So please welcome Professor Daniel Suskind. Thank you very much indeed. It's a great pleasure uh to see everyone here uh this evening to talk to you about education and its limits. And what I want to do in the next 45 minutes or so are six different things with you. Uh the first is I want to uh set out some context for the particular moment we find ourselves in. The second is I want to talk about an idea which is very common in thinking about how we use education to respond to the challenges of automation, which is the idea of future-proofing. Uh, and I'm going to be quite critical of that idea. Uh I then want to think about three different dimensions in which we ought to be engaging with the idea of education, rethinking what it is that we are educating people in, how it is we are educating them, and finally when in their lives we are educating them. And finally, I want to close with some limits to education. Uh, people often talk, particularly among policymakers and politicians, as if education is a sort of panacea uh for responding to the challenges of technological disruption in the world of work. And and I want to set out some limits to explain why why that isn't so, why I think we need to go beyond education. So, first let me begin with some context. And for those of you who have been following the other lectures, you'll know that the argument I have made is that we live at a remarkable time. In the last few years, we have seen an extraordinary wave of so-called generative AI systems, whether it's Chat GPT from OpenAI, Claude Anthropic, Gemini at Google, Grok from X, whatever it might be, these are systems and machines that appear to be taking on more and more tasks and activities that until very recently we thought only human beings alone could ever do. And there's been a huge amount of excitement, I think, justified with respect to these technologies. But again, for those of you who have followed the lectures until now, you'll know that one of the arguments I've been making is that this recent burst of technological progress hasn't come out the blue. It's in fact a chapter in a far longer story. A story that's been unfolding now, in fact, for many decades. And again, it's a story where not just AI, but a whole variety of different technologies have gradually been creeping, you know, again, you know, falteringly, stutteringly, but fairly relentlessly into our working lives. So I think it's important as we think about the technologies of the last few years. Yes, they are remarkable, uh, but we ought to put them in the context of this longer story of technologies gradually but relentlessly becoming more capable. And this context is important because technological change isn't a new phenomenon, and until now, our response to technological change has essentially been more education. Again, those of you who have attended other lectures will will have picked up on this theme that more education in various forms has been how we have responded to automation in the labor market. At the start of the 20th century, the more here essentially meant more people, um, getting more people into education. And it was remarkably slow in coming. So in the nine in the 1930s, as some economists noted, the US was virtually alone in the world in providing free secondary school education. Uh but as time passed and as the 20th century unfolded, other countries caught up and copied that initiative. And today that sort of education is commonplace, it's widespread in most parts of the world. So by the end of the 20th century, what that more means in more education had changed. Uh it no longer simply meant educating more people, getting more and more people into school, uh, but more advanced education, uh, with a particular focus on going to college or attending university. And you can see this in the language that politicians and policymakers were using at the turn of the century. Uh, in 1996, US President Bill Clinton uh introduced a set of tax changes that he hoped would make the 13th and 14th years of education, so those more advanced years of education, as universal, in his words, to all Americans as the first 12 are today. A few years later, UK Prime Minister Tony Blair declared that he had no greater ambition for Britain than to see a steadily rising proportion gain the huge benefits of a university education. And in 2010, then President Barack Obama proclaimed that, quote, in the coming decades, a high school diploma is not going to be enough. Folks need a college degree, they need workforce training, they need a higher education. So, what should we do in response to the technological disruptions we see in the world of work today? How should we respond? Well, it seems to me that for now, more education remains the best response, the most effective response that we have. And the reason is if you again remember, again, for those of you who have attended previous lectures, one of the arguments that I've made is that there are two different ways in which we can think about the impact of technology on work. Two different ways that people might find themselves without work because of these remarkable technological changes that are taking place. The first is what I've called structural technological unemployment. And here there just aren't enough jobs to be done full stop. But there's also a second type of technological unemployment, a frictional one, where there are jobs, but for various important reasons, people are not able to move into those jobs to take up the work that is available. And the argument that I've made is that for now and in the medium run, the main challenge that we have in the world of work is not the structural one, the one that you know tends to uh pop up in lots of um popular commentary on the future of work, but it's the frictional one. It's that there is work out there, but for various important reasons, people might find it difficult to do it. And the best response we have to this frictional challenge in the world of work is more education. Uh, particularly because one of the main reasons that people are unable to do the