Titans of Science: Mike Wooldridge
PodcastsThe Naked ScientistsTitans of Science: Mike WooldridgeFascination sparked in a little computer shop in rural Hereford. . . 03 March 2026Our Titans of Science series continues with Mike Wooldridge, Ashall. . .
PodcastsThe Naked ScientistsTitans of Science: Mike WooldridgeFascination sparked in a little computer shop in rural Hereford. . . 03 March 2026Our Titans of Science series continues with Mike Wooldridge, Ashall Professor of Foundations of Artificial Intelligence at the University of Oxford. He has conducted extensive work in the field of agentic AI, specifically systems composed of multiple interacting AI systems known as multi-agent systems. In this episode, he tells Chris Smith what drew him to computers and AI, the pioneering work of Geoff Hinton, why ChatGPT isn't made to speak the truth, and what's in store for us as AI continues to develop. . . Mike Wooldridge: from childhood dreams to the AI explosionMike Wooldridge, University of OxfordMike - Artificial intelligence means many different things to many different people, but for me, what it means is about getting computers to do things which at the moment only human beings can do. Getting computers to be able to understand ordinary language like we're having now, we don't think of that as being a remarkable capability. We're not surprised when somebody understands when we talk to them, but actually getting computers that could understand a conversation like this, which we're just at the point of doing, turned out to be phenomenally, phenomenally difficult. Chris - What got you into this in the first place?Mike - So I was of that generation that grew up with the Apollo space programme in the background. I can remember my parents waking me up to see one of the splashdowns from the Apollo spaceships, and that idea of space travel and technology was very, very much in the air. So I was always scientifically minded. I very much wanted to be an astronaut until I discovered just how difficult it was to be an astronaut. But a big moment for me was early 1980, and I grew up in the town of Hereford, a rural town. The main industry was making cider, and the main leisure activity was drinking it. In this rural town, I met with some friends on a Saturday afternoon, and they said, there's a shop that sells computers down the road. And this seemed completely outlandish to me, because in my head, computers were things that cost millions of pounds. The idea that you'd go into a shop in Hereford and buy one just seemed ridiculous. But we went down to this shop, it's called a Tandy store, and indeed, there in the window was a computer. And we went into the shop, and we chatted to the guys behind the counter, and they literally said, go ahead. They gave us some programmes to type in. We had no clue what we were doing at the time, and we typed them in. And then over the next couple of weeks and months, I went back to that shop, really a lot. They must have regretted being so generous to me, I'd suspect. But I was sat in the window of that shop, and I learned to programme. And the computer we used, a TRS-80 model one, was very, very crude. But it was the first time that an ordinary family could afford a home computer. Chris - Did you buy one, though? Did their investment in you, letting you sit there in the shop window, did you buy one?Mike - I would have loved a TRS-80. We couldn't afford one. I mean, they were 400 odd pounds at the time. That was actually well out of my pocket money range. But my parents, for a birthday, bought me a Sinclair ZX-80. We bought it secondhand, and it was 70 pounds, and I remember being so excited when I got it home. And much of the following year or so was wasted with me learning how to programme that and writing in computer games from magazines, typing them in, and running them, and debugging them, and so on. I remember it being a very, very happy time that you could, in your head, think up a programme and type it in and then get the machine to do something. You know, it would follow your instructions, and you would create something. And that was really, really great fun. But that set me then on the path for computing, and I knew that that's what I wanted to do. Chris - Did you know that you could do computing at university, though? And did you know that that was the logical step?Mike - Nobody in my family had a degree and only had the dimmest understanding, honestly, of what they were. Both of my parents had left school certainly by 16. But I think in my dad's case, he left at 14. I'm afraid despite my passion for computing, I wasn't a very diligent student at school or sixth form, so I didn't work terribly hard. But I did get a place on a computing degree, and that just confirmed for me that computing was what I wanted to do. You know, this was then the mid 80s, and this is when IBM PCs became a thing, and all of a sudden, businesses, ordinary businesses could start to think about having computers and how they might use them for word processing and spreadsheets. That's when all that stuff was invented, and so there was a big, one of the big expansions in computing at the time. Chris - This is obviously in the pre-Internet era. That was 10 years away from mainstream, wasn't it at that time? But it still existed. There were still networks because I remember my school at that time had, I mean, we used BBC microcomputers, and we had Econet and all that kind of thing, but we still had