Will AI really deliver growth?

0:01You know what I love, Damo? Things that save me time. You don't have YouTube premium, mate, so I just don't believe that. Granted, I'll give you that one. However, I've got one for you. A great time saver in personal finance is Money Week magazine. They spend a lot of time distilling the biggest stories in personal finance down into consumable chunks, so you don't have to scroll and scroll. They give practical tips on savings, investments, pensions, the UK economy, the global economy. It's like your five a day, but for finance. If you want to give Money Week a try, you can get six issues in print and the app absolutely free by visiting moneyweek.com forward slash money.

0:34After your trial, you'll save an extra£5 a quarter on the subscription, which is exclusive to Making Money listeners. And that's moneyweek.com forward slash money. But there's a link in the description if you just want to click that.

0:50Yes, I mean, it's a bubble, obviously. Is OpenAI worth the gazillions of dollars people say? Probably not. Is AI going to solve our productivity problem? Diane Coyle is an economist and professor at the University of Cambridge. Her new book is called The Measure of Progress, counting what really matters. So I think it's going to be quite a long time before we see the use of AI translating to big increases in productivity. I think I should hit that button because maybe we need to talk about what that means. Go on, hit it. Hit it, first of all. It's very... I want to start with something that Keir Starmer said.

1:25He said that he wants to make the UK a champion of AI. Do you think that that's something that we should aspire to do, first of all? I think it depends what you mean by champion of AI. We are really good in this country at frontier AI research. There are some fantastic people in the universities. So if you think about how can we make this an export success and create great jobs for people and upskill people, then that's a really sensible aspiration. We've only got one big company. It's DeepMind. It's owned by Google. So can we grow from those university researchers, some companies that are producing really useful services that people want to pay for and we can export them?

2:08That's a great aspiration. There's another side to it, which is using the AI created by the big American tech companies. And I'm much more hesitant about how that's playing out because there are people saying this is going to completely reform our public services, but they're short of money, but this will fix that. And I don't think that's true at all. And I've got a lot of reservations about the way that's being done. DeepMind does incredibly exciting frontier AI. It did the program that beat the world champion at Go. So it's tested it all on games, but it does things like protein folding, which is a way of understanding how molecules in medicines are going to behave.

2:51It's got the best weather forecaster in the world through an algorithmic device. So it's doing really fantastic kinds of research. The other companies we've got are startups. So they're all really fantastic too, but they're tiny. And they're not in that sort of stage of commercial maturity where we see lots of jobs and lots of growth coming out of them. Do you think the UK has that potential to deliver commercial growth and big AI companies? I do, except the trouble is all the startups want to be bought out by Google or the big tech companies because that's the easy way to make your money out of it.

3:24Yeah. So we could, but I think it will take a little bit of work to make it happen. You said that you're worried that it's not going to deliver the productivity shift that people think it will. Yeah, well, think about what it is. It's an information technology. It's a way of collecting up a whole load of information that is gathered from the Internet and other sources of data and using that to help people make better decisions or predict things. and that means that companies and other organizations have to figure out how to use it. So you've got all this new information. How have you got the data?

4:00What data are you giving access to to help you make your decisions better? How are you going to change your organization so that you can use that information to do things better? What does that mean about accountability and hierarchies? So for example, there are loads of great startups in health tech using AI. They can do things like, you know, help doctors write letters more easily or provide information to nurses on the ward more easily. Does that nurse then have the authority to make a decision about something or will they still have to go to the doctor who's busy in the ward down the corridor to get the approval to make the decision?

4:39So there's a lot of restructuring that gets done. People have to learn how to use it. They've particularly got to get the data in shape and most organizations don't have the data that they need. You can't match TURPS, different formats. You're worried about GDPR and regulation. You're worried about data security. So I think it's going to be quite a long time before we see the use of AI translating to big increases in productivity. You said productivity. I think I should hit that button because maybe we need to talk about what that means. Go on, hit it. Hit it, that's all. It's very satisfying, isn't it?

