258. How AI changes everything we should teach kids

That's mad. Clinically insane. Completely mad. If AI makes it harder for, let's say, us to Test people cannot take away the absolute best tools for learning that maybe humanity has ever invented. On the one hand, there will be people displaced. On the other hand, if we artificially limit our companies from making people redundant, then they'll become uncompetitive.

And there are those people will lose their jobs anyway. Threading this needle of what's the right level of displacement and how do we make the best of that is actually a really hard problem as well. There are political risks. There are There's the concentration of power. To our mind, the best mitigation that we can think of for those risks is the

Hello and welcome to the Rest is Money with me, Robert Peston. And with me Steph McGovern and today Uh we've got a fantastic guest for you, Mark Warner. He is a founder of a fantastic London-based AI company that he's in the middle of selling to Accenture for a billion dollars, making it the UK's newest.

Tech Unicorn Robert. Is AI gonna take our jobs? Why he is sold out to the world's biggest consultancy, Accenture, and how he persuaded The Prime Minister at the time, Boris Johnson, to lock down during COVID. Mark it's an absolute treat.

to have you here on the show um robert and i are both fascinated by your business well done and all the success with it but can you just tell us a bit about what faculty is you know because this is a business that started with you and two friends, two colleagues and has become, well, the UK's latest t unicorn, tech unicorn, which is an incredible achievement. Yeah. Well, thank you. Lovely to be here. Um so faculty Is an applied AI company. So sort of 2014 when we set up the business.

we could see already that AI was gonna become the most important technology of our time. Like I was actually a physicist by background and sort of decided physics was a science of the twentieth century. led to these kind of incredible technological revolutions. um that we saw over the course of the twentieth century. But AI was going to be the equivalent.

of that for the twenty-first century. But we wanted th at the time there were lots there were OpenAI and DeepMind and these were research labs. So they were primarily concerned with how do you push forward the state of the art of the technology, how do you push forward the scientific frontiers?

And we thought there was a place for bridging the frontier technology and getting it into the real world. So we thought it was like it would be a great thing to be uh to make it valuable in people's lives. And so that involved, you know, doing things in health, in education, helping companies make their products better and cheaper. And so we've been doing essentially that for the last ten years.

uh watching as the world has recognised more and more that this is gonna be like a huge deal um across the across the economy. And can I just ask in terms of that initial judgment that you made about AI. Your academic background, um uh your research was in quantum. So and and quantum is obviously at the moment uh something we're all pretty excited by. Why did you make the leak from quantum to AI? Uh a couple of reasons.

One, AI is like fundamentally easier. Like as just as a as a kind of mathematical topic, um it's simpler. And I could see that it was just seemed more important. So, you know, quantum mechanics is a hundred year old field. It's very mature. If you want to make important breakthroughs, you have to be unbelievably talented these days. Whereas AI was a much less mature field, or at least let's say The AI techniques that were working, deep learning, was a much less mature field

And so there just felt like there was way more opportunity. There's a a kind of um funny quote from one of the pioneers of quantum mechanics. He said, In those early days It was a time when second rate minds could do first rate work. And in my life I'd read that in my quantum days and always thought, Well, if there was ever the opportunity to do that, that's where I should be. And so I've kinda like when I saw the AI uh like taking off.

that just felt like so much more um energetic and and young as a field. And just and and this is not to denigrate you in any way'cause you're obviously a lot brainier than I am, particularly in this space. But you broadly took the view that your you were much more likely to make a breakthrough in terms of commercial application of AI rather than sort of fundamental research that was gonna push for

forward the frontiers of knowledge. Actually it was less about where I would like to make a breakthrough and what I thought was good for the world. So, you know, I'm always part of the reason I got interested in this was because I cared a lot about safety. And so I've always taken the view that pushing forward the the bounds of what these models are capable of

Is co at least very complicated from an ethical perspective. I think there's very good arguments to do it, but I think there are arguments against doing it. Whereas if you take technology that already exists and you just apply that, that seems to me like almost obviously good. And so it was more that choice that pushed me in that direction rather than where I actually thought I could make most impact.

We're proud to say that the rest is money is powered by Octopus Energy this year. Greg Jackson is back to answer another question. Now this is something that we see a lot. That I wanted to ask you about. When you're starting something new, how do you think about risk and momentum? I think people think entrepreneurs love risk.