dial-up modems, and we would dial in to this network and share academic data among schools and universities and things. And there was a little club where we would make pages that would go up, and that was really a prelude. It's what lit my fire. I mean, were you doing that kind of thing as well?Mike - Yeah. So, my story is that I did what we now call an internship, but used to be back in the day called an industrial placement for a group called the Joint Network Team, JNT, who were based at Rutherford Appleton Laboratory, and they basically managed the UK branch of what became the internet. It wasn't called the internet really, at that time. I think it was still called ARPANET, and the UK branch was called JANET, Joint Academic Network, and not every university, I think at that point, in the UK was even connected to it, but virtually nothing outside universities, a couple of government establishments and military establishments were connected. But the Joint Network Team were busy thinking about the future of computer networks. And after that period with them, I absolutely understood that the future of computing was going to be networked, and this was not a big thing at the time, but I really, really understood this is where it's going. I mean, I can remember getting my first electronic mail message and being so excited about getting an electronic mail message sometime in, it would have been 87. But I say, the point is at the end of that period, I really, really got it that the future of computing was going to be networked. Chris - It's a bit like Bill Gates coming back from that conference at Microsoft and saying, it's all about the internet now. That was a sort of epiphany for you, albeit earlier. But how does that become an interest in AI?Mike - In my final year, I got to specialise, and I absolutely wanted to specialise in networks. But I also found AI really, really interesting, and in my head, these two things came together. And I realised, look, if the future of computing is going to be networks, then that must be the future of AI as well. It won't just be a chatbot that's having a conversation with a human being. It will be AI programmes that are talking to each other. I had that idea as an undergraduate, beginning of 1989 or thereabouts, and that's what I went and did my PhD in. But it was an uphill struggle to convince people that we were going to build these AI programmes that were called agents, that were going to communicate with one another. People were like, well, why? What's the point? And we're like, because it's inevitable. It's the future. The idea that you would have these AI programmes that operate on behalf of users, that would be, you would have your agent, Chris, and I would have my agent. We want to arrange a meeting. Why doesn't it just talk directly to your AI programme? Why don't the agents just communicate directly with one another? So we had this idea of AI programmes operating on behalf of their users and interacting with each other. And that became this very substantial research field called multi-agent systems, and what's now called agentic AI. Chris - I think you're pointing towards one of the big problems that the industry has, a sort of image problem in the sense that it seems to be omnipotent. AI can do everything. It's lots of different things, lots of different platforms, lots of different technologies. I'm going to ask you an almost impossible question, which is, can you try and break it down a bit for us so you can explain what these different elements of AI are and why they matter, each of them?Mike - Sure. When I was an undergraduate in the 80s, the big thing at the time was what's called symbolic AI. And the question is, suppose you want to build AI that could, I don't know, translate from French to English, how are you going to do that? Well, the big idea of symbolic AI is what you need to do is you need to model the cognitive processes that a human being doing that task would carry out. In other words, the kind of mental conversation that they would have with themselves, you would explicitly model that. It was called symbolic AI because the things that you were working with were more or less like words in a language. They were symbols. And symbolic AI was good for some things, for mathematics, but it was hopeless at anything that requires perception, vision, or understanding a spoken word, driving a car, riding a bicycle, anything that involves understanding the world around you. It was just really, really hopeless at, and it kind of ground to a halt, and by the late 80s, it was very rapidly going out of fashion. There weren't, I have to say, big ideas that replaced it at the time, but lurking in the background was an idea that had been around since the 1940s called neural networks. So remember, in symbolic AI, what you're trying to do is model the mind with neural networks. Instead, what you're trying to do is kind of model the brain. We can look at a brain under a microscope, and we can see what's going on there. And what we see is enormous numbers of nerve cells connected to one another in vast networks, and we know that all of human intelligence somehow reduces to the 90-odd billion neurones in a human brain, which are arranged in an enormous network. So the idea of neural networks is, can we do, in software, what those natural neural networks do in wetware, the stuff that's in our brains? Theoretically, the answer was known a long, long time