5:12It's cathartic. But yeah, go on and define. Well, you know, politicians are talking about this all the time. People in business don't use the term. What we economists mean is it's what you get out for what you put in. So given the resources that you have, the people you hire, the equipment, everything else, paying for electricity and all of that, how does that help you produce an output that your customers or users are going to value? And if you're in business, that they're going to pay for. So it's what you put in at one end, how do you use it and what you get out that's valuable at the other end?

5:43Are there certain characteristics of technology that make it, you could look at it and go, this is going to improve productivity? I would always look for how much time is it going to save. If you look back at a lot of technologies in history, what they fundamentally did is save a lot of time in the process. Washing machines. Washing machines, great example, although they don't count in the formal economy that we measure because it's household use. But arguably, one of the most important innovations of the 20th century, because of all the time it's saved from washing clothes by hand. But steamships versus sailing ships would be another example.

6:20Or the just-in-time revolution in manufacturing in the 1970s, which was about making the assembly line work in a much more efficient way. Similarly, that would be how AI does improve productivity if it's going to do it. Because I was surprised to read that 5 % to 10 % of companies admitted to being able to use AI. The rest were like, we don't really know what to do with it or how to use it. Yeah, the numbers vary depending on the survey, but it's definitely a minority. And, you know, it is quite hard to figure out in many businesses what you would do. If you're thinking about chat GPT summarising things, that's okay.

7:00Lawyers use it and accountants use it. Sometimes that gets them into trouble because it makes things up. It's called hallucinations. And so the information is inaccurate. But many companies, never mind AI, they don't really know what to do with digital. You've got to have the mind space and the time and also enough workers with skills to start to use these new technologies. And it's only 5 % to 10 % that have seen their productivity go up because of previous digital technology, never mind AI. How long do you think it will take before we start to see productivity improvements? That's a big question.

7:39I think we ought to stake something on it and reconvene in 10 years' time. Okay. I'm going to say 10 years. 10 years. 10 years. And what will that look like, do you think? There's another big bet to make. Are we all going to be on universal income because the robots have replaced all of our jobs, or is it just going to be that? No, I don't think that. It's really hard to do that kind of forecast. If you again look back, electricity took about 50 years to have an effect on productivity. And that required people to get away from the tall factories that you see in Manchester or Glasgow, which were built around one steam engine or two steam engines.

8:20and you laid out flat factories like the Ford plant at Dagenham because the assembly line could have an electric motor on each machine. So you had to completely rebuild that and they were in different places. So that meant building houses there. So you started to get the suburban growth that we saw in the 20th century. And then you got domestic appliances, which completely revolutionised home life and allowed women to go into the workforce. So that sort of long term big picture is really hard to predict from this end of a technological revolution. So I'm not going to place a bet on that. But you do think it is a technological revolution?

8:55I probably do. Yes, I probably do. You think like the internet, like that kind of scale? Like printing, like the internet, a big information technology change. That's nice to hear. Because you do feel really like, are we just all sipping the Kool-Aid a bit or are we going to be carried away? A little bubble and it's going to be when no one's talking about it in a couple of years. Yeah. But you can get both. I mean, it is bubble, obviously. Is OpenAI worth the gazillions of dollars people say? Probably not. No, no. But then after the bubble burst, the technology's still there. We've spoken before about inshittification and this idea.

9:27I mean, we can hit the button if we want. Damien's favourite word. Two in the first half an hour. I think that's sort of good Anglo-Saxon phrase. You can figure that out. Yeah, yeah, yeah. But, you know, we've seen... So if we look at some of the big tech movements, so Google has clearly got worse in the past few years in terms of the search results are poor and there's more ads. Uber, I think, is a classic example of a service that seemed to be like, they've improved taxes. Now I would just rather have a black cab, honestly, when I'm in London. Do we think AI will just go that way where it promises to be everything and then a few parties will control it and just make it crap for everyone?