But I'm I'm not sure they do. I hate it. I never buy individual stocks and shares on the stock market. Uh I uh don't gamble. I think the thing about an entrepreneur for me is I've got more control. Uh when you're working for a company every day, you're at risk of what that company chooses to do with you.

If you're an entrepreneur, actually you've got far more say in what happens tomorrow and next year than when you're, you know, at the mercy of your bosses. Nice one, Greg. Well, thanks to Octobus Energy for powering this episode of The Rest is Money.

So how did you work then? Where to start in all of this? Um well we started with the place that we knew. So we were PhD students, or well actually we'd got our PhDs at this point. but we knew the academic world really well and could see that transitioning from a very mathematical subject, something like physics, maths or engineering into the world of data science as it was called at that time and now AI.

was quite tricky because you needed experience to get a job and you needed a job to get experience and it was just sort of hard to break that kind of loop from inside the academic system. So we started as a program to help PhDs become uh AI engineers, AI researchers, data scientists. Then pretty quickly we got applications from about ten percent of the UK's maths, physics and engineering PhDs.

And so we realised that we just had these incredibly talented people throw it flowing through the programme and we could just hire them ourselves and start doing um building AI systems. Then we realised that or or we had actually the resources more, it's fairer to say, to start doing a work on the safety side and um trying to figure out how to make these uh algorithms do what we wanted them to do. um there's still a big disconnect between AI and decision makers.

and the different departments inside businesses. And so there's sort of these silos where marketing only really talks to marketing and procurement only talks to procurement. But if you really want to make intelligent decisions, you want to connect all that up and you want to get that into the hands of decision makers. And so then we built a software product to solve that problem. Just explain to uh How

You uh y y your work is about applying AI as opposed to creating these large language models. And I suppose w one question is why did you decide not to go down the route? of building, you know, your own equivalent of, you know, whether it's um open AI or Gemini or uh

you know, deep seat. Why didn't you go down that route? Well I mean the first thing to say is it's really hard, right? If you those companies that are building those models are essentially the y mae'n ymwneud â phobl sy'n ymwneud â phobl sy'n ymwneud â phobl sy'n ymwneud â phobl sy'n ymwneud â phobl sy'n ymwneud â phobl sy'n ymwneud â phobl

And so even if we'd tried, I'm not sure we would have succeeded. And we kind of see that in the world today where like the the big three are taking off and seem to be ex slightly accelerating away from some of the smaller players. um uh in the market. But ultimately, uh, we've always had this hang-up about really trying to push forward the bounds of the technology where it's just not unambiguously good to us.

And so our kind of the the choice we We've never really had the energy to throw behind it that it would take to succeed in doing that. because we just can't quite bring ourselves to think that it's really, really the best thing to be doing. So tell us what you do do in the AI space.

So the way we think about it these days is we have this incredible wave of technology washing over us. And all organizations are gonna have to change some of their core business processes. They have to discover drugs put them through clinical trials and then market them to the world. And each of those elements is going to change very substantially over the next five to ten years. Let's talk about clinical trials, for instance. So

There's an amount of time that you have to test something for, like test a drug for, to ensure it's safe. So, you know, if you want to know whether it's safe over a few years, you have to test it for a few years. no one should ever touch that bound of uh of like safety. But actually in a real trial, there's lots of extra time, administrative burden before, after, and all of that is very costly, but also stops us getting drugs that really do work to patients as fast as possible.

And so Um AI is already letting us plan trials more effectively. uh make them like happen, you know, much closer to this truly safe amount of time. And then that ultimately results in cheaper drugs getting to market faster and you know better like healthier patients um out in the world. Why is it a case that it can't also do the trial faster? Why

Is it that you have to still keep that in the same time, but it's just the abdomen that can be better? Ultimately, it's likely that we'll be able to build sufficiently clever models of the human body to start testing things. Uh like they call it in silico. Isn't that what Denis Azabis thinks will happen? I th I mean on some time frame that will certainly happen. It's just a question of is the technology right now

uh sufficiently good that it should be trusted to fulfill that kind of um that kind of role. And it's just not there yet. Like no one would tell you that there's any chance we are um able to model like anything like the complexity of a human body with sufficient uh like quality that you'd want to trust it to say this is safe or this is not.