ago. Yeah, in principle, you can, but people just didn't know how to really build these things. And one of the really interesting things is all of the key scientific ideas underneath the current generation of AI, which is all based on neural networks, were invented as far back, in fact, as the 1980s. Many of your listeners, and I know you have, will have heard of Geoff Hinton, who's often described as the godfather of AI. Geoff and his students and his colleague, and there was a large number of people working in this space, they invented all of the key techniques to make neural networks work. What they didn't have were computers that were powerful enough, and you need training data in order to be able to build these things. Chris - So that was the bottleneck then, it was the computing grunt that we just didn't have. We had the aspiration, we had the insight, we didn't have the technology to power it. Mike - Yeah, and my honest belief is if Geoff and his colleagues had had the computer power that they now have available, which is, I mean, it's hard to contemplate just how much computer power is used to build AI programmes at the moment. They're literally astronomical quantities of computer power. If Geoff and his colleagues had had that at their disposal in the 1980s, I really think the AI revolution that we've seen over the last 15 years or so would have happened then. But they didn't, and Geoff, I think he's pretty much on record as being still really quite bitter about this. Neural networks were unfashionable. And at the turn of this century, just 25 years ago, I had colleagues that were working in this space who would have their grant proposals rejected, and the reasons for rejection were that people would say, this neural network stuff's not going anywhere. It's pseudoscience. It's not proper science. We're not going to fund this. And that's how unpopular neural networks were. It's really remarkable to think about it now. Chris - Well, what's also remarkable, though, don't you think, Mike, is that you're saying that we've got computers that have extraordinary power and a massive power consumption to run them. And they're mimicking what's running in your head and my head, which is using about 20 watts. Mike - 20 watts. And what that demonstrates to us is that current machine learning algorithms are far too power-intensive, and they require far, far too much training data. A human being learns to read with much less training data than an AI programme requires. How many hours does a typical human driver spend on the road learning how to drive? 20 hours or something. Driverless car companies have spent thousands, almost certainly millions of hours on the road. And we don't have driverless cars yet. Chris - So what's wrong? Is that our model is wrong? Is it that the computers are just not powerful enough yet? Why do you think the problem is?Mike - No, I think it's fundamentally that the basic way that we train neural networks at the moment, which is a technique that Geoff pioneered in this space called backpropagation, is just not a very efficient algorithm. If somebody could invent an algorithm to training artificial neural networks, which required literally as little electrical energy as the energy that a human brain requires, that would be a completely transformational moment, but I don't believe anybody's seen anything yet that will do that. So that's how the space got transformed. I mean, and it started to happen around about 2005, and then it became supercharged in 2012, because people realised that if you could use GPUs, graphics processing units, for building neural networks. `you got 10 times as much bang for your buck as you could build neural networks that were 10 times bigger, or you could train them 10 times faster, but you got 10 times more neural network for your money, and that then supercharged the whole space. Chris - Can you explain for us, when you say that Geoff Hinton's big insight was this backpropagation approach, how does it actually work? How do these networks work, albeit in a slightly inefficient way at the moment, but what's the basic premise of their function?Mike - So the key point in a neural network is that you want to automatically configure the neural network, so that when you present it with an input, and the input maybe is a picture of Chris Smith, and the desired output, the name Chris Smith, you want to automatically adjust the network, so it's getting closer to the right output for the given input. Geoff and his colleagues came up with the backpropagation technique, which is basically based on calculus. It's not very advanced calculus. Leibniz would have understood the calculus that's required for backpropagation, it's called the chain rule, for those that know about calculus. But basically, it involves working backwards from the output of the network, and trying to automatically adjust the network going back to the inputs. Now, the mathematics is not very sophisticated, but there's a heck of a lot of it that you have to do. That was the challenge; it's just that, you know, the amount of computing that you have to do to be able to reconfigure a neural network using that backpropagation process. And you have to do this training process repeatedly, you can't just show a neural network one picture of Chris Smith labelled with the name, right, so you've got an input, a picture of Chris, you've got the output, the name of the person that appears there, you have to train