A World Without Work - Daniel Susskind

In the future, we may face ‘structural’ technological unemployment in the labour market – where there is no longer enough work to occupy the human workforce. This lecture explains how such a phenomenon is possible at all, particularly given that repeated bouts of automation anxiety in the past have turned out to be wrong. Understanding this challenge is critical given recent claims by the leaders of the large technology companies – that they hope to build an AI that can outperform human beings at every economically useful task, within a decade. 

This lecture was recorded by Daniel Susskind on the 20th of April 2026 at Bernard’s Inn Hall, London

Dr Daniel Susskind is a writer and economist. He explores the impact of technology, and particularly AI, on work and society. He is a Research Professor at King’s College London, a Senior Research Associate at the Institute for Ethics in AI at Oxford University, a Digital Fellow at the Stanford Digital Economy Lab, and an Associate Member of the Economics Department at Oxford University. 

His new book, Growth: A Reckoning (2024), was chosen by President Obama as one of his ‘Favourite Books of 2024’ and was a runner-up for the Financial Times Business Book of the Year 2024. He is also the author of A World Without Work (2020), described by The New York Times as "required reading for any potential presidential candidate thinking about the economy of the future” and a runner-up for the Financial Times Business Book of the Year 2020, and co-author of the best-selling book, The Future of the Professions (2015). His TED Talk, on the future of work, has been viewed more than 1.6 million times. He is currently working on his next book, What Should Our Children Do? How to Flourish in the Age of AI. 

Previously he worked in various roles in the British Government – in the Prime Minister’s Strategy Unit, in the Policy Unit in 10 Downing Street, and in the Cabinet Office. He was a Kennedy Scholar at Harvard University

The transcript of the lecture is available from the Gresham College website: https://www.gresham.ac.uk/watch-now/world-without-work

Gresham College has offered free public lectures for over 400 years, thanks to the generosity of our supporters. There are currently over 2,500 lectures free to access. We believe that everyone should have the opportunity to learn from some of the greatest minds. To support Gresham College's mission, please consider making a donation: https://www.gresham.ac.uk/get-involved/support-us/make-donation/donate-today

Website:  https://gresham.ac.uk

X: https://x.com/GreshamCollege

Facebook: https://facebook.com/greshamcollege

Instagram: https://instagram.com/greshamcollege

Bluesky: https://bsky.app/profile/greshamcollege.bsky.social 

TikTok: https://www.tiktok.com/@greshamcollege

Support Us: https://www.gresham.ac.uk/get-involved/support-us/make-donation/donate-today