So for the example you've given there is the productivity and efficiency gains around the admin side. But you you guys have also done some really fascinating stuff with the NHS in terms of like the early warning system in COVID and

Can you explain that?'Cause it it it it's quite fascinating what you did there. We started working with the NHS many, many years ago, um helping them in all all kinds of small projects and eventually um started helping them with their AI strategy or at least a small bit of the NHS called NHS X at the time. And um, you know, we started that project in sort of late January twenty twenty.

uh where we were all planning to sort of multiple years build up this uh AI lab quite slowly and um and focus on rolling out capabilities across the system. Obviously history didn't pan out like that and sort of late by late February Um, we went to them and said, Look, we can all clearly see that this uh Covid is gonna become a an enormous deal. Would you like us to help with that?

And so actually the sort of senior technical leadership of the of faculty kind of moved so me, the CTO, the chief AI um scientist all moved out of faculty for I forget how long, let's say three months, and went to work full time for the NHS to help them uh figure out how to predict how many patients were gonna turn up in hospitals across the country.

Because of course you'll remember at the start of the pandemic particular, they had to make all kinds of difficult decisions about where to send things like oxygen and ventilators and how to move patients around and where to send PPE. And if you don't know where the disease is gonna peak, you're sort of doing that blind and you can move stuff and then it can turn out it's easy in other places. And so We built a system for them called the early warning system.

that sucked in a bunch of data, uh anonymous data, as in there was no personally interfa identifiable information, and we built models to predict how many patients were going to turn up um uh across the entire uh the entire NHS. And of course that gave us very deep insight into what was going on across the system. And we were able to help um like some of the senior people, including the Prime Minister, to understand

like what was likely to happen and off the back of that he was able to make a bunch of decisions about things like lockdown. I mean what Dominic Cummings said to me pretty much immediately after lockdown, as if it hadn't been for the work that you and your brother did, um, the Prime Minister might not have knock locked down. I is is is that how it felt at the time?

I think uh he would have locked down and the like the later lockdowns show that he would have done it anyway, but I think it would have been a couple of weeks later. And that couple of weeks could easily have overwhelmed the NHS. And, you know, as a result of doing it two weeks earlier, of course, significant numbers of lives were saved. So that was that that does that did that feel we you know, like the most meaningful moment of your life in a way? Not really. I mean it felt uh panic

uncertain. Uh you know, it was not the it was not the suave movie version that I would like uh if anyone ever dramatizes it. So it was not very cool. But um yeah we we kind of You know, it was it was m it it was all much more real than that. Like we were just trying to do what the sort of our uncertain best rather than thinking. Every minister, every official and it just the chaos

The what experience we're talking to people. So being on the inside must have been must must have done your head in a bit. That chaos was uh was very noticeable from the inside as well. It feels like a real sliding doors moment as well, given you just happened to have started this project with them and then I mean, how easy was it to like kind of suddenly scale up and move so quickly to do it or'cause as you say three of you went to then work for the NHS, was it

How easy was it to suddenly do that? We put a team of I don't know, I think it was probably like started off at about fifteen and went up to about forty. Um yeah, I mean I it it was I don't know, uh y you guys probably r recognise it as well. It just felt different at that time. Like it w it didn't feel hard. We phoned our investors and they were like, Obviously that's the thing you should do.

Uh and like, you know, we knew there were gonna be repercussions downstream, but it was it just felt like one of those moments where uh it was just kind of in some sense the emergency was so clarifying that it didn't you know, it didn't feel hard or difficult. It was just like obviously we just have to do this. I mean I wish actually I wish we could recapture that level of clarity, uh, particularly in in sort of government work, um, uh, more regular.

we know that the people in government can operate on those kind of timescales and achieve those, you know, those kind of like that system was absolutely the best in the world when it was built. uh over the course of about six months. And and so I mean obviously incrementally with useful things almost immediately, but but And this is the system that forecast essentially Um bed demand, uh NHS resource demand. Exactly. Exactly.

What what's happened, is that system still in operation? No, that's one of the slight tragedies um of the of COVID was that obviously we'd spent and not we but the government had spent an enormous we as the country had spent an enormous amount of money on COVID. And when we needed to rein that spending into under control, it was done fairly indiscriminately.