jobs that have to be done out there is because they don't have the right skills or the right capabilities to do the work that has to be done. And so more education remains our best response. But what I want to explain is why the meaning of more education has to change given the nature of the technological changes that are taking place. And I want to turn now to focus on a pretty profound sense, I think, in which our current approach to more education uh is is misplaced. That it might have helped in the previous century, it's not going to help this time around. And I want to begin with that. And it's what I call the idea of future proofing. Um and the story I want to tell about the idea of future proofing actually begins back in 2009, 2010. Uh, and um before I began writing and researching and thinking about the impact of AI on work and society, I worked in the British government. Uh so I I first worked in the uh Prime Minister's Strategy Unit under Gordon Brown, and then in the policy unit in Downing Street under the coalition government. So with David Cameron and Nick Clegg, and this is in fact a photo of me 15 years ago, although you can't see my face, you can see the outline of my slightly more adventurous hairstyle uh in 2009, 2010. The the reason I go back to this moment though is because I I was there for a couple of years and then I left, and something quite interesting happened uh a year or two after I left, which was that the then Prime Minister David Cameron decided that the UK, well England and Wales in particular, would become the first place in the world to mandate that all children from as young as five had to learn to code. Um so in state-funded schools up and down the country, the old ICT curriculum, so teaching people how to do word processing, how to uh you know, word processing in Word, how to, you know, uh uh you know play about numbers in Excel spreadsheets, that was out, and in its place was an entirely new curriculum drawn up in partnership with a whole variety of professional institutions and large technology companies that would um that would play a really important role in um preparing the next generation, it was believed, for the future of work. So you can go back and have a look at the policy papers that were published at the time. Here it is, document published at the end of 2014. This year, England became the first country in the world to mandate teaching coding to children at primary and secondary schools. Um the then education uh minister Michael Gove uh described these reforms as being designed to equip every child, every child, with the computing skills they needed to succeed in the 21st century. And the point I would make is that you can understand, looking back on it, the promise of coding. So 2007, you know, Apple released the iPhone, the smartphone era begins. Uh, 2010, you have AI companies like DeepMind, fine uh DeepMind in particular, founded with the sort of unthinkably ambitious aim of solving intelligence. Um you had extraordinary hardware, then you had lots of sort of interesting software bubbling up. It was clear in the air that something interesting was happening from a technological point of view. And it was in fact this sort of technological activity in 2010, 2011 that led me to start working on the first book that I wrote, The Future of the Professions, which was looking at the impact of uh technology, particularly AI and the internet, on professionals, which we we published in in 2015. It was a really interesting technological moment. And again, the UK was a um really was a pioneer in mandating that all children in primary and secondary schools had to learn to code. And and what happened in the years that followed was that almost every country in the world, sensing that something really interesting was happening from a technological point of view, published an AI strategy or some form of an AI strategy. And a key component in almost all of these AI strategies that I looked at, again, was this idea that we needed to teach young people to code. Um, some countries took it pretty far. So, Finland, for instance, today celebrated for its educational standards, dropped long division from its national curriculum and replaced it with coding in 2016. And it was really hard for a while to find any country in the world that didn't have some version of teaching young people to code as part of their strategy. As time passed, enthusiasm for coding continued to grow. It was no longer just a way to prepare young people for the future of work, but it was also a way to help those already in work, uh perhaps in roles at risk of automation, move on to secure jobs. So at a new Hampshire rally in 2019, uh then President uh Biden explained to Pres now President Biden, then not President Biden, explained to the crowd assembled before him that he had been asked by President Obama to figure out the jobs of the future. Uh and he had this remarkable line in the speech. Anybody who can go down 300 to 3,000 feet in a mine can sure in hell learn to program as well. Anybody who can throw coal into a furnace can learn to program, for God's sake. Um, which was sort of remarkable. Fast forward to 2023, and then you get the release of uh uh Chat GPT by OpenAI. It took just two months to become the fastest growing consumer app of all time. Here was an AI system that could answer questions on anything in the world, uh, in more cases, in many cases, to a standard that seemed to rival what we were capable of doing. And what did it turn out to be particularly good at doing? What's the the punchline that I'm chasing after? What it turned out to be particularly good at doing was writing code. Um in March 2025, the founder of the uh OpenAI rival, Anthropic, a man called Dario Omode, predicted, on the basis of the progress that he had seen uh in AI-enabled coding, that within three to six months we may find ourselves in a world, quote, where AI is writing 90% of the code. And then in 12 months, we may be in a world where AI is writing essentially all of the code. At the time, that might have sounded science fictional. Uh, what