Transcript

SPEAKER_00 0:04

Please join me in welcoming Professor Daniel Suskins. Thank you so much. Well, thank thank you. Thank you. Thank you so much for that warm introduction. A great pleasure to be with you all this evening for my fifth lecture, A World Without Work. And for those who have been following the lectures, you will know that the central economic argument that has run through all these lectures is that new technologies have two very different effects on the world of work. There are two different forces at play when new technologies are invented. On the one hand, machines substitute for human beings when they displace us from doing particular tasks and activities, reducing the demand for us to do that work. And that's relatively easy to see. But on the other hand, machines complement human beings when they raise the demand for the work that we do at other tasks that haven't yet been automated. And until now, again, what we've seen in previous lectures is that this balancing act between these two forces, these two different effects that technology can have on the world of work, has tended to tip in favor of the helpful complementing force. It's tended to outweigh the harmful substituting force. And that explains essentially, again, as we've seen, why past anxieties about the impact of technology on work, the idea of technological unemployment, the idea that some people might find themselves without work because of technological progress, why those worries have been misplaced. In this clash between these two fundamental forces, people have until now tended to pick the wrong winner. It's why, as we saw the Luddites, so those technological vandals who went around industrial Britain in the 18th century smashing new technologies from spinning jennies to roller splitters, why they were wrong to worry. They picked the wrong winner. They worried too much about the substituting force, didn't focus enough on the helpful complementing force. And fast forward to today is why those who worry today, like those who recently set fire to driverless cars in Los Angeles in the US, are wrong as well. They too tended to pick the wrong winner in this battle between these two different forces. Until now, we have lived in what I call an age of labor. A time when there has always been enough work for human beings to do. And it seems to me, at this time of remarkable technological progress, asking the question, might the age of labor come to an end, is a really interesting and important question to be asking. In the last lecture, I argued that in the short and medium run, the age of labor is not coming to an end. I said there were two different types of technological unemployment, two different ways that people might find themselves without work because of technological change. There was the frictional type, where there is enough work for people to do, but for various reasons, and I explored what those reasons were, people might not be able to do that work. And then there was also a structural type where there just isn't enough work for people to do full stop. But I said that for now and in the medium run, our focus ought to be on the frictional type. Our worry should not be that there isn't enough work for people to do. There is going to be work, but for various important and difficult to resolve reasons, people might not be able to move in to do the new work that has to be done. Today, though, I want to turn to this second type of technological unemployment, the structural one. And what I want to do is I want to explain why, as we look further into the 21st century, the problem that we might face could shift from that frictional problem where there is work but for various reasons people can't do it, to one where there isn't enough work to be done full stop. And I want to set out some of the problems that we might face in that world. And in particular, I want to do five things tonight. I want to set out how the technological changes that are taking place might strengthen that harmful substituting force, but also explain how those very same technological changes might also weaken that helpful complementing force. I want to describe a change of heart that has taken place and is still taking place in the economics profession from a sort of benign optimism about the impact of technology on the labor market to a more uh to greater concern uh about the possible impact of technology on the world of work. I want to set out the problems that I think we will face in a society if we do have to confront this structural technological problem. And then finally, I want to close, in spite of what might sound like a relatively pessimistic uh you know, uh scenario, I want to explain why, nevertheless, I remain optimistic uh in light of these technological changes that are taking place. So that's the plan. Uh and as you heard, I'll talk for about 45 minutes or so, and then we'll have time for me to hear some reflections and take some questions as well. So let's dive straight in. And the the first aspect of the story that I want to tell is how the technological changes that are taking place at the moment strengthen that harmful substituting force. And in a sense, this is the most obvious part of the story. Again, for those who have followed uh the lectures until now, one idea that I've returned to time and time again is what I've called task encroachment. The idea that the best way to capture the impact of technology on the world of work is to think of what these technologies as doing is just gradually but pretty relentlessly encroaching ever further into the realm of tasks that until recently only human beings alone could ever do. And whether it's manual tasks or manual capabilities, those that involve dealing with the physical world, or our cognitive capabilities, those that draw on our capacity to think and to reason, or our effective capabilities, those that draw on our capacity for feeling and emotions, what you see and what I've described are how machines gradually, but again pretty relentlessly, are just encroaching in more and more areas of our areas of activity that once required these sorts of capabilities from us. It's this process of task encroachment that I think has really important implications for the world of work in general. And when we think about the generative systems that have caught people's imaginations in the last few years, you know what these technologies are capable of doing, and again, I've explored this lots in in previous lectures, it's remarkable. But the point I've made again and again is that this is a remarkable chapter in a far longer story. And again, this longer story is this story of task encroachment. These technologies gradually but pretty relentlessly becoming more and more capable. But