And so You know, I would very, very much like it if we had kept something like the early warning system going, because it is clearly the future of public health. Totally. I mean it seems absolutely astonishing that it's been I mean, you know, I mean th there's two things. One is You know, there will be another pandemic. Correct. Um and it would be y absolutely essential to have that kind of predictive model uh within the NHS. But w w w why wasn't this system simply useful?

for um assessing NHS demand On a normal day to day basis. I mean it seems to me if it can do a crisis, it can also do normal times. So wh why is it gone? Well the th I mean, the future of public health, like we should have Uh, on the TV every day, you should get a weather forecast, a pollen forecast, and an infectious disease forecast.

that should actually like m let you make good decisions about what you want to do with respect to your health. And that should be constantly being done with this wastewater surveillance. So you you monitor sewage for the genetics of all the infectious diseases. you predict how much is gonna be in every region across the country, in every city, and then you provide that information to the NHS, to companies, to

Um to citizens. I mean relative to outcomes, that's not all you know, the the cost is i insignificant. So why isn't it happening? Um, we have tried very, very hard in lots of different directions to try and get that to happen again. We are now, as I understand it, somebody told me, we are now behind Malawi in our ability to to uh this wastewater surveillance and uh early warning.

It's such to me it is such a no-brainer win. It is ultimately a piece of generational infrastructure like the motorways or the Met Office that we should be building now for our kids, for our grandkids.

Um and at the moment we're not. It's really depressing. Yeah, it is. Can I just ask therefore on facul faculty AI more broadly, widely seen um as an important British success story. But you did take a decision uh a few months ago and the deal's going through now.

to sell the company to uh I don't know if it's the world's biggest consultancy. It's certainly an enormous consultancy. It's the world's biggest. I mean it employs, you know, almost eight hundred thousand people across the world, Accenture. Um I mean, it's been reported you sold for, I don't know, a billion dollars plus, which is why.

Steph used the unicorn tag about you. But why did this feel the moment for you to sell? Did you feel constrained in your ability to grow by your existing independence? So I think there's a few different elements that sort of figured into the decision. So the first and probably the most important is that we do think of AI in terms of this tremendous upside. Like it's gonna change lots and lots of things, there's gonna be

um a personal tutor for every kid, a doctor for everybody uh in their pocket. These kind of amazing, wondrous technologies. But there are also risks associated with it. So there'll be unintended consequences of what good actors do, and there'll be malicious use by bad actors. We think of those as like the technical risks.

And then there are political risks. There are there's the concentration of power um that will come with uh if we don't manage things correctly and there's the risks of conflict as countries get increasingly concerned about others having access to this powerful technology. Rydyn ni'n mynd i'n mynd i'n mynd i'n mynd i'n mynd i'n mynd i'n mynd i'n mynd i'n mynd Um and it to our mind the best mitigation that we can think of for those risks

Is the safe widespread adoption of AI. So the safety, building the kind of technical safeguards into the technology itself. And then having it wides widely adopted means you don't get the same concentration of power and the same risk of conflict. Right. Now Ultimately, we think this is happening right now. Like, if you have not used the latest coding tools, You should just sit down and try them. They are extraordinary. Anyone can now basically program in English. And so

Um you know, autonomous cars are here. They work. They're gonna come to London this year. Things like this. And so with the best will in the world and let's you know, I have hold faculty in extremely high regard, but nevertheless, if we did brilliantly for another five, ten years, we'd be a m multi billion dollar company. And we wouldn't be able to touch the sides of c safe widespread adoption at a global scale. And that and and so why does Accenture as a partner

help you do that? Because they are the biggest consultancy in the world. So immediately we have access to essentially every organization in the world, whether it's the labs themselves, whether it's uh, you know, country like, you know, however many countries they're in across the world. essentially every important decision maker in essentially every large company in government, we can now help think through this problem and help build this technology out in it like safe and widely.

And and you're gonna be chief technology officer, what does that mean in practice? Um, it means that uh we uh I help set the strategy, the technology vision and the technology strategy for the organization. which sort of leans everything in an AI first direction. I mean, uh you know, the the chief technology officer role has evolved a lot over the last few decades. It used to be about sort of building computers in basements. then it became about building software products.

But I think with the latest iteration of these coding tools, it's gonna become primarily an AI role and primarily about how does AI actually drive like the fundamental elements of your business. And on your point about the ethics around this, which clearly mean

so much to you. You you advocate, don't you, human led AI over full automation. So do so why do you think preserving, I guess, human judgment is non negotiable? I I was also interested in, I saw that Um Accenture uh is basically as I understand it, training pretty much all its people to use AI agents. What does that mean in practice?