is really interesting is that in January of this year, Anthropic reported that 90% of the code behind Claude Code, which is their code uh AI-powered coding assistant, 90% of the code behind this system was itself written by Claude Code. Now, why am I really laboring the this sort of coding saga? Why am I going on and on about this point? Because I think it's a really good example of how in the world we live today, it is incredibly difficult, and I'd say impossible, to really identify future-proof skills. We simply do not know enough about the relevant skills of the future to be telling young people that they ought to remake their entire career, their entire lives in pursuit of them. Again, just think of those words from the then education secretary Michael Gove, designed to equip every child with computing skills they needed to succeed in the 21st century. Not only did it not prepare them for the century, it barely prepared them for the next five years. Um I think the temptation for many politicians and policymakers today today is to want to dismiss this as an unfortunate blip. They just got it wrong. But the idea of future-proofing is nevertheless still a good one. I think that's a mistake. I think these leaders got it wrong because they believed in this idea of future-proof skills. That there is some set of valuable skills out there that AI is not going to be able to do for some time, if ever. And most importantly, that if we together think cleverly and smartly and thought, you know, thoughtfully enough about the future, we can figure out what those skills are going to be. Um it's not just policymakers and politicians either, as many of you will know, um, the idea of future-proof skills is is far more pervasive. So this is just one way to see the explosion uh in how people have appealed to the idea of future-proof skills, almost barely mentioned before 1980. So this is a Google Books N-gram view-ish. So it shows how often uh this phrase future-proof appears in all uh published material. Uh barely mentioned before 1980, and then you get beginning in a gradual uh ascent, and then beginning in 2010, again, that sort of moment of technological uh and that sort of you know new technology started to bubble up in really interesting and exciting ways, a sort of explosion in interest in this idea. It's a far more pervasive idea, this idea of future proofing. You know, career advisors rely on it when they tell young people what to study at school if they want a good job. Business leaders draw on it when they plan their learning and development strategies for their employees to follow. Researchers appeal to it when they claim, as you often hear, that some job X is at risk of automation, but other jobs Y are somehow out of reach. Parents turn to it, reflecting on what they ought to tell their children to do. Journalists use it, as we will all have read, when compiling lists of future-proof professions for their readers to focus on. And I know the idea of future-proofing sounds very reassuring, but I think it is, I'm increasingly coming to the view that it's it's an illusion. Um, and and not only is it an illusion, but it misunderstands the problem that we currently face in the world of work. And in my view, the fundamental problem that we face at the moment in the world of work is one of uncertainty. Again, this idea that it's just incredibly difficult to anticipate exactly what skills and what capabilities are really going to be most valuable and important in years to come. Um I think we are spending far too much effort at the moment trying to resolve this uncertainty, trying to think really cleverly about the future and figure out exactly what skills and capabilities can be done. And it's not just coding. Um, you know, you can flick through one of the enormous number of publications that have been written in the last few years that have tried to identify these future-proof skills, whether it's corporate strategies, newspaper articles, government reports, academic papers, whatever it might be, to see just how difficult it is and how often people trip up. Um, given the technological changes that are taking place, huge in scale, unpredictable in nature, it just seems to me to be a really difficult, an impossibly difficult exercise. Um I had a really um interesting experience last week, in fact, um uh reading the the tweets of this man, the great uh British mathematician, uh Timothy Gowers. Um he really is one of the great. So in 1998 he received the Fields Medal for his work in mathematics. So he's you know the essentially the equivalent of the Nobel Prize. And back in April, and I didn't see this at the time, but on the first, I should have realized given the date, the 1st of April 2025, he wrote this tweet about the idea that we might be able to use AI in mathematics. So it's finally happened. After several unsuccessful attempts, I found a prompt that got Grok uh to solve a maths problem, the well-known Dubnovy Blasen problem in graph theory. I've been working on it for over a year. Um, how long till it's better than human mathematicians across the board? Now, of course, April Fools, this was uh an April Fool's trip, there was no such system. In fact, I'm told that uh Dubovny Blasen is Czech for April Fool's, um, which made me smile. Now, the reason I've said this is uh the reason I put this up is because um only a few days ago, uh Timothy Gowers wrote this tweet. If you're a mathematician, then you may want to make sure you're sitting down before reading further. What he was referring to uh was this story uh from OpenAI in the last few days uh that a model has solved what's known as the planar unit distance problem. Um said to be the best known uh problem in in combinatorial geometry. Uh yeah, an incredibly well-known problem among mathematicians. Um we went from a situation in which the idea of using AI in mathematics was so far-fetched it made sense to you know have uh do an April Fool's joke on it, to just over a year having systems that are solving some of the most famous problems in mathematics. Now, in fairness to um