10:03As possible. I think it depends on whether this time around the competition authorities in the different countries are on the ball because they weren't in that previous wave. And it's why we've only got the handful of big tech companies dominating everything. And they are then able to get away with the interpretation. I think there's more chance of stopping at this time. OpenAI is obviously a different player. There's a company called Anthropic. There's a French company called Moustral. I think there's more hope as well from different versions of the technology because these big, large language models are really energy hungry.

10:40They cost a fortune to train. and so the the technological people are working on can we do it more cheaply more efficiently can we use a different approach to building much smaller models and if that happens then you get much more competition and you don't get that that kind of phenomenon a lot of people are building on top of them as well so you might get it's like powered by chat gpt but they've they've narrowed down the use case i don't know if perplexity is chat gpt or not but i use that a lot and And do you think it could be, they could almost be like base layers and we see lots of different types of applications built on top, almost like an iPhone with apps?

11:19That's certainly true. And that would be much more competitive. But there's a level at which you're not going to get competition. So these data centers are huge and costly and you don't like, you know, like water infrastructure or something. You're not going to have competition there. So there will always be some level at which there are people who are gatekeepers to the technology. you said you were writing um at the beginning you said you were writing a paper on the laws of what do you call it um governance is the government you better hit the button yes paper on the governance of ai right yeah um what are some of the like fears or like negatives you have about ai because i see it all the time um with these fake ai videos and now it's got so bad that you see something on social media and people like oh that's ai and then someone else was like not everything you see is AI.

12:06So now people are getting so sceptical that they're like, oh, it's a fake video, it's AI, because the AI is getting so good at fake videos. So what do you think are some of the dangers of AI? Well, that's obviously one of the misinformation, broadly, fakes, really big concern. And who knows what effects that's been having already, we've not really got a handle on that. Invasion of privacy, and that takes subtle forms as well, because if I post some information, it reveals information about my family and friends or about people who are like me. So there are a lot of concerns about privacy. The concerns we just talked about, about who's actually getting better off as a result of this and sure we get free search or some free AI, but at some cost.

12:52And then just broader questions, I think, about who's in control of society. and it's really striking to me that things that used to be seen as really central to what governments do are being outsourced to private tech companies. Things like identity systems or the operating system for a lot of the public sector sitting on Amazon Cloud. That's really striking. It's a real difference. These are things that governments have always sought to be in charge of And so there's a question, you know, not a day-to-day question, but fundamentally a question about who runs society. There are the economists like me who go, oh, we're not measuring this properly.

13:36And oh, what's the effect going to be on jobs and productivity? And then there's a political discussion. And somehow joining up these things is what I think is going to be important to make this amazing technology work for us. And a lot of it comes down to power. we've got these huge technology companies we've got multi-billionaires they're buying elections they're running governments there's so the missing bit for me is we're not actually talking about about those power relations and how does that play into the economics conversation and the technical conversation so what i want to do is to get each people in each of these little silos to look outwards and pay attention to what's going on.

14:19Do you think there can also be an element of scaremongering around it? Could there be that we overcompensate and we stifle the evolution of it by coming in, regulating it heavily? Because this is what they will probably say, won't they? They would say that. And I think what they're most concerned about is being forced to compete and being forced to operate in a way that would allow others to come in with better technology in the way that Facebook back in the day overtook MySpace because it had a better platform. And it's really hard to see that happening now. So that's definitely what they will say.

14:54Could there be an overreaction? I mean, sure, but I think that partly depends on how the companies behave. And if they do carry on degrading and shitifying all of these services that we have loved, then that's more likely. The backlash will get all the more severe and that means that there is more likely to be overreaction. I love that the n-shitify is like the accepted term. It was the word of the year in one of the big dictionaries. Shall we tell the people what it is just in case they don't know? We already have. Not in this episode. I mean, it's the n-shitification of something. It's making it worse intentionally.