Right. So the first thing to say is the deal hasn't actually closed yet. Right. So I haven't uh we haven't started with them. But I think that is v like a broadly a very sensible step for every organisation. Um they uh like if you're not If you haven't experienced these kind of latest coding agents

They're like superpowers. You know, anything you've ever dreamt of in so I've always wanted to build this particular type of planning software for myself so that I can just quickly build out what's called a Pert Diagram, basically a gancha. And just the last night and this morning I just started building that. And I just built it in English. And like that power to put software in every like the ability to build software in everybody's hands is remarkable. And so I think everybody should be just

getting their people started in this. And for some roles it'll make more sense than others, but there's like an activation barrier where if you don't if you don't play with these tools, you never quite get started. You don't know what they're capable of. And so you never get the benefits. So I think it's very sensible to just be able to do that. lean everybody into it to start with and then kinda let the dust settle and let the people who really

like get value out of it, continue. You see, the businesses doing well are the ones that no matter what their business, whether they think they're digital or not, is the they're encouraging their their staff to to to just find ways of you using AI. And even on a personal level, my mum and dad were staying with me last night'cause they were looking after my little girl. And they were saying to me, What's all this about AI? We we won't need it, will we? And I said,

Right, ma'am, what's your what's going on with you in your world? What's the biggest? You said, Well, we'd love to get more people at Stockton Rambling Club. And I said, Right, ask AI. to tell you how you can do that. And she went, what? So then we did it. And honestly now she's downloaded the app. She's like, right, I'm gonna go to the committee meeting. I'm gonna tell them all what AI has told us to do to get more people in Stockton Rambling Club. I think that's

Perfect example. Yeah, just using it, finding ways to use it in your life then just makes you feel more comfortable about it.

Yeah, I think that's exactly right. So so just to my question before, why is it you think that human judgment is still needed even though the models are advancing? So I guess right now it's still true that the models um can't aren't powerful enough to do things without human judgment. So we're sort of in this um intermediate period. where I had to spend a lot of time saying, No, no, don't

put the diagram like that, do it like this, this is better than that, add this feature, these kinds of things. And even though it can do the coding, it can't um pull together the the sort of bigger picture for me. So I think right now it's just necessary from a technical perspective. in the long run, um, we have to find ways to stay in control of the models. So But that's what I wanted to ask you, um, because, you know, there are Plenty of people. Uh expert.

uh people actually at the cutting edge of creating models. who um are either warning um that we're quite close to seeing very significant numbers of human jobs replaced by uh AI, um or actually trying to do that. You know, they're actually in the business of trying to do that. Um Uh I mean that is gonna happen, isn't it? There I mean, you know, we are going to see jobs eliminated by AI, aren't we? There's no question, but I think it depends on exactly how you see that.

what context you set that in as to how serious you consider that to be. So So what do you think? I think of I mean AI is ultimately better software. And so I see this as part of a trajectory that we've been on sort of since the invention of the computer where software has consistently replaced jobs. Um and so, you know, the number of secretaries over time has uh has diminished and all kinds of things like this. But you don't think this is the biggest leap, the biggest step change? Um so I it

I don't think it's gonna feel like a single moment in time. Actually, my and this is now I wouldn't say this was like general wisdom as in other people, very serious people, have a different perspective on this. Well Dario Amade definitely has a different perspective. It's slightly hard to decode exactly what um exactly the details of what he's saying. He certainly thinks he'll say things like we're gonna get a country of geniuses in a data centre in two, two, three years.

Um but what exactly he means by geniuses is like slightly hard to unpick. And so I I ca I don't think I can make strong statements about like what he means.

what I can say is I think this process is gonna sort of be much more continuous. So these algorithms are going to get better and better and better. And every so often they're gonna sort of um uh get to capability levels that pop through into the mainstream and suddenly become like the Chat GPT moment as the kind of most obvious example of this where

those models were improving for quite a long time beforehand and if you were paying attention you could see that they were just getting better and better and better at natural language processing. The chat GPT moment was primarily when it became like a mainstream phenomenon. And so that feels very like a step shape, but actually there's a more continuous background going on.