you know Timothy Gows, as many of his followers and supporters told me online, he's actually been at the forefront of using these systems, uh using AI systems in mathematics. Um it's not as uh much of a uh, you know, it's not quite as striking as as these two tweets might suggest. But in any event, what this points at again is the uncertainty. Just quite how difficult it is to say for certain what it is that these systems are going to be able to do and not going to be able to do. Yeah, that's another just example that popped up on my social media in the last week. But uh again, you know, you can read almost any newspaper report, you know, corporate publication, whatever it might be, and find similar examples. So I think the starting point in thinking about what more education means for us now is not to try and think really cleverly about the future and resolve the uncertainty and figure out exactly what skills and capabilities are going to be most valuable and important, but instead to take this uncertainty as the given, to not try and resolve it, but to say the most important future feature of the future of work that we currently face is this uncertainty. And the challenge for all of us is to prepare those who are wanting to enter the world of work or those who are in the world of work looking to move around, to prepare them to navigate this extraordinary uncertainty that we currently face, rather than attempting to resolve it. And I know that can feel quite that sounds quite constraining to just say, you know, the challenges that we need to respond to this future that we know very little about, but I think there is still a huge amount we can do, and that we're currently not doing if we take this uncertainty that we face as the problem. And I want to spend a bit of time thinking about what we can do in life. Of the uncertainty that we face. And the first dimension that I think we should be thinking about with respect to this uncertainty is the what. What skills and what capabilities. I think there's a huge amount to say here, but I just want to pull out two themes which I think are particularly important for saying for thinking about how we prepare people to respond to this uncertainty. One is getting back to basics, and two is teaching people to make critical use of AI. And I want to say a little on each of these. First, back to basics. One of my big worries about the powerful new AI systems that we have is that they are going to lead us to forget some of our most basic skills. Literacy, humoracy, critical thinking. It's an idea that people have begun to write about. Are we living in a golden age of stupidity? Uh wrote The Guardian at the end of last year. You can see similar pieces of writing and also academic reports elsewhere. This is not a story, though, that begins with AI in 2023. It's actually a far longer story. So these uh this is a chart here for um uh PISA. Um so let me explain what what this is. What one of the big challenges in comparing educational attainment around the world is that different countries have very different methods of assessment, um, and it's very difficult to compare them. And so what the OECD did, uh the Organization for Economic Cooperation and Development, what they do is at fairly regular intervals, um they test directly 15-year-olds around the world on basic skills. Mathematics, literacy, uh sorry, uh yeah, mathematics reading or literacy, and less basic in my view, but they also test for scientific knowledge. Uh and in the latest round published in in 2022, just to give you a flavor, there were 81 countries involved, 700,000 students took part. Um and this is the story that you see, which is that essentially not a lot happening uh to basic skills between 2000 and 2009, and then you get what looks to be quite a significant fall in certain basic skills. Um many people who are critical of social media point out that this decline begins in 2009 when some of the most uh addictive features of social media are introduced. Uh things like you know, like buttons on Facebook, infinite scrolling on you know uh X and Instagram. There are many people who point to that decline in basic skills among young people starting in 2009 as being um uh uh uh associated with the increasing use of social media. Um my worry with respect to AI is that it is going to encourage this pre-existing trend. Um that in a sense, if what social media has done, and I don't want to talk too much about social media today, but if what social media has done is it has distracted us, it's taken up too much of our attention, um, and it's led us to you know not pay enough attention to the you know learning these basic skills, my worry about AI is that it is going to lead us to just simply forget some of these basic skills altogether. That it's just too easy uh to turn to one of these AI systems for to help you with a problem that in the past you might have used a basic skill like mathematics or reading or or critical thinking, critical thinking to solve. The other thing to say that's really interesting, this is young people, but um the OECD also do a very similar exercise, the PIAAC, uh for adults, and you see a very similar downward trend in uh basic skills. So one reaction to this is why bother? Why do we actually care about the decline of basic skills? Uh, you know, if these systems can do things that require, would have required basic skills from us, but they can do it you know better, more efficiently, more effectively. Why bother? Why do we care? We can just use these systems instead. Um, think you know, just down the road, um uh these you will see these gates uh to tallow chandler's hall, um, home of the worshipful company of tallow chandlers. Tallow chandlers a few hundred years ago were people who had expertise, extraordinary expertise in the skill of tallow chandlering, which was the craft of hand-making candles. You know, we don't lament the loss of that skill thanks to the technologies of the Industrial Revolution. Um, why should we lament the loss of uh literacy, humoracy, and critical thinking today? You know, skills