the neural network repeatedly with many pictures of you with the same name over again, and actually you have to show it the same picture over and over again. But what you're doing throughout that process is every time you show it the input and the output, you're adjusting the network using backpropagation, so it's getting closer to the desired output for the given input. And that process repeats until what's called the loss, that is, the errors that it's making are getting sufficiently small that you're sufficiently happy with its capabilities. Mike Wooldridge: why AI confabulates, and looking aheadMike Wooldridge, University of OxfordChris - If we apply that to what we call large language models that work a bit like predictive tech, so I can go on to ChatGPT, and I can say summarise this and get a summary of a piece of work or something, what it does do from time to time is make stuff up. I mean, I've seen it for myself. We've put it to the test, we've asked it about various medical things to see what the rate of so-called confabulation is. Why does that happen? Why does it generate stuff that just doesn't exist? And what's the mechanism of that? And can we cure it?Mike - Well, I think the fundamental point, it's a really important point for your listeners, is that this technology is not designed to speak the truth. That's not what it's doing at all. What it's doing is exactly as you say, it's doing prompt completion. Given a piece of text, it's designed to produce the likeliest next word, and then the likeliest next word after that, and the likeliest next word after that, given its training data, and the training data for large language models is, roughly speaking, all of the digital text that's available in the world. You scrape the whole of the World Wide Web and every bit of digital text that you can get, which is all thrown at the neural networks so that it can learn, given this phrase, what the likeliest next word to appear is. And then you repeat that process, what's the likeliest next word after that, and the next word after that, and the next word after that. So you're appending each word that you generate to the previous ones. And that's literally, as you see the words appearing on a large language model, when you give it a prompt, that's literally what it's doing. So why does it make mistakes? It's not looking things up in a database of the truth and producing the output. It's not computing the right answer for you. That's not what it's doing either. What it's doing is just based on the training data and the prompt that you give, what's the most plausible next word to appear? So there's no reason that it should actually be speaking the truth. In the absence of having seen anything in the training data, intuitively, it will make its best guess about what should go there. So there's an example that you will relate to, I think, very much from my life when I was playing with GPT-3, which was really one of the breakthrough systems. It was the predecessor to ChatGPT, and I asked it about me, and it did a pretty good job of a few sentences coming up with known for AI and multi-agent systems. Great, this is what I made my career. And then it said, studied an undergraduate degree at Cambridge, a marvellous institution, which I never went anywhere near. I never studied there, never had a job there, was never a student there, never a postdoc or anything. Why is it doing that? Well, intuitively, my guess is, for an Oxford professor, studying at Oxford or Cambridge is very, very common. There will have been lots of biographies in the training data where it will have seen that. And so in the absence of any more detailed information, it just fills in the gap in a plausible way. Another way of thinking about it is that what these models are doing is compressing large amounts of data. So they've seen all of the digital data in the World Wide Web, and somehow that's being compressed into the neural network. But that compression process can't be perfect, because you can't store all of the digital data in the world in a neural network with a couple of billion parameters. So it's lossy. And so what we're seeing, another theory is that what we're seeing there is just a lossy output. It just didn't manage to store that information. But I say, forget about that. The simplest way to think about this, honestly, is it's not designed to tell you the truth. It's to tell you the most plausible thing, given its training data. Sometimes that will coincide with the truth, but very often it won't. Chris - Well, shall we have a look at how good it is now? Because you put it to the test back then. We've asked ChatGPT to write the introduction to a Titan of Science episode of the Naked Scientist. I'll read it to you. And if you spot anything along the way that makes you generate antibodies, I want you to say, that's not right. So here we go. Mike - It depends on how flattering they are. Chris - Today's guest is Professor Michael John Waldridge, a British computer scientist and one of the world's leading voices on artificial intelligence. Well, that's not true, is it? They've got that wrong. Mike - No. Chris - Born on the 26th of August 1966 in Wakefield, England, Mike studied for a BSc at Wolverhampton Polytechnic. Is that correct?Mike - Yeah. Chris - No longer Cambridge. He didn't go there anymore. Before earning his PhD in computer science from the University of Manchester Institute of Science and Technology in 