something quite interesting is happening at the moment, uh, which is that if you listen to the leaders of the large technology companies, whether it's Demis Asabis at DeepMind, Dario Emoday, at Anthropic, Sam Altman at OpenAI, all of them, as we've explored before, are saying roughly the same thing, which is that within a decade, they are going to build systems that can outperform human beings at every cognitive task that they do. All economically useful tasks is another way that some put it. Now, I know the temptation is for many of us to want to dismiss this as science fiction, and I think, you know, the most important reason to take some of these claims with a pinch of salt is the strong financial interest that these companies have in talking up the capabilities of these technologies. But nevertheless, I think there are three important reasons that I've emphasized in the past for taking these sorts of claims seriously. The first are just the extraordinary financial resources that we're putting into developing these technologies. Uh, we I think this year we're going to invest something like two times what we did during the entire 10-year project to put man on the moon, the Apollo project. We've never invested so much in the pursuit of a single technical problem. Um this is an interesting chart that I saw, in fact, just this week. Um, although you might not be able to see it at the back, that sharp red line there showing expenditure on data centers, and then all the other mega projects that we might be able to conceive of in recent history, you know, falling well below in terms of total expenditure. The other aspect of this, though, is just the extraordinarily talented people who are going into developing these technologies. This is the chart that really captures that for me, showing share of bachelor's degrees in the US by major. What you can see is that in 2022, the proportion of bachelor's degrees in computer science was neck and neck for the first time in history with the proportion of all of bachelor's degrees in all of humanities combined. You know, there's a real sense, particularly in the US but elsewhere as well, that we're currently re-engineering higher education in pursuit of these technologies. And finally, and almost more importantly, these, you know, we're making progress. These companies are making progress. Um and this is something that I've seen in my own research, is that you know progress is so striking, it's quite difficult at the moment to, and this is what this fascinating article in the Financial Times described a couple of years ago, but it's still the case, uh, it's quite difficult at the moment to generate benchmarks quick enough with which to adequately measure the progress that's taking place. So I say all of this, just the extraordinary inputs that are going into developing these technologies and the progress that is being made, the outputs that we're achieving, to just make this first point, which is that if we think of this as just this ongoing story of task encroachment, these systems gradually but relentlessly taking on more and more activities that until recently only human beings alone could do, the first part of the story is that this substituting force, because of this technological progress, is getting stronger. Um and that I think is hard to, that that conclusion is pretty hard to avoid. But there's another far subtler part to this story, uh, which is how that very same process of task encroachment might also weaken the helpful complementing force. The force that has made sure there has always been enough demand for displaced workers to do tasks that haven't been automated in the past. Now, I explored this in greater depth in previous lectures, but I just want to do a recap for those who might not have been there. There were three different ways I thought about how this helpful complementing force might work. What I call the productivity effect, the bigger pie effect, and the changing pie effect. And I want to just take each of these in turn and think about how this process of task encroachment might affect each of these helpful forces. So, first, the productivity effect. And as I said, this is perhaps the most obvious way that the complementing force has helped human beings and continues to help human beings, is that new technologies, even if they displace workers from performing particular tasks, often make other workers more productive at unautomated tasks. So here we can think of technology complementing, helping human beings directly by making them more productive or more efficient at performing particular tasks. And if that productivity increase is passed on to consumers and perhaps a better quality product or lower prices, what we'll see is increased demand for those goods and services and increased demand for the workers who have to produce them in the first place. And you can see this productivity effect appearing in lots of different ways throughout economic history. So if you go back again to the Industrial Revolution, think of British weavers who were fortunate enough to find themselves operating the power loop. You know, they were able to weave more cloth, they became more productive. Or spinners who used a spinning jenny, they could be far more productive, they could spin more yarn. So these technologies in the Industrial Revolution, they you know, they weren't displacing those particular workers, they were making them better, more productive at their work, increasing the demand for what it was that they did. And fast forward to today, and you can see the productivity effects working again in lots of different ways. So take a driver who uses a sat nav system to follow unfamiliar roads. Or think of an accountant who uses an Excel spreadsheet or some other computational software to do harder, more intractable tax calculations. In all these different cases, if these productivity improvements are passed on to consumers in the form of lower prices, better quality products and services, you're going to see an increase in demand for those goods and services, and again, an increase in demand for the work of human beings to do all the activities required to produce those things. So that is the productivity effect. And you can see how it helps to you know, you can see how this this you can see how technology might help, might increase the demand for workers in this sort of way. But in the future, you know, new technologies are no doubt going to make some people more productive at certain tasks. But, and this is just the key point, this is only going to continue to help workers if they remain better placed to do those tasks than a machine. And as task encroachment continues, that becomes less and less likely for more and more tasks. So take sat nav systems