This episode is brought to you by Hargreaves Lansdown. Now, if the last four decades have shown us anything, it's that change, Kate. Coming. There have been so many huge economic events that have shaped our thinking today. Back in the 80s, we had Margaret Thatcher's deregulation of the stock market. Then there was the dot-com bubble of the nineties, followed by the credit crunch of the noughties and the pandemic in the decade after that, plus lots of other significant developments too.

For over forty years though, Hargreaves Landstand has helped Britain invest through it all confidently, whatever is going on in the world. Join over two million Brits using the UK's number one investment and savings platform. Learn more at ho.co.uk. Investment returns vary. For claim verification visit hl.co.uk slash platform.

Should governments our government other governments be doing more to prevent um uh even if you think the most likely scenario is not a mega labour market shock Nonetheless, there y you're you're saying there is a trend of certain jobs over time being replaced by AI. Economies will need to adapt. New jobs will have to be created, education will have to change. do you think that governments are conscious enough of the role they should be playing in helping

uh particularly younger people acquire the relevant skills for this new economy? The slightly easy answer is no. Um I mean governments face enormous amounts of incredibly urgent day to day stuff. And w we know two things. Exponentials are really hard for humans and we know uh that long term planning is really hard inside the like institutions of government as we've got them currently set up today.

And so it would be sort of inconceivable to me that we could be planning effectively for AI. But it it is worth pointing out just how hard a problem this is, in the sense that Um on the one hand there will be people displaced. On the other hand, if we artificially limit our companies from doing uh from sort of uh uh making people redundant

then they'll become uncompetitive and they'll those people will lose their jobs anyway. And so finding the exact like threading this needle of what's the right level of displacement and how do we make the best of that and cycle people round. Is actually a really hard problem as well. But so but just to look at schools for a second. At the moment, schools are

often saying to their pupils, you can't use AI. Now that seems to me to be That's mad. Clinically insane. Completely mad. Yeah. I think it's in absolutely insane. uh to do that. I I think that's completely the wrong approach. if AI makes it harder for, let's say, us to test people or whatever, we have to c reconfigure the tests, we cannot take away the absolute best tools for learning.

that maybe humanity has ever invented. It's so out of date. You know, I've got a six year old who, you know, her her IT lessons at school are just absolutely out of date with what she can do already.

in in the house. Can I ask you then, just you know, w w Robert and I have talked a lot about the problem of youth unemployment, the problem of young people who are just disengaged with the jobs market, with education. What could What should they be doing to get a j you know, a job like what if you were gonna sit with a young person now, this is what I'd learn to do. Is it coding? What what is it that I you wish should all of our young people should be doing to stay relevant?

So I think there are slightly different answers at different ages now. The technology's moving that fast. So I think for my three year old, there are different answers to sort of like an eighteen year old today. So we're about to go through this incredible transition um of the kind of industrialization of software development. So if you think about like the production of cotton before and after the industrial revolution.

before it was this incredibly artisan thing that meant there are very small volumes of it and, you know, it was very, very highly prized. And then after We can do it industrial scale, it becomes like just, you know, the default um uh material for for clothing. I think software development is about to undergo a similar transformation.

We have seen n absolutely no lack of demand for software. So I think there's just gonna be this unbelievable elasticity and ultimately we're just gonna build more and more and more. But instead of it being this artisan craft, it's going to be available to everyone all the time, no matter what your software development skill is. And there will be people who are better at understanding other people's problems and better at seeing the opportunities.

and uh better at specifying what it should be. There's obviously an enormous amount of debate about what they call recursion, which is when the AI itself is better than humans at improving itself, how close do we Uh how close are we to that? So I this is another one of those massive open scientific questions. Um

It depends on your perspective. So my personal perspective, we are gonna make like programming really easy, but that is not going to have the like recursive effects on other parts of the economy. Uh so um I would say, you know, in principle if you uh you know, you could take the world's smartest mathematician and say, well, why aren't they brilliant at gardening? They know they know maths. So they can just figure out gardening and they should have the best garden. Of course.

Real knowledge doesn't work like that and I think there are good good reasons why it doesn't. A recursion you think will work within a very narrow area. So you set an AI a task and it will be able to self improve within a very narrow area. But the idea of general Intelligent Constantly um improving on its own in an exponential way, you don't think that.

will happen. Uh well I would say if you that I don't think if you put that intelligence in a box, it can just recursively self improve its way to brilliance at everything. Um, what I think it has to do is I think it has to interact with the universe.