1991. Mike - 1992. Chris - Okay. So we've one minor glitch there. Mike's academic career has taken him from lecturing posts in Manchester and London to a full professorship at the University of Liverpool, and since 2012, he's been at the University of Oxford, where he's the Ashall Professor of the Foundations of AI and a senior research fellow at Hertford College. His work helped found the field. Mike - Okay, so far. Chris - We're doing okay?Mike - Yeah. Chris - His work helped to found the field of multi-agent systems - well, you told us that - AI systems composed of interacting autonomous agents, and he's published hundreds of research articles and several influential books on AI. Doesn't mention he gave the Christmas lectures there. But is that right? Other than that?Mike - Yeah, yeah. Chris - It goes on to say Mike's contributions have been recognised with many honours, including election as a fellow of the Association for Computing Machinery, ACM, the Association for the Advancement of AI, that's triple AI, the British Computer Society's Lovelace Medal, the Royal Institute, oh, it does mention here, the Royal Institution Christmas Lectures, and the Royal Society's Michael Faraday Prize in 2025, among others. Off the academic track, you're passionate about communicating science to the public, where you're here, so you must be doing that. From writing accessible books on AI to giving public lectures and media interviews. And he balances his professional life with family time with his spouse and children. So that's an improvement by the look of it. You've nodded along to most of that. Mike - Yeah, it's interesting. I think it very much depends. I mean, one other thing about these models is that they use randomness when they generate their answers. So the fact that you get an answer like that today doesn't necessarily mean that even with the same model, that you would necessarily get the same answer. I mean, going back to the Cambridge example, which is completely true, and I was giving a lecture on this a couple of weeks ago and mentioned this, and somebody opened up ChatGPT and said, it still claims that you went to Cambridge, and now it's got anecdotes about you, and we went and had a look. And if I hadn't been a lazy teenager, if I'd studied a bit more and had gone to Cambridge, then the dates that it had would have been exactly the right dates. That was a very strange thing. But, you know, it was like he rode for this college or whatever. It's all complete nonsense, but very, very plausible nonsense. And that is definitely something I think that the listeners need to be aware of. Not only does it get things wrong, it gets things wrong in very plausible ways. But honestly, Chris, if I studied at Cambridge, I deserve a pay rise. Honestly, I'm much better qualified. Chris - It occurred to me that perhaps because your profile's gone up because of the amount of exposure you and your work have had because of AI and the burgeoning interest in the field, perhaps there are now more templates, there's more training data on the internet for it to ingest. And so that has become a more plausible description of you. I say that because I was just reading a newsletter from a tech journalist in America, and he opens by saying Stanford and Yale researchers extracted 95. 8% of a copyrighted novel word-for-word from Claude 3. 7 Sonnet, and Gemini 2. 5 Pro gave up 76. 8% of Harry Potter. So basically, they're saying that in some cases that it's just memory retrieval. So do you think now that the reason we're getting so much accuracy is because it's got a lot to go on you? If we asked it about me, it might still say that Patrick Moore started The Naked Scientists and trained me, which is what it used to say. Mike - Yeah, that's a good point. I think the technology is getting better, and it's got markedly better just in the few years, although far, far, far from perfect. About me, that could well be the case. I don't know if that's true. But yeah, doing the Christmas lectures, I guess they've probably boosted the amount of data that was online available about me. And is it just regurgitating? That's a really good point. The point about, for example, Harry Potter novels, I mean, the issue there is that clearly copyrighted data has been used to train these models, and I had a couple of letters from lawyers in the US recently about a class action lawsuit, where it was found that my book was used to train AI models from a very large AI provider, and I have the opportunity now to enter a claim. And clearly, I say, if you can reproduce 70% or whatever it was of the Harry Potter novels. Chris - 95. 8% in one case. Mike - Wow. Chris - And it's interesting, because he goes on to say, Google told the US Copyright Office in 2023, there is no copy of the training data, whether text, images or other formats present in the model itself. So it's kind of hard to reconcile that argument with the fact that you can, with the right prompts, get back three-quarters of Harry Potter word-for-word and 95. 