again. You know, today they make it easier for taxi drivers to navigate unfamiliar roads. It makes human beings better at the wheel. But that is only going to be true so long as human beings are better placed than machines to steer a vehicle from A to B. But in the coming years, indeed already in some parts of the world, that's no longer the case. You know, software drives cars more efficiently and safely than us. And at that point, it doesn't matter how good people are driving a car with or without a sat nav, the machine will simply do it instead. Chess, I think, provides another really interesting illustration of how the productivity effect might fade away in the years to come. So this is a moment I hope lots of you are familiar with. It's when Gary Kasparov in 1997, who at the time was the world chess champion, was beaten by a computer system owned by IBM called D Blue. And it was an amazing achievement, and it's become part of the sort of you know the intellectual canon of uh of um of uh of artificial intelligence. What's quite interesting about this though is that if you look at how Kasparov tried to make sense of this moment in the years and and decades that followed, he he fell back on an idea, phenomenon that he called centaur chess. And the idea was that uh centaur chess involves a a human player and a chess playing machine working together as a team. And Kasparov's thought, uh, and indeed his hope, in light of the defeat that he had in 1997, was that such a combination, human plus machine, would beat any chess playing computer alone. Uh, that a human plus machine in a way would be more productive than just a machine working alone. And for those of you who are interested, this article captured that sort of self-reflection from Kasparov really nicely in the New York Review of books in 2010. I encourage you to have a look. But you can also see it in action. So a website, playchess.com, in 2005, online chess playing site, hosted what it called a freestyle chess tournament in which anyone could compete in teams with other players or importantly with computers. So normally these sites at the time had these anti-cheating algorithms that were designed to prevent or or at least discourage players from cheating by using machines, but not this time. In 2005, that wasn't the case. And what what happened um was that sentence in the middle of that paragraph, they're reporting what was found, the teams of human plus machine dominated even the strongest computers. So it looked like Kasparov's uh centaur chess, and I thank um GPT for for generating this image for me. Uh you know, this really was the productivity effect in action. It was an example of new technologies making human beings better at what they do. Kasparov felt defeated, but he took comfort from the fact that if he worked alongside a machine, he could be any other machine. Human plus machine was better than machine alone. The problem is that Kasparov's Centaur has now been, in some sense, decapitated. And again, this was GPT's interpretation of that idea. So what happened in 2017, so uh uh a few years on from Kasparov's self-reflection, is that Google took uh a system called Alpha Go Zero, which was um it was a Go playing machine, played the the Japanese board game Go, uh, and it and it tweaked the algorithm so that it could play other games as well, including chess. And it gave it the rules of chess, and and they called this new system uh Alpha Zero. Um and what was I mean, there were a few things that were really interesting about this. So this was you know almost 10 years ago. It did a few things that were really interesting. One is um that instead of absorbing the lessons of past games by the best human chess players, which was one of the ways these systems had worked up until then, this machine had no human input at all. Um it self-trained, it learned by playing itself at the game of chess. And yet, after only a day of self-training, it was able to achieve unparalleled performance. It beat the best existing chess playing computer in a hundred-game match, and it didn't lose a single game. Um and what was you know fascinating at that particular moment uh was that it was very, very difficult to see any longer what role human beings might have, a human player might have alongside a machine like Alpha Zero. Uh Tyler Cowan, the the uh the economist, put it really well when he said, quote, the human now adds absolutely nothing to man machine chess playing teams. Indeed, there's a sense in which having a human being at the table alongside that machine just got in the way. Um but there's a really there's a deeper lesson here, and I think it's a really important lesson, which is that Kasparov's experiences in chess led him to declare that human plus machine, human plus AI partnerships, are the winning formula not only in chess, but across the entire economy. And you hear this slogan, those who um you know, those who are interested in in artificial intelligence, particularly in the corporate setting, will have heard something like this: that human plus AI is better than you know than human beings. But I think 10 years ago, AlphaZero's victory showed that this is wrong. That is only true. Human plus machine, human plus AI is only stronger so long as the machine in any partnership cannot do whatever it is that human beings are bringing to the table. But as machines become more capable, as that process of task encroachment continues, the range of contributions made by human beings diminishes until partnerships like this just essentially dissolve. The human in human plus machine becomes redundant. That's the direct effect, the productivity effect. But the what's important to remember is that technological progress also complements human beings indirectly. It increases the demand for human beings to do unautomated tasks in two other important ways. One of them is the what I call the bigger pie effect. So remember, if we think of the economy as a pie, what technological progress does is it makes the pie bigger. As productivity increases, incomes rise, more goods and services are demanded, and so demand grows for workers to perform tasks that haven't been automated to produce those new goods and services. And this bigger pie effect is you know incredibly consequential. You know, over the last few hundred years, global economic output has soared. So what you can see is that for most, in fact, you know, this is from 2000 to zero. But in fact, for most of the 300,000 years that human beings have been around, it's the same story. Economic life is essentially stagnant. Whether you're a hunter-gatherer laboring away in the Stone Age or someone's toiling away at the start of the Industrial Revolution, you know, your lives would have differed in lots of respects. But