come and go, why why why be bothered by it? The reason I think we ought to be really bothered by it is because we can be confident that whatever the most valuable skills of the future turn out to be, and again, huge amount of uncertainty, but whatever they turn out to be, they are going to rely in some way on those basic skills of literacy, numeracy, and critical thinking. Indeed, one way, you know, when I talk about these skills as being basic, it's not because they're easy or simple, but it's because they provide the base, the foundation for all other skills that we need and we might put to use in our lives. This is a point that many experts have made uh about these skills, from you know, Nobel Prize-winning economists like James Heckman and Amartya Sen, esteemed social psychologists like Jean Piaget and Howard Gardner. Some of them call these core or essential skills rather than basic skills. Others talk about them acting as building blocks or gateways to other skills. But the point is the same. Um that these are skills which in a some in some sense provide the foundations, the building blocks, the basics for all the other skills that might be important and might uh you know be valuable uh in the future. And so, you know, we don't need to know anything really about what particular skills are going to be mo what more advanced skills are going to be most valuable and important, to know that having a mastery of those basics, literacy, numeracy, critical thinking, is going to be uh incredibly uh important. Um put another way, it seems to me that making sure the next generation is equipped with basic skills is what game theorists would call a no-regret strategy. It's a strategy that we aren't going to regret, however the world turns out, however the uncertainty that we face um resolves itself. So I think this has to be a core part of the what of education in response to technological change. The other part is making teaching people to make critical use of AI. I think the only thing we really know about the future is that it's likely to be saturated with technologies that are far more capable than those that exist today. And so teaching people how to make critical use of these technologies is is really important. And there's two parts to this. One part is, of course, how to use these technologies effectively. Uh I won't go into the details, but just to give you a sense of the ambition, I think we ought to be spending a third of our time in education, whether it's primary, secondary, university, professional institutions, teaching people in each of these different settings how to make effective use of these technologies. But it's not simply how to use them, it's how to make critical use of them. Um one of the things that is really distinctive about these technologies is that they are, although they are extraordinarily capable, they're also in many cases in many settings extraordinarily flawed. Um, just one uh example of this uh a gentleman called Damien uh Charlatan has been collecting examples of legal decisions where generative AI uh produced hallucinated content has featured. Um so he's found almost one and a half thousand cases where lawyers have used generative AI to inform their legal arguments, only to have those only for it to turn out that some of the cases those arguments have relied upon have been hallucinated, have been entirely made up. It's one of the kind of interesting uh but also uh uh irritating features of these systems is some is that sometimes they they just make things up and they do so extraordinarily confidently. Um we cannot treat these systems like we might treat a calculator. You know, when we use a calculator, we know that the answer it gives us is going to be uh correct. We cannot be sure of that. And that's why, in my view, teaching people to make critical use of these technologies, to not only understand how to use them, but also how to understand their limitations, their weaknesses, uh, their shortcomings is incredibly important. Um another way to think about this, Jeffrey Hinton, one of the so-called godfathers of artificial intelligence, he won the Nobel Prize in in physics for his work on neural networks. He called AI, these AI systems we have today, idiots savants. Um it's really important that we teach people to be able to tell when these systems are being savantic, when they're being extraordinarily capable, and when they're being idiotic. Um I suppose the big question though is how we do both. How do we teach people to use AI effectively without also undermining uh those basic skills that we think are really valuable and important? Uh, this is, I can't say too much more on it today for time, but this is a big theme of my uh my new book, which is out later this year, how we can do both of these things at the same time. The second thing I want to turn to though is the how. How are we training and educating people? Um, again, there's a huge amount I could say here. I just want to pick out two themes, two ways in which I think we ought to be rethinking how we educate people. Uh one, the idea of personalized learning, two, the idea of simulated learning. So, personalized learning. One of the really interesting results from educational research, and it was first captured in this paper in the early 1980s, is the extraordinary effectiveness of one-to-one tuition. So, an average student who receives one-to-one tuition will tend to outperform almost everyone, 98% of students in a traditional classroom setting. Um, it's known as the two sigma problem. That's what Benjamin Bloom, the original researcher, called it. Two sigma because an average student will tend to outperform two sigma to standard deviation, 98% of students in a traditional classroom setting. But it's a problem because, of course, um providing everyone with a human shooter is extraordinarily expensive. It's not something we can do at scale. Or at least it's not something we could do at scale until relatively recently. Uh, one of the developments that was very interesting when uh ChatGPT was released was that one of the first partnerships