly what we're doing, is exactly what OpenAI, Anthropic, and Google are doing, and nothing else is possible.

6:25And even if asking that question is sacrilege. No, AI is such a versatile technology. It is the ultimate information technology, which means technology that provides information to humans. Many pioneers of computer science, starting in the 1950s, envisaged a day, they thought it would come sooner, but envisaged a day where digital technologies would be a tool, powerful tool in the hands of humans to expand human capabilities. Going back to our discussion from the last episode, to enable humans to contribute more to the production process, to productivity, to society, not sideline them. But the current obsession, especially around the banner of AGI, artificial general intelligence, or sometimes artificial super intelligence, is that these models are soon going to become so powerful that they can do everything that humans can do.

7:28And hence, humans are going to be dispensable. They'll eat the fruits of this, perhaps, you know, who knows? I'm very doubtful of that. But they won't be so important because the models are going to be so much better than humans. We can debate whether that's a feasible path or not. I actually think it's not anything that will happen anytime soon. But even if it were feasible, it's disastrous. We cannot have a democratic society. We cannot have a functional society. We cannot have social peace when a very large fraction of humanity feels they have nothing to contribute to society. They are dispensable.

8:09All they can hope for is live off the crumbs of some billionaires who control AI models and go home and play video games or just do leisure all day. I think that would go against our drive for having meaningful social relations and having this sense of contributing to society. Of course, it's possible, perhaps in 50 years' time, we will have different social values and different things such that people could have a feeling of contributing to society that doesn't work through some sort of building skills and giving advice or giving information or doing things and solving problems. But today, that's how we get our sense of dignity.

8:49We solve problems. We do things from the home, in parenting, from the community when we give advice to people, when we share information, to the production process where we make things or we solve other problems. I think that is a very critical part of it. If AI takes these things over, I think liberal democracy will collapse as well. The good news, AI doesn't need to do that. As an information technology, it can make workers more productive. And that's the idea of my pro-worker AI agenda.

9:18Robert Colvile:And I want to drill into that in a minute or two. But just to paraphrase, you genuinely believe that it is more practical to essentially stop these juggernauts, you know, essentially investing hundreds. I mean, I think there's going to be more than a trillion dollars this year invested by just three companies in the development of generative AI with a view to, they hope, reaching artificial and general intelligence. you think it is more practical to stop that juggernaut than to prepare for a world where we would need essentially a super tax on these AI companies in order to deliver universal basic income to the millions and millions of people who won't have a job?

10:15Well, no, I wouldn't say we should be thinking of a tax on the companies and the universal basic income as the main levers. Let's clarify what I am proposing a little bit, and then we'll talk about tools for achieving it. I am not arguing we should stop AI. It may come a point where that may be necessary. But the image I have in mind is that we are in a fast car driving uncontrollably towards a cliff. You may want to hit the brake, but the point is so that you can steer away towards somewhere where there is a better pathway. So I'll be in favor of slowing down for a short while so that we can redirect.

11:02It's the redirection that I'm asking for. And that redirection is more pro-worker AI and it's also more pro-human AI in other domains than the production process. We'll come back to that. But let me explain why two easy solutions are not on the table. Yeah. One is we cannot just dictate that new direction. No government bureaucrat, no government regulator is capable of steering technology that precisely. What we can do is use the democratic process to provide a framework, but also change norms and expectations and aspirations among the very talented people working in the tech sector so that they work in a direction congruent with human betterment, liberal democracy, shared prosperity.

11:56And no, I don't think UBI would work for the same reasons that I have articulated. UBI would, if funded at a sufficiently generous level, would indeed support living conditions for people that is much more decent than today. And if nothing else was on the table, I would, of course, support that. But the problem with UBI is that it doesn't solve the dignity contribution problem that I outlined. If we have a society in which, say, 80%, 90 % of the population is on UBI, they have officially been designated as not contributing to society, while the remaining 10%, 20 % generate all the wealth or control all the wealth.

12:49That is a very, very unequal society from the view of social status, a two-tier society, and it would come with many, many deep challenges and probably will be as difficult for liberal democracy to take root in as a very highly unequal society without UBI. There is a sense in which UBI is also throwing the towel. It says, no, we cannot do anything to control technology. No technology is indeed inevitable, as the Silicon Valley moguls are telling us. All we can do is try to sort of lessen its effects by redistributing. I am all for a better social safety net. More redistribution is necessary for a number of reasons, but not as a substitute for creating jobs, creating meaningful professions, and pro-worker AI.