And it has to be able to test theories. So this is world AI. This is the so called world AI. Yeah, sort of embodied AI, some people will call it. And so but I I should say that this really like there are lots and lots of very smart people that would totally disagree with me. and uh that would say we are plausibly gonna get this recursively self-improving AI.

within the next few years. So what are you telling your three year old just out of interest and what should I be saying to my six year old about skills? I think we already know huge classes of jobs that um we do because We care about the experience um rather than the outcome. So to take a slightly silly example to illustrate this, we can already move a hundred kilos of mass over a hundred meters much faster than you saying bulk can run.

Right. It's never about the economically efficient way of getting Usain Bolt over a hundred meters. It's about the experience, it's about the competition, it's about what it means to us. And so, you know, Magnus Carlson, the chess player, already knows what it's like to have a job in a world where super intelligent systems can already completely dominate him at chess. He will never ever be at best chess computers. no matter how hard he tries. And so

But actually, when you start to think about it, there's actually like quite a lot of these types of jobs. So anything sort of artistic or artisan that involves like human taste. or live music or live comedy or professional sport or, you know, podcasting or these kind of things. There are like a million and one ways um that we do things because we care about the process and that process is actually somehow importantly interwoven into the outcomes.

So the thing I'm thinking for my three year old, which is kind of annoying for me,'cause he's very good at maths and I know how to teach him math, is that that sort of that job where you're pushing the frontiers of knowledge for economic value, I don't think we'll uh I don't think that will be open to him in twenty years' time.

But I can easily imagine a world where we we still do maths but we do it more as an artistic pursuit um than as a kind of economic pursuit. I really like that as a vision. I mean obviously obviously it's conditional on us not becoming the slaves of the superintelligence. Uh can I actually yeah, just to make that point a bit more like I do actually think that um, you know, if you look back at at sort of humanity and where it evolved, we are kind of egalitarian hunter-gatherer tribes.

and our instincts are very much geared towards that. And in some ways. to sort of force us to be part of this large industrial machine. We have to like train those instincts out of ourselves. So a lot of education, some people would say, is like You have to force people to think about themselves as part of a hierarchy so that they can then fit into a kind of more industrial workplace.

But that's dead, is w that should be dead, is and I like think that's the wonderful thing about all of this. Out the other side, I actually think we'll we'll have much will be more human, like people will look back on that p on this period being like, what, you had to go to work and be told what to do by somebody you didn't like? And like That was how you had to do it to survive. It'll feel much more like coal miners do to us.

Well of course they were the sort of Silicon Valley of their time, but we look back on them as having to take these terrible risks and get sick to survive and these kinds of things. I think out in the long run, looking back, we will feel similarly to those people as we as coal miners do to us.

But I do think this bummer. So you believe in the world of plenty, you do believe that this will generate significant amounts of income, better lives, better health and all the rest of it. I but again it gets back to my government point. At the you know, that can only happen And I... N ninety nine point nine nine percent of the profits don't all go to eat on mask.

Uh um you know, the the fruits of all this stuff are properly shared. I think that's exactly right. Like what I'm saying is if you zoo if you get the path right and you go out a thousand years, I think there's brilliant futures available.

But it's a v it can be quite bumpy and quite path dependent on getting there and there's no guarantees that it gets you there. Yeah. I also think it plays to the point that Robert and I often talk about, which is with education is we we often refer to the most important skills in life as soft skills when they're very much essential skills, which is what you're saying about communication and creativity and everything else is

stuff we often sideline in pursuit of these, you know, pure academic um roots. And actually computers say everything's gonna do that for us and we should focus more on all the things that make us human. And one like one weird sort of perspective on all of this is Silicon Valley is actually commoditizing the very skills that it is particularly good at and all the artistic skills that they don't prize so much are actually then gonna become

the much more like competitively valuable skill set. And I think there's definitely some truth in that. Yeah, God, I hope you're right. I hope you're right. Can I can I just ask a f uh a a a a slightly narrow semi patriotic question that I have to say is in the context of the really big

um ideas we've been talking about is probably a bit trivial. But we do talk on this program quite a lot about both the sadness and potentially the economic cost of the UK of great pioneering companies like you selling out to overseas interests. I mean, you know, Accenture, Dublin based. I sorta think of it as American but it's Dublin based. Was there a route where you could have stayed British and achieve everything you wanted to achieve? Yeah, so we wrestled with all kinds of uh

all kinds of possible futures. I ultimately I think it's true you know, it's always a bit hard to say, so I don't want to overclaim anything here. But I think it's true that if A couple of years ago we'd had really ambitious British capital that would have come in and supported us. uh, to compete at the top levels of the world. We would have taken that and this wouldn't have happened, or at least wouldn't have happened in quite the same way.