8% of another novel. There must be in these models, there must be a nearly word perfect impression to get that to work. Mike - Yeah, but the point is, you can't point anywhere in the neural networks and say where it is. You could say, look, here's my neural network. It's just a list of numbers. It's not like it's in a database. If it's in a database or it's on a web page, you can point out and say, look, there it is. But with a neural network, you absolutely can't. And what this points up is that copyright law was not invented with large language models in mind, and in particular, they are not storing the text that they've been trained on in any conventional way that we can understand, and they are not derivative works in any conventional way. And one of the arguments that was made in these lawsuits was that actually all the AI is doing is reading a book in the same way that you or I would read a book and then takes inspiration from it. So copyright law needs to be fairly nimble to try to catch up with what's going on here. Chris - When we hear that the operators of these networks put in what they call guardrails, so that people can't use them for nefarious purposes, how does that actually work? Because if you've got this layer of neural connections, effectively storing information, and you have an input and it generates an output, how do you make it so that you actually constrain that? Does that have to be a physical thing that there's something between the output and the human being, and it just interrupts stuff and says, no, you're not having that?Mike - What's going on now is behind closed doors. So we don't have detailed access to how this works, but it looks like there's a range of different mechanisms. One thing people are very much trying to do is once they have their model, they try to basically train it good manners. They do this by what's called reinforcement learning with human feedback, where basically humans give feedback on the answers that we're given, and the model is adjusted to reflect those answers. So if a model comes back with an inappropriate answer, like a recipe for building an improvised explosive device or planning a terrorist attack, a human judges that that's a bad answer, and that feedback is given to the machine. So that's one big thing to try, a big way of trying to improve the models, just so that they're better at not producing inappropriate content. But ordinarily, I suspect what's happening is when you give a prompt to a machine, there will be some examination of that prompt before it goes off, and partly that's likely to involve keywords like explosive or murder or something like that, which will trigger a kind of a guardrail response. I'm sorry, I can't answer that question. But probably much more sophisticated than keywords, actually, you'll have AI that's looking at your input. But then also, there'll be scanning of the outputs as well in exactly the same way, you know, because it's possible that you could just inadvertently prompt one of these models, which then produces inappropriate output. So, in other words, your prompt looks completely innocuous, but you end up with something extremely unpleasant coming out of the other end. One of the most interesting recent developments I saw is trying to look for the patterns of bad behaviour within the neural networks themselves. When a network is doing something which is evil, and spotting those, and then being able to automatically adjust the network to suppress the evilness, if you like. I'm struggling to find metaphors for this, but for me, it's like you're kind of almost like you're drugging a network the same way that a human being that was violent might receive some treatment. You're drugging the neural network to suppress those negative activations. Fascinating stuff. Chris - Yeah, it is. Let's finish by asking you to crystal ball gaze then. If we get together in five years' time, what do you think we're likely to see?Mike - Well, I think, firstly, the generation that are growing up with this technology, like all technologies, the people that grow up with it think about using it in ways that its inventors never, never, never imagined. And for us dinosaurs, let's be honest, Chris, we are going to be bewildered by the things that kids are going to be doing with this technology. Students that we're taking at university now have actually had ChatGPT available to them throughout their A-level studies, right into their degree studies. In a few years' time, we'll have kids that don't remember a world before ChatGPT and the like. And the way that they're going to think about it is just going to be transformative because this always happens with technology. So that's the first thing. We are going to see the world that we're heading into very, very quickly, for better or worse, a world where TikTok-length videos can be generated to order. You just think of what you want the video about, and you'll get a TikTok-length video about it. TikTok, one of the most popular social media platforms in the world, and that's just going to be the beginning. We're heading to that world very, very quickly. What's beyond that, you know, when AI hits virtual reality and so on, is going to be remarkable. But will it fundamentally change humanity? Look, I live in a Victorian terraced house built in the late Victorian era, and it's drafty, and it's cold, and I go home to cook a dinner and light a fire in ways that, you know, 100 years ago, 125 years ago, people would absolutely have recognised. Walk into a classroom today, and you'll have a teacher at the front of the classroom, and the kids might be a bit scruffier, but a Victorian teacher would have recognised that. So, despite all this technology that will go around us, there are so many fundamentals of humanity that have remained constant. But at the same time, it’s going to shake things up in many, many different ways!