13:42Robert Colvile:So let's talk a bit about pro-worker AI. And we, you know, as certainly an alternative path to the sort of slightly fatalistic one of, you know, everybody's going to lose their jobs and we've got to have universal basic income. I mean, in my own case, just as a sort of one, I think, semi-optimistic case study, you know, I find the, you know, the AI that I use massively increases my productivity. You think that we just got to think more creatively across all occupations to come up with examples where AI can do that? Absolutely, 100%. First of all, I absolutely reject the view that AI models are stochastic parades of no good.

14:24I mean, as you said, its capabilities are already pretty striking. But the real key is in journalism, in nursing, in electrical work, in plumbing, in blue-color work, for AI to create new capabilities and new expertise for workers. Today, it may appear like if you are at the forefront and AI does some of your work for you, it is contributing to your productivity. Perhaps you're going to be advantaged. That's a temporary thing. And if I want to illustrate that best in the UK context, I would think back to the days of black cabbies and minicabs in the UK, 35 years ago or so. Black cabs knew all the streets of London, and they had unique, very valuable skill.

15:19Minicabs, cheaper though they were, couldn't compete with that. Now, imagine one of those minicab drivers being given a GPS device, which didn't exist 35 years ago, but did exist, I guess, 29, 28 years ago. That worker would benefit from this new technology. In fact, black cabbies would too. I mean, they have amazing knowledge. They had amazing knowledge of the streets, but not all of them. There were some gaps. If you give them a GPS device, they would have benefited too. But once those GPS devices spread everywhere, it commodified that knowledge. And no longer could these drivers receive the same premium for what they did.

16:10Now, you can say consumers benefited. In every case, that's going to be different. But that commodification is part of the agenda of AGI. So the advantages that we generate from AI today, if they are going to be durable, they really need to be in the form of truly complementing humans. And that means expanding what we do. For an electrician, not just enabling that electrician to do what they were doing faster, but enabling them to do new things, learn more quickly when there are new equipments, deal with much more complex troubleshooting issues, deal with new problems in a much more expert way.

16:55For a journalist, it would expand what they can do. So those are the kinds of things that AI and generative AI at that is actually very good at. Because the knowledge of humanity to a first approximation is out there. It's on the internet. It's in books. It's in encyclopedias. But good luck finding it when you need it. What generative AI is capable of is to find the relevant, reliable information that is context dependent for you. That's where the pro worker AI lies.

17:29Robert Colvile:Darren, we need to go to a quick break. Fingers crossed. When we're back, we won't have been replaced by an artificial intelligence super brain. See you in a minute.

17:41Robert Colvile:This episode is brought to you by Accenture. Now, when your advertising operations fall out of sync, campaigns slow down, insights get buried and opportunities get missed. That's why Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics and smarter workflows to simplify campaign delivery and access better data across the business. The result, less time spent on operations, more time connecting brands with the moments and fandoms that matter most. To learn more, check out Accenture.com slash Spotify dash UK.

18:21Robert Colvile:Welcome back to The Rest Is Money. In a minute or two, we'll be back with our chat. But I wanted to remind you that the podcast is doing its first ever live show. It's at the South Bank Centre, 7.30 on Friday, the 4th of September. Part of The Rest Is Fest. Now, there are just a few tickets left. And one of the treats that Steph and I have for you is for one night only, you can be Chancellor. Tickets can be bought from the Southbank Centre website and the link is in the episode description. I was very struck, however, by one of the things you say in your great book, which is that you don't believe that if we became very dependent on, or totally dependent on large language model, generative AI, conceivably in the form of artificial general intelligence or super intelligence, you actually don't believe that we would have the intellectual and scientific breakthroughs that its proponents claim.

19:26Robert Colvile:So if you take somebody like Demis Asabis, one of the great pioneers and fathers of AI, he believes, I mean, and he says this passionately all the time, that he is on a mission to understand the entire universe through AI. I mean, you know, It's literally something he says all the time. Now, if I read your book correctly, you think he's just wrong. Demis has a fantastic understanding of things. I wouldn't say he's wrong. I would put the emphasis differently. I think AI as a scientific tool, just like the much better microscopes, such as carbon dating, radioactive dating, can vastly expand what we do.