fundamentally, from an economic point of view, it would have looked remarkably similar. You know, you would have been stuck in a pretty unforgiving, relentless struggle for subsistence. So economic growth only really begins, modern economic growth, growth that is significant and sustainable. Only really begins 200, 300 years ago. But the consequence is that over the last few hundred years, global economic output has soared. And this growth in output was overwhelmingly driven by technological progress. So again, if we think of the economy as a pie, the UK, for instance, has seen its economy grow 113-fold from 1700 to 2000. And that's nothing compared to other countries that were less developed at the start of this period. So over the same 300-year period, the Japanese economy grew 171-fold, the Brazilian economy almost 1,700 fold, the Australian economy over 2,000-fold, the Canadian about 8,000 fold, the US economy over 15,000 fold. The reason all this matters is because this bigger pie effect, the way in which technological progress makes the economic pie bigger, has often been a source of optimism for economists trying to think about where demand might come from for displaced workers. So David Otter, one of the great labor market economists of the moment, wrote, people are wrote when thinking about the impact of technology on work a few years ago now, people wrote, he wrote, people are unduly pessimistic. As people get wealthier, they tend to consume more. Again, it's this bigger pie effect. And so that creates demand. Again, the idea of more income in the economy, more demand for goods and services, more demand for the work of human beings to produce those goods and services. Or Kenneth Arrow, one of the great late economists, the economy does find other jobs for workers. When wealth is created, people spend their money on something. So again, you know, historically, this bigger pie effect has been a source of optimism, a source of reassurance that as technological progress continues, as this process of task encroachment goes on, the economy is going to grow, incomes are going to rise, demand is going to increase for goods and services, people who are displaced are going to find jobs elsewhere in the economy to do. But in the future, economic pies will no doubt continue to grow. Incomes are going to be larger than ever before, and demand for goods is going to soar. That seems very likely. And yet, it seems to me we cannot rely on this to necessarily bolster the demand for the work of human beings as it has in the past. Why? Because just as with the productivity effect, the bigger pie effect, this sense in which the economy just gets larger and larger, incomes rise, is only going to help if people, again, it's the same reasoning, if people, rather than machines, remain better placed to perform whatever tasks have to be done to produce those goods. And as task encroachment continues, as these technologies take on more and more tasks and activities, that becomes less and less likely. And you can see this already unfolding in certain parts of the economy. So take the UK agricultural sector. But it's just not created more work for people to do. So that top line there showing how UK agriculture now today produces about five times what it did back in 1860. But look at that bottom line there. It requires just 10% of the number of workers to do it. Far bigger pie, but in absolute terms, not even in percentage terms, just in absolute terms, it requires far fewer people to do it. 10% of the number of workers, five-fold output. It's not just a story about agriculture either. Think about UK manufacturing, the same story. The sector now produces about 150% more than it did in 1948. Look at that top line there. And yet again, look at that bottom line, requires just 40% of the number of workers to do it. That you know slice of the pie is even bigger today than it was 70 years ago, and yet again fewer and fewer people required to do it. Now, you know, these stories are clearly only unfolding in you know particular corners of the economy at the moment, but they capture what I think is the essence of the problem with the bigger pie effect. That rising incomes might lead to rising demand for goods and services, but that doesn't necessarily mean rising demand for the work of human beings as well. It's only going to be the case if human beings remain best placed to do whatever tasks and activities have to be done to produce all those goods and services. Uh, and as task encroachment continues, that seems less and less likely. The other indirect way that the complementing force has traditionally helped um to increase the demand for displaced workers to do uh tasks that haven't been automated is that is that it it's not only made the economic pie bigger, but in a sense it's also changed the ingredients in the pie as well. So uh what's happened is that as time has passed, people have spent their growing incomes in very, very different ways. So they've changed how they spread it across existing goods and services, but they've also developed tastes for entirely new goods and services too. And what that has meant is up until now, new industries have been created, new tasks have to be done that haven't yet been automated, and that has meant more demand for human beings. So 300 years ago that British Pi was made up of farms, 200 years ago it was made up of factories, and today it's overwhelmingly made up of offices. And the story has been that people displaced from tasks in those old types of Pi could tumble into performing new tasks that hadn't yet been automated in the new bits of Pi instead. Importantly, I I think it's you know uh this changing pie effect is again a kind of common source of optimism. And and the point that's often made, and I think it's a really good one, is that it's often very difficult for us in any particular moment to anticipate how the economy might change in years and decades to come. And you know, one way to see how difficult it is is to think historically. So imagine yourself transported back to the past. You're standing on a street in Britain in the 1780s, you're thinking of joining the Luddites, you know, to go out and start smashing some machines. Most people around you are still working on farms and factories. It would have just been impossible, could anyone have predicted back then, that a single healthcare organization, the National Health Service, would eventually employ more people than the number of men working on all the farms in the country at the time. You know, back then there was no healthcare industry, certainly nothing funded by the state. Everything was voluntary. And yet within a century and a half, the NHS would be the biggest employer in Europe, the fifth largest in the world. Or