15:24There you go. Normally with the attempt to make profit, to improve profitability. Last time we recorded, Tamein, you were having some real dramas with your accountant. So how's that been going, mate? They're sacked. So drama sorted. They're a big corporate firm. they didn't really reply to my emails very quickly like took a week or two at times and they charged me way too much I mean I've got pretty simple taxes and yeah they were charging me thousands they saved me some money but yeah I had to move on slow and expensive pretty much yeah this is one of the reasons that we're really happy to be partnering with tax app it's a tech platform that makes self-assessment simple whether you're self-employed like me a freelancer or a director like demo big dog instead of sending endless emails bills and spreadsheets to your accountant, you just connect your bank, answer a few questions that are only relevant to you, and your tax return can be ready in as little as 15 minutes.

16:13TaxApp is really easy to use, and it's HMRC-recognized software, so it's safe, secure, and legit. The price is also decent, so if you're self-employed with one income stream, it's just£89 as a one-off fee, no big accountancy fees, and we also have a discount code, of course. If you need to file a self-assessment this year, give TaxApp a try. We've left a link in the description and use the code MONEY10 for 10 % off your first tax filing. That code is MONEY, M-O-N-E-Y 1-0. So Mr. O 'Carolet, I hear you are a salesman. Elite salesman, yes. One of the best, they say? I've got a little bit of experience in the game, yeah, I could say.

16:50Done a few deals. A bill, a bill. What would your compliance team say about you? They would say that I am always nagging them and that essentially I just have beef with compliance. I love the team. Compliance, slows down all my deals because every time I get to the finish line, they've got to check documents, KYC, GDPR, and it's just a nightmare. It slows the deal down by like two, three weeks. It's always on both sides as well, isn't it? Sometimes it can be blocked on the other side. Well, that's where today's sponsor can help. Indeed. Vanta helps companies of all sizes get secure and compliant fast.

17:24And they stay that way. They do it by automating compliance with over 35 security and privacy frameworks like SOC 2, ISO 27001, and HIPAA. Yeah, all of them. And this saves businesses so much time and money. According to a recent IDC study, Vanta customers save over half a million dollars a year in costs. Not bad. And they also help you complete security questionnaires up to five times faster, which is great because everybody hates filling out forms. If you're a business that needs to prove security and compliance, visit vanta.com forward slash making money to sign up for a completely free demo today.

17:57That's vanta.com forward slash making money. There's a link in the description though, so you can just click that. You recently wrote about time tax. Can you tell us what that means, please? And can we have a talk about that? Yeah, so we were talking a little bit earlier about one of the good things about technology is saving people time to do things they really want to do. And modern life taxes us with time because we sit in voicemail queues trying to get through to a human being to fix the problem for us. or in the supermarket, we no longer have somebody scanning the groceries. We've got to do it ourselves and the machines don't work properly.