they did was with uh Salman Kahn and Kahn Academy. So uh for those of you who know, uh for those of you who don't know, Kahn Academy, an extraordinary collection of online practice problems and instructional videos. Really high quality resource. So I used to teach mathematics at Oxford, uh teaching economist maths, and I'd often direct my students towards incredibly high-quality resource. One of the first things that Salman Kahn did was he teamed up with um with uh Chat GPT with OpenAI to develop this. This is the logo for Khan Migo, which is a little chatbot that sits beside you as you navigate um Khan Academy, answering problems, concerns, you know, complications that you might have as you're working through all the different problems, providing precisely the kind of interaction that you might have with a human tutor, but doing it at a far lower cost. And the truth is, when I use the latest systems, what I see is a level of instruction that is more effective in explaining ideas, in asking questions of the student, in guiding them through problems, in tapping into their particular strengths and weaknesses, like a great human tutor would do, um, that is more effective, in my view, than many of the teachers I've watched in action over the years. And I include myself in that. Um one of the things I find particularly remarkable, just very quickly, is their breadth, just the extraordinary breadth of these systems in providing personalized learning experiences. So, you know, at one end of the spectrum, you have my five-year-old son asking at bedtime to try and delay bedtime. Daddy, where did the first human come from? Um, you know, I love it, it's it's not the sort of question you want just before they they close their eyes. Classic piece of bedtime delaying. You know, you put this to Chat GPT with the right prompt. I'm sitting here with my five-year-old, um, he's trying to avoid going to bed, but he's asked this question can we come up with something that that you know will answer it? Here was the little story that it told. I'll just read it quickly. Uh, a very, very long time ago, there were no people, only animals that looked a little bit like today's monkeys. One day a monkey mum and monkey dad had a baby who was almost the same as them, but tiny, tiny things about him were different. Maybe he could walk a little straighter or make a new kind of sound. That baby grew up, found a partner, and had babies of his own. And each time new babies were born, some got tiny, helpful changes, better brains for thinking, thumbs that grip tools, mouths that made clearer words. Imagine changing one Lego brick at a time. You can't point to the exact brick when it suddenly turns into a spaceship because the changes are so small. After millions and millions of baby steps, the animal at the end of the line wasn't a monkey anymore, it was a person. So the first human wasn't a single magic arrival, it was one baby in that long, long family who looked just human enough that if you met him today, you'd say, hey, that's a person like me. Uh now I love that. It just, you know, far better than I could do, exhausted at the end of the day, trying to get my my five-year-old to sleep. But then you go to the other end of the spectrum, and I thought about a fellow economist who told me that he'd be using GPT not only to solve the Ramsey growth model, which um for those of you who don't know uh you know, don't know uh uh the the problem here, this is a problem which would take you know a good first-year graduate student in economics a few hours to solve. Not only was the system able to solve it in a matter of seconds, but it was also able to provide very clear step-by-step instructions. And I I remember thinking about, you know, when I tried to solve this problem for the first time myself, how I had to wait for a week until I saw my tutor next uh to ask them particular questions I had about the problem. How I had to ration the time because we only had an hour, so I had to have a short list of questions that I could put to them. And just thinking that the system could do, you know, it was never going to get tired, it was never going to get bored, you could ask it as many questions, it's just an extraordinary, um extraordinary, uh an extraordinary tool. The other aspect of rethinking how we educate is simulated learning. And I'll just make this observation. It's a lesson from engineering, which is that if you think about a nuclear power plant manager, they don't learn how to keep a reactor core stable in an actual nuclear power plant. You know, they train in virtual reactors and full-scale replicas. An astronaut doesn't learn how to spacewalk while suspended above Earth in orbit. Instead, they use simulators, neutral buoyancy labs here on Earth. A Formula Run aerodynamics engineer tests their prototypes in a virtual wind tunnel, not with an actual human driver on a track. In so many areas of engineering, people master their roles in a simulated environment before moving on to do the tasks in the real world. We don't really do it anywhere else, and I think this is an extraordinary um shortcoming. Um I think the reason we don't do it anywhere else is because the technology wasn't quite there yet, um, but now it is. So you could think about how you might use these simulated environments in the you know white-collar professional setting. Think of a junior lawyer, for instance, who spends the first five to ten years of their career doing relatively routine work document assembly, document retrieval, document review. They could now take the leading role in a simulation of a leading dispute, handling AI-generated materials, interacting with AI generated clients, arguing their cases against AI-generated opposition in front of an AI-generated judge. You know, they could engage in the sorts of opportunities to learn their trade that would traditionally come far later in their career, if at all. The third aspect of more education, I think, is the when. At what moment in people's lives do we step in to provide them with education and training? So if you go to public