put yourself in the early 1990s again, um, before the internet era really begins. No Yahoo, no Netscape, no Auto Vista, no Google, no MSN. Now imagine I came up to you and said, within a couple of decades you're going to find a job as a search engine optimizer or as a digital social media manager. How would have you reacted, right? Complete amusement. Or even go back as recently as 2021. You know, would the prospect of working as a prompt maximizer have meant anything to you in a world before ChatGPT? It's not simply that these roles were hard to imagine. You know, they were just impossible to imagine. They relied on concepts that just didn't yet exist. Time and time again, what we've seen is the technologies that would transform our lives and provide work for us to do just hadn't yet been invented. And again, this observation has been a kind of continual source of optimism for those looking for where demand might come from for displaced workers. So David Dorn, another one of the great labor market economists a few years ago, writing technological progress will generate new products and services that raise national income and increase demand for labor in the economy. New products and services. Sense something novel, something new, something that didn't exist before. Or Joel Mokier, who won the Nobel Prize in Economics very recently, wrote how the future will surely bring new products that are currently barely imagined, but will be viewed as necessities by citizens of 2050 or 2080. And people can find work producing those sorts of goods and services instead. But and it's the same, again, it's it's exactly the same reasoning as before. You know, the economic pie might change. Just as the economic pie might get bigger, the economic might change. And perhaps it will change in ways that are inconceivable to us today. But in exactly the same way. As that process of task encroachment continues, it becomes more and more likely that machines rather than human beings will be better placed to perform whatever new tasks have to be done. And if you look at newer parts of economic life today, you might worry that you can already see something like that unfolding. So in 1964, the most valuable company in the United States was ATT. It had 758,611 uh employees. In 2018, it was Apple with only 132,000. 2019, Microsoft with only 131,000. There was an interesting piece of work done looking at new industries a decade into the 21st century. They found that new industries that were created in the 21st century in that first decade, a remarkable time of technological technological progress and creation, accounted for just half a percent of all US employment. So just to gather all of these observations together, it seems to me that this process of task encroachment might not only, you know, there are good arguments to think it might not only increase the strength of that harmful substituting force that has tended to displace workers in the past from performing particular tasks and activities, but it might also weaken the helpful complementing force that we've traditionally relied upon in the past to create sufficient demand for those displaced workers. And that I think is how we might find ourselves in a world with less work. That's the scenario that I see. Um how we might find ourselves in a situation with structural technological unemployment. As time goes on, machines continue to become more capable. They take on more and more tasks that once fell to human beings. The harmful substituting force displaces workers in a familiar way, as it's done for centuries. And for a time, the helpful complementing force continues to raise the demand for displaced workers elsewhere, whether it's through the productivity effect, the bigger pie effect, the changing pie effect. And our challenge, so long as that goes on, and that I think is our challenge for the moment, and I think it's our challenge for the medium run, our challenge is going to be this frictional technological unemployment. There is going to be enough demand for the work of human beings. There are going to be enough jobs for people to do, but for various reasons, it's going to be difficult to get people into those jobs. Set out in the last lecture, what I think those reasons are reasons of skills mismatch, place mismatch, and identity mismatch. But what I worry about is the scenario where as task encroachment goes on, and more and more tasks fall to machines, that helpful complementing force is weakened as well. That we find ourselves retreating to an ever-shrinking set of tasks in which we retain the upper hand over these technologies. And the key point here is that there just is no reason to think that the demand for those particular tasks will be large enough to keep everyone in well-paid work doing them. You know, there's a lot of a lot of people spend quite a lot of time today trying to imagine what, okay, what particular tasks and activities might machines never do in the future. Um and you know, it's possible to, of course it's possible to come up with lots of tasks and activities that machines might never do in the future. Um but it's one thing to say that there are activities that machines might never do in the future. It's another to say that there's going to be enough demand for those particular tasks and activities to keep everyone in well-paid work. And if that's the case, then I do think our challenge becomes one of structural technological unemployment. So, in a sense, the world of work comes to an end in that famous phrase, not with a bang, but with a withering. A withering in the demand for the work of human beings as that harmful substituting force for the first time gradually overruns the helpful complementing force, and the balance between the two no longer tips in favor of human beings as it has until now, as it has in favor, uh as it has in this um age of labor that we've been fortunate to live in uh over the last few hundred years. Now, this argument that I've set out for why I think structural unemployment is some structural technological unemployment is something we ought to take seriously, um, comes in the context of a change of heart that has taken place in the last few years within the um economics profession with respect to thinking about the impact of technology on work. Um the great British uh economist Um John Maynard Keynes, a famous quotation uh attributed to him is you know, when the facts change, I change my mind. What do you do, sir? Uh well, as the facts have changed, economists have changed their mind uh about the impact of technology on work. Um and for those of you who have followed the lectures, what you will have seen in earlier lectures is that actually in the beginning, for most of the 20th century, economists found themselves