But even then, e let's say that did happen, even then we still would have been faced by this dilemma of the moment where we decided that the mission was actually better served. by having much more global scope. And so Maybe, maybe if you wound back ten years and we could have gone on a very, very dramatically accelerated path, which I'm not sure we would have been capable of as entrepreneurs, but nevertheless, let's say we could have pulled that off and we could have been

fifty X, a hundred X the size by now, then may maybe something could have looked different. But Um so I I we wrestle with this and I think there are arguments that this could easily be end up being better for the UK because

faculty as a s as a multi-billion dollar company was never gonna prop up the economy in any meaningful way. Um but this way we can do a bunch of um We have this now global rate, global reach, global scale, and uh are still gonna be London based, all the people are gonna be here, all the skills are gonna be here. all the um all the money that flows through the ecosystem when you get an exit like this that tends to have lots of good consequences downstream is gonna be here.

So it's it's really hard to know. And kind of the thing we hung our hat on in making the decision was that ultimately the mission was best served by this. and the mission of safe widespread adoption actually massively benefits the UK as well. Did did you feel any pressure to go to Silicon Valley as well? It would obviously be better to be in Silicon Valley, um, as uh uh when we were independent. It would obviously have been better

uh to be there. I we stayed for family reasons primarily and like patriotic reasons, not because we thought it was absolutely the most effective place for faculty. There's actually some slight subtlety in that, in that to get the quality of talent. Faculty would never have had the same quality of talent in Silicon Valley um as it did in London because um Silicon Valley's so competitive.

And we had this weird access to talent. But And we probably are almost out of time. Th there was one other question I wanted to ask you, which is we do have unbelievable talent in this country and a lot of it's connected to our university system.

If there were one thing that any government could do to um tr uh essentially convert more of that talent, into wealth generating projects that benefit the UK, what what would that be for you? That is a very a very big uh question. Um I mean ultimately uh I think the UK There's there's really two things that you need to do to make the UK um the

home of the next generation of tech companies. So I guess let me just unpack some assumptions first. I assume that most economic growth is gonna come through technology companies that grow really fast. I mean that's what we see in America. Doesn't have to be true, but that's a kind of built in assumption here. And I assume that the best technology companies need the the most talented, most ambitious founders.

paired with lots of very ambitious capital. And so if I was in charge, I'd be trying to make those two things happen. I'd make it very, very easy for ambitious founders to move here with visas and I'd like reduce the cost of uh them coming and the you know, at the moment they have to pay into the healthcare system in a way that's makes it is quite a big burden for a small start up and things like this. So I try and make sure we were really, really accessible to talented founders.

and then I'd try and make sure we were very attractive to um to ambitious capital. And unfortunately the only real way I can see to do that is to give essentially tax breaks to um to uh like fast growing startups that end up um becoming very big. Now, I don't love that because I recognise that in some ways these people are gonna become very rich and like taxing them less doesn't feel necessarily fair.

But I also think that in the long run, the country as a whole will look back on that and say, Actually, you know uh we desperately needed some growth in our economy. We cannot flatline and that was the necessary consequence of or that was the necessary trade off to me. Yeah, that's really interesting'cause it would be such a hard sell to people here, wouldn't it? But y y like you say, it's about thinking about the longer term. The only thing that I would say that makes it slightly easier is

you would be giving tax breaks to extremely tiny companies when they were founded. So the actual break you'd be giving them would be minuscule. Ultimately if they did become successful, then it would look like a lot more money. But that would be ten, twenty years down the line. Yeah. Fascinating. Mag, thank you so much. That's we could carry on for another hour, I think, but we should probably let you get back to your

That was absolutely brilliant. Honestly, so much uh gripping stuff that we covered. Uh thanks again. And well good luck with your new life. Thank you very much. Lovely to be here. Yeah. Yeah, thank you very much. And that's it from us. On the rest is money. Bye bye. Goodbye.