school in America, the state will spend about $18,000 a year on your education from the age of five until 18. If you live in New York, that figure might be as high as $32,000. Um, if you go to any sort of college or university once you're done, you can expect a further $13,000 in financial support on average as an undergraduate every year. And $11,500 as a graduate, excluding any subsidized state loans. Education is one of the biggest investments the country will ever make in you. Nothing else compares. But once your childhood is over, basically that's it. You're out on your own. As an adult, you might receive, on average, $10 a year through state-funded programs if you're lucky to get anything at all. And the US isn't alone. Uh, around the world, countries, including this one, spend a small fortune educating people at the start of their lives, often several hundred thousand dollars, hundreds of thousands of pounds, but after that they spend almost nothing at all. So, one of the most ambitious programs in Singapore, for instance, I think comfortably the most ambitious program, Skills Future, it still only provides one-off credits worth about $3,000 when you're over 40. Um, this cannot be right, given the extraordinary uncertainty that we face. You know, the best response we have, or one of the best responses we have to uncertainty is flexibility, a willingness and a capacity to retrain and reskill later in life, with the same intensity and seriousness with which we currently engage in education at the start of our lives. And the fact that education is so lopsided that we invest so much at the start and so little uh later on, uh, it seems to me is uh ill-preparing people for the uncertainty uh that lies ahead. Um for those of you who are interested, uh what it's important to note that the challenge here, of course, is not simply how much we spend on adult education, but it's also um it's also the nature of adult education as well, which bluntly is a complete mess. And for those interested, do have a look in in the at this report, for instance, on uh adult education in in the United States, uh done by a team at Harvard. It's it's really uh quite sobering stuff, and and the the reference to this is in the the notes for this talk. So um, what does more education mean today? In my view, it means rethinking education through these three different dimensions. The what, what skills, and what capabilities are we going to provide people with to provide to prepare them for a future that is actually extraordinarily uncertain. I think there are two key things here. One, back to basics, to teach people to make critical use of the AI systems that are going to saturate their lives. Second, the how. You know, the way in which we train and educate people hasn't really changed for centuries, and now we have extraordinary technological possibilities. The ones that particularly excite me with respect to personalized learning and simulated learning, and then the when. Education is extraordinarily lopsided, huge investments are the start of people's lives, very little later on. That is not a good way to provide people with flexibility later in life to respond to uncertainty. But in the last few minutes, I just want to close with some thoughts on the limits to education. Why education is not the panacea that many politicians and policymakers currently treat it as. Just again remember that I think there are two different types of technological unemployment, two different ways that people might find themselves without work to do in years to come. One is that there simply aren't enough jobs to be done full stop, this idea of structural technological unemployment. Two, that there might be jobs, but for various reasons, people aren't able to do them. I think for now it's that latter type, that frictional type, that's our main challenge. As you will have seen in previous lectures, I think the reason. For frictional technological unemployment are threefold. One, that people don't have the right skills. Two, that they might not live in the particular place that work has been created. And three, that people might not have a sense of themselves, uh, might have, sorry, have a sense of themselves that is at odds with available work and they're willing to stay out of work in order to protect that identity. And for those of you interested in exploring some of these more, do take a look at um at previous lectures. The reason though this breakdown is important though is because if you think about it, education, more education, is only really a response to the skills mismatch. It's only a response to the challenge of people being unable to work because they lack the right skills. But what does education have to do with the fact that sometimes people might not live in the particular place that work has been created? And how does education engage with the idea that some people might be staying out of work, not because they lack the right skills, or not because they don't live in a place where there aren't jobs, but because the work that's available isn't the sort of work that they think people like them ought to be doing. I don't think education, as currently understood, is a good response to these two other types, other causes of frictional technological unemployment. The other reason I think there are limits to um education, not simply is it a partial response to frictional technological unemployment, but it's not a response at all to the structural technological unemployment. You know, what is the role of education in a world where there might not be enough well-paid work for people to do full stop? Again, I don't think this is the challenge we face now. But as I've argued in previous lectures, as we look further into the 21st century, this is a scenario I think we need to take seriously, and more education alone isn't an adequate response. So I think we need to look beyond education. Although education is our best response, uh, I don't think it is a complete response. So just to give you a flavor, let's suppose, for instance, that we take this structural technological