in possession of a pretty benign view of the impact of technology on the labor market. Um it's the one we saw where it was this idea that technological change was skill-biased. Um what that meant was that it was biased towards skilled workers in the economy relative to less skilled workers in the economy. And that was captured, you'll remember, for those of you who follow, through this idea of the skill premium, which was the ratio of what you would get paid if you went on average if you went to college relative to what you would get paid if you just went to high school. What you saw in many parts of the world in the second half of the 20th century was something like what's set out here with the US skill premium. A rising skill premium. The wage that you would get if you went to college relative to um if you went to just high school was rising. Um and the argument was, you know, why is it rising? Well, because technology is skill biased. Because technology in the second half of the 20th century, whether it was the personal computer, whether it was all this innovation and software, whether it was the internet, was increasing the demand for skilled workers who were capable of using all these new technologies and putting them to effective use. Increasing the demand for people who could use word processing software, increasing the demand for people who knew how to organize numbers in an Excel spreadsheet. Um what's interesting, of course, about this is I noted in the past was that you know this increase in the US in the skill premium is happening at the same time as the number of people who are coming out of college around the world, is you know, taking off in the second half of the 20th century. Um wages are rising, even though the supply of educated workers is also rising, which suggests that the demand for those educated workers is rising even faster than the supply of those uh educated workers is rising. And and so what emerges, and I I've I've returned to this idea uh several times, is the idea that there is this race uh between education and technology, and that the the way for us to respond to technological disruption, and this is how most economists and policymakers were thinking at the turn of the century, is more education, to provide people with the skills, with the training, uh, with the education that allows them to put these new technologies to productive use. Um and it it's and it's benign because in that story, no workers are left worse off. Um it's true that skilled workers do better relative to unskilled workers, but in that story, and most of the models and mathematical models and the empirical work that economists were doing, it also suggested that those low-skilled wages were rising as well. It's just that for the higher skilled people, their wages were rising even more. It's a relatively benign story. Everyone benefits from technological progress, some people benefit more than others. It's that benign story that I think has has uh has started to unravel in recent years. Um and for those of you who are interested in seeing how uh this uh you know, seeing how economists are thinking about it today, um, I'd encourage you to have a look at um, for those of you who are interested in some of the more um academic research, uh really interesting recent uh gathering of uh of you know leading economists looking at the impact of transformative AI. So that's AI, you know, transformative AI is how economists at the moment are talking about um what I call artificial general intelligence, those systems and machines that are capable of doing all cognitive tasks that we do more efficiently and more effectively. Some of the accounts, some of the papers, some of the research in here tell a very different story about the impact of technology on the world of work to that sort of relatively benign view that emerged from the skill bias story. For those who are less interested in looking at the um looking at the academic papers, uh I um uh edited a uh a collection of essays, many of which were uh drawn from that uh academic research uh into this volume. Um again, um you know, uh encourage you to have a look. But there are conversations happening now. This this is the main point. There's conversations happening now which would have just been unimaginable um only you know a year or two ago. So earlier in the year, David Orter, Anton Koronek, and Natasha Sarin, what if labor becomes unnecessary? Um, just at the start of this month, exactly the sentiment that I'm describing. Another piece in the New York Times reporting on some uh empirical work done on that how economists' views of the impact of AI on work are changing. Economists once dismiss the AI job threat, but not anymore. Remarkably interesting and important shift is taking place in how economists think about the impact of technology on work. Um the problem though, and you you you come up against this problem, is that um you when you try and pin economists down on exactly what this might mean, you get sort of scenarios that look something like this uh in the Financial Times um uh at the end of last year, where this is them plotting um GDP per capita. Um there's a sort of middle scenario, the AI boosted growth path, where things just continue as they did before and AI helps you um to grow as you did before. Then there's the end of scarcity story where all of a sudden technological progress leads to some extraordinary explosion in economic prosperity, and we soar to unimaginable greater heights, and then there's also the sort of human extinction story where uh everything you know collapses into economic uh penury. So you you get these kind of extraordinary divergences of opinion. And you know, that there is a remarkable amount of uncertainty about the moment that we live in and about the impact of, you know, the likely impact of technology on work as we look further into the 21st century. So the the spirit in which I think the argument that I've made this evening ought to be taken is as a scenario. Um, is the idea of structural technological unemployment inevitable? No, absolutely not. But is it possible? And is it possible, particularly in light of the sorts of technological claims that the leaders of the large technology companies are making at the moment? I think it is, yes. And what I find really um uh significant is that at the moment, very, very few politicians, policymakers, business leaders are taking these sorts of scenarios seriously. And I and I think that's uh you know a big mistake. You know, the sort of motivating observation for a lot of that work on transformative AI is that exactly as we saw earlier in the lecture, we are investing a huge amount of financial resource into developing these technologies more than we've ever invested in