20:09But if we slavishly follow it, we're not going to be as innovative. And if someday it became so smart that we are completely unnecessary, then I think our capabilities, our collective knowledge would start declining. In the same way that we're seeing it in classrooms, where generative AI sometimes is very powerful in helping you with homework. But if you let the students become slavishly followers of generative AI, their learning deteriorates. You need the humans to work with AI. And that's why AI as a tool is a very important part of this agenda of pro-worker AI and pro-human AI.

20:52Robert Colvile:Yeah, I was going to ask you about the whole education point. I was very struck, actually. There was some data out just the other day, a study of Chinese kids using AI, which, as you may have seen, showed that the 80 percent who used AI all the time ended up with significantly worse exam results than those who didn't use AI. Now, passing exams is not the be all and end all. But one of the things I'm massively struggling with is we do have to incorporate AI in the classroom, but we've got to do it in a way that doesn't turn, you know, essentially kids' brains into sludge. It's a very big and important challenge.

21:27It is. And that's, again, another one of a directional challenge. What is the industry investing in? It's investing in getting AI in the hands of every 11-year-old, every 10-year-old. It's investing in tutorial programs and things like that, that cut out the teachers so that the parents or the students learn without the teachers. And it's investing in tools that reduce the effort from the students. I think all we know from the field of education suggests that teachers are critical. Effort is critical. It's just like learning for kids and actually for adults too. It's like muscle building. You have to get those muscles exercised.

22:07So the tools we need are very different than the ones that are being developed and marketed right now. First of all, you need to empower the teachers so pro-teacher AI can make individualized, much more targeted education with still the teachers and inspiration to students, a very important part of learning. They can get the students to put the effort and then help them at the margins, which then is this complementarity point that I'm talking about. But it's an excellent example, thanks for bringing it up, Robert, that shows how the direction is so important. The details are going to matter. We need to understand those details.

22:42We need to understand what's on the menu. But absolutely, we need to choose the right direction. Otherwise, we're doomed.

22:48Robert Colvile:Yeah, well, certainly one of the ideas of yours that I thought was most compelling was this idea use AI to evaluate which kids will benefit more from personal intervention by human teachers. And I think that is absolutely the right direction. Now, frustrating, we're almost out of time. I'm going to do one, what we would call in the UK, handbrake turn. I'm going to ask you about one other issue, because on one issue, you are unusually, I would say, optimistic compared to many economists at the moment. And I thought it would be good to end on an upbeat note. And it is that you've recently published a paper where you believe that you have demonstrated that we shouldn't be as panicky about this collapse in the birth rate all over the world.

23:30Robert Colvile:That you see that where you have a declining young population, and this will seem slightly paradoxical given everything we've been talking about in these two episodes. you actually see the kind of investment by companies that actually increase national income and increase, I assume, individual income to compensate for the declining birth rate. Yeah, that's actually indeed a paradoxical result when you first put it into context of many other things that have been written. But the reason is very much congruent with something that I emphasized earlier on. Labor scarcity helps. Labor scarcity helps because it is more conducive to wage growth.

24:14It encourages employers to invest in workers because workers are scarce. You don't want to let the workers go. You want to get more out of those workers. And it also makes automation not as bad. Automation is going to be part of our future. I love automation. I don't want to go back to the days where we use shovels and we don't have cranes and people in the factory do such dangerous things that they lose their hands and fingers. But if automation just displaces workers, that's the problem. But if workers are scarce and automation is sort of helping with rectify that scarcity, that's also good, more productivity improving.

24:50So, scarcity provides more solutions, but we have to make the right investments. We have to really support the technologies that deal with that scarcity and we have to support training and other programs so that that labor scarcity actually translates into wage growth, into shared prosperity. So it's not an automatic thing.

25:13Robert Colvile:Brilliant. Darren, thank you so much for joining us over these two compelling episodes. So I'm going to sum up by saying that on the one hand, we do have to be concerned with the way that AI is being developed and we all have to essentially try and take control and move it towards a more worker-friendly path, perhaps not panic quite as much as many people have been doing about the declining birth rate. But Darren, brilliant to have you on the shows and I hope you'll come back when you come up with your next book. All the best from me. Thank you very much, Robert. It was a true pleasure to talk to you.

25:47Thank you. Goodbye.