unemployment seriously, that at some point in the 21st century we do find ourselves in a world where there just isn't enough work for people to do full stop. What are the challenges we might face in that world and how should we respond to them? I think the big economic challenges you will have seen in previous lectures is one of inequality. How do we share our income in society if our traditional way of doing so, paying people for the work that they do, is less effective than it has been in the past? Um it's not obvious how education helps us respond to that challenge. I think really the only response that we have to a labour market which fails to share our income sufficiently in society, is for the state to take a larger role in sharing out income in society. And I think the idea of a bigger state, a state that takes a larger role in sharing out prosperity in society, if our traditional labour market does an increasingly poor job of it. I know some people bristle when I talk about the idea of a big state. I think it's important to say that what I have in mind here is not the big state of the 20th century. It's not teams of smart people sitting in central government offices trying to command and control affairs from a distance. Um it's not a big state of production, it's a big state of distribution. It's a state that takes on a larger role in sharing out the proceeds of technological progress if the world of work does a less effective job of doing it than in the past. Another challenge, which you will have heard me talk about, is the challenge, the con the contributive challenge, the challenge of contribution. Today's social solidarity 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 but they're able to work, there's an expectation in most countries that people will actively look for work if they want support from the state. How do people make a contribution to society in a world where there isn't enough well-paid work for them to do? Again, it's not obvious to me how education helps us respond to this challenge. I think if we're going to respond to the challenge of contribution, how we can provide people with an opportunity to contribute to society and to be seen to contribute by others to society in a world where there isn't work for them to do, um, I think we have to explore very other, uh very different types of proposals. For instance, providing people with the opportunity to make non-economic contributions to society, to do other types of work which we think is socially valuable, socially meaningful and important, even if it's not paid work in the labour market. A third issue, the power of uh of the large technology companies who are responsible for developing these technologies in the first place. Again, this is a challenge I've I've spoken about in previous lectures. That in the 20th century, the main concern we had with large companies was their economic power. But in the 21st century, it seems to me our main concern with large technology companies is going to be far more with their political power and the impact they have on things like liberty, justice, democracy, and whether those things are under threat. Again, it's not obvious to me how education helps us respond to these challenges. Uh, and the tools that we have to respond to large concentrations of economic power, things like competition policy, antitrust law, do a very good job of helping us identify concentrations of economic power and for thinking through how we might intervene and break it up. But we've got almost nothing available to us as a toolkit for helping us identify uses and abuses of political power and making sense of how we uh intervene uh in those settings. And finally, there's the challenge of meaning and purpose. Um work, it's often said, is not simply a source of income, but it's also a source of direction and fulfillment. And if that's right, the challenge of a world with less work, that structural problem, isn't simply that people might not have an income, that the world of work might be hollowed out, but that sense of meaning and purpose and direct direction and fulfillment might be hollowed out as well. Again, it's not of, and so the question, you know, what provides people with a sense of meaning and fulfillment and purpose and structure in a world where work, paid work might no longer sit at the center of their lives. Again, it's not obvious to me that education, understood today as you know providing people with the skills and capabilities to flourish in the world of work, necessarily engages with this problem. Um, you know, we have a huge variety of labor market policies that help us think about how to prepare people for the working world, how to intervene and shape the labor market effectively. Do we need to start thinking about leisure policies, how we intervene to shape people's non-working lives instead? So I will just close on that note, which is again, I think more education is our best response to the challenges that we face in the world of work for the moment. I think we need to rethink what we teach, how we teach it, and when we teach it. But I don't think education alone uh is a response, is a sufficient response to the challenges that we face. It's a partial response at best to the challenge we currently face, which is this frictional problem, how to help people, uh, how to help people move around the world of work and take up the work that has to be done. But it doesn't seem to me to be a response at all to the sorts of structural problems that we might uh you know want to think about if we are looking further into the 21st century. Not problems where there is work, but for various reasons people can't do it, but one where there might not be enough work to be done full stop. So I will finish there. Um I do encourage you, if there are ideas and themes in this lecture that interest you, I've drawn on lots of uh ideas that I've set out in previous lectures. So do uh take a look at those and it will illuminate some of the ideas and arguments that I've set out this evening. But thank you very much indeed. Thank you.