the pursuit of any single technical problem. We are investing nothing compared to that in terms of understanding what the likely impact of these technologies is going to be, not in the next few years, but further into the 21st century. So I think we ought to be taking this idea of structural technological unemployment seriously as a scenario. And I think there are four big problems that we need to grapple with if we do take this scenario seriously. The first is the distribution problem. It's an economic problem. Today the labour market is the main way that we share our economic prosperity in society. For most people, their job is their main, if not their only, source of income. How do we share out material prosperity in society when our traditional way of doing so, paying people for the work that they do, is less effective than in the past. That I think is the big economic problem. There's also, though, a second problem which has less to do with economics, which is the problem of contribution. You know, today social solidarity often comes from a feeling that everybody is pulling their economic weight through the work that they do and the taxes that they pay. And if people aren't in work, there's an expectation that if they're able to work, they ought to be willing to look for work. And a lot of you know welfare is conditional on actively seeking work. How do we provide people with an opportunity to contribute to society and to be seen by others to be contributing to society if our traditional way of providing people with an opportunity to make that contribution through the work that they do and the taxes that they pay is no longer available? I think that is a really significant problem. I think we spend a lot of time on that distribution question at the moment, on questions of distributive justice. How do we share out prosperity in society fairly? We spend far less time on this problem of contribution, on the issue of contributive justice. How do we provide everyone in the future with an opportunity to contribute and to be seen by others to be contributing as well? The third problem is the problem of power. In the future, our lives are likely to become even more dominated by a small number of large technology companies that are responsible for these developing these technologies. I think in the 20th century, our main worry about large companies tended to be their economic power. We worried about things like market concentration, predatory pricing. And you know, we developed a whole framework, competition law, antitrust law, for thinking about how to identify abusers of economic power and intervene accordingly to break it up, if need be. In the 21st century, I think our concern is going to be far less with the economic power of these companies, and far more, as I think we see already, with their political power and the impact they have on things like liberty, social justice, and democracy and whether they threaten those things. And the challenge, I think, for us at the moment is that while we've developed very good frameworks for thinking about the uses and abuses of economic power, we don't have anything analogous to that, to thinking about abuses and uses of political power. The final challenge is the challenge of meaning. You know, it's often said that work isn't simply a source of income, but it's also a source of direction and fulfillment in life. And if that is right, then the challenge of automation and structural technological unemployment isn't just that it hollows out our labor market, leaving some people without an income, but it might also hollow out that sense of meaning and purpose in life as well. You know, these are you know really important challenges. Again, do I think they're inevitable? No. But do I think there's a scenario in which these things are possible? Yes. And I think we have to collectively start grappling with these challenges. And yet, in spite of all of this, and I think this is really important, I do remain optimistic. Um and the reason for this is simple, which is that if we step back and think about one of the charts that I showed earlier, in the decades to come, technological progress is likely to solve the economic problem that has basically dominated humanity until now. So if we think of the economy as a pie, the traditional challenge, the challenge for 300,000 years basically, has been to make that pie large enough for everyone to live on. So at the turn of the first century AD, if the glow if you had taken the global economic pie and divided it up into equal slices for everyone in the world, most people would have had just a few hundred of today's dollars. Most people lived on or around the poverty line. If you had gone roll forward a thousand years, roughly the same would have been true. Roll backwards 300,000 years, it would have looked remarkably similar in every millennia then. But over the last few hundred years, as we saw, economic growth has soared. And this growth was driven by technological progress. Economic pies around the world, as we saw, have exploded in size. They've become much, much bigger. So today, global GDP per capita, the value of those individual slices, is already about$11,000,$12,000. So technological unemployment in a very strange way will be a symptom of that success. In the 21st century, technological progress will solve one problem: the question of how to make the pie large enough for all of us to live on, but as we've seen, it's going to replace it with four others. The problems of distribution, of contribution, of power, and meaning. And as I said, you know, these are huge problems. And there is going to be immense disagreement about how we ought to respond to them. But in the final analysis, it seems to me that these problems, however large they might be, and however intractable they might seem, and however much disagreement they might introduce, it just seems to me that they are far better problems to have to grapple with than the one that haunted our ancestors, not simply for centuries, but for millennia, uh, which was how on earth do we make this economic pie large enough for us all to live on in the first place? So I will finish there and just give you a little taster of what I want to talk about next time. I want to explore what I think our best response at the moment to these technological challenges is, which in my view is an educational one. Um but I also want to, drawing on some of the ideas I've explored today, suggest that there are limits to education, that it's not the sort of panacea that many politicians and policymakers present as a response to the technological disruptions that are currently underway. So thank you very much indeed, and I look forward to hearing some reflections and taking some questions. Thank you.