18:38So that takes time. So all of these things that are taxing our time for the profit of different companies, and they might look like they're improving productivity. Those companies will be doing better, but we are paying not a monetary price, but a time price for it. So the good thing about the tech will be if it starts saving us time instead of costing us time yeah the um i mean like ads right i i i i kick off at these guys in here when we put youtube on and there's an ad i'm like the time has a value you know i pay for the subscription just to get rid of the ads but i think that's another way that you sound very impatient yeah yeah a little bit yeah no i just um no you value your time i value my time i think most people don't they don't realize that that's like the hardest currency they've got and that it disappears and yeah I think we ought to be we economists ought to think much more about it and uh you know we always calculate in money but um as you say we've got 24 hours a day you've got to spend it all so we should be focusing on how do you do that productively in work and how do you do it enjoyably when you're not in work I think what made me realize it was that because of YouTube what I do on the back end of YouTube I can see how many hours people spend watching my content you're talking like hundreds of thousands of hours you know some videos can be tens and thousands of hours and i work it out and i'm like you know 200 lifetimes a year watching my content amazing mr beast is hundreds of millions of views of video thousands of people's lifetimes a week are watching his content you know that you literally mine the attention like it's a commodity and they do that by slapping ads in front of it and i think when you see it in that way you start to realize like this has real value i might watch a 10 minute video but if 100 million people watch that minute that video the productivity loss if it's not something that helps people if it's just him running around like you know giving away money it's it's massive i feel yeah i feel the same for social media like i dame knows i've got an app that blocks my social media because i look on every week and it says you spent two hours on instagram today you spent one hour and i'm like i could have done so much in that one hour and i learned very little and i just was because it sucks you in into the like it's deeply addictive yeah they want you to just keep scrolling eternally there's no end to it so i'm like i'm gonna go to bed early tonight get into bed at 11 and i'm like why am i still on instagram at midnight so i just block the apps but i feel like they all have a way tiktok instagram to suck you in and they were designed by the guys who built vegas weren't they and they and my biggest worry about ai is we turn the best humans against everyone in terms of attention and now we're going to design an AI that will know us even better than we ourselves and it's just going to become so efficient at grabbing your attention that we will find it like it'll be like heroin trying to pull your phone away from your face there was a novel uh called brave new world uh which drugged people into uh just you know being passive and happy with everything and we've devised that but the drug is is the social media scroll isn't it yeah although I do sometimes find that when i've seen enough videos of hamsters being cute or whatever i go what am i doing yeah that's it it's when you have a moment of clarity you're like what am i because they show me basketball then i'll show me like someone in a bikini then i'll show me like football then i'll show me very different algorithms you to me hamster in a bikini hamster playing basketball instagrams that's just what they show everyone you are in charge of what they show you because on my new instagram page which i literally only follow people that have been on the podcast they show me they still show me like random girls in bikinis and i do not because they get you know they're like people tend to like random girls in bikinis ronaldo or cakes yeah so take your pick but it's a brand new account and they're just showing me all these things to keep me so i'm just i'm only going to follow professionals and still they're showing me the same like our nonsense that's just dragging me in so yeah it's um it's a bit annoying okay the good question i think we've been a bit negative where do you see the biggest impacts in terms of positivity over the next 10 years oh there's some amazing uh there's some amazing potential um already being seen in science, in science discovery, new medications, solutions for people who've got disabilities.

22:48The Neuralink stuff, which is scary because it's Elon Musk, is actually, it's got great potential for people who've got neurodegenerative diseases. So there's a whole swathe of potentially fantastic medical things that could come out of this. I talked about the weather forecasting and in a time of climate change producing extreme weather stuff that's going to be really important so all of that kind of scientific domain I think there's there's huge potential probably also in the kind of organizational improvements that allow us to get conventional economic productivity because you know it will help people in lots of jobs that are quite routine automate the boring bits, automate the routine bits.

23:35So in my work, we don't do it very much at the moment. I mean, some academics use it to try to generate ideas or titles of their papers, but we're experimenting. Part of our research is looking at policy documents. And normally, you get a research assistant to read them and summarise them and pull out the important points so it's been human. We're experimenting with getting ChatGPT to do that. And that means that if it can be reliable, you can look at thousands of policy documents. So you can expand the scope of what you're doing. In firms like lawyers or accountants, you can see the potential for that to really improve their productivity by automating the more routine things that they do.

24:17So I do think there's a lot of amazing potential. But the questions are, what kind of things we've been talking about really, including can you trust it? Is it reliable enough? Can it drive itself into the floor with inaccuracy? It's what the technical people call model collapse. Yeah. If it doesn't get enough new information and correction, then that happens. And so there's something called reinforcement learning with human feedback, which is exactly about people going, oh, that's really stupid. Let's correct that. And it improves next time around. And one thing I wanted to talk to you about around this was the incentive to make the AIs look more sophisticated than they are.

24:52I'll give you an example. There was the Amazon shop where people could just walk in and grab something off the shelf and walk out and it would charge them. And then they realised that they just had people in India following everyone round on cameras. Do you feel that there's an incentive from big tech to make it look like their AI offerings are more compelling than they actually are? Well, there's a huge hidden human input into all of these models. and it all started with the image labelling, which was done by people in low-income countries being paid not very much to label images so the AI could learn from that.

25:23And when you say the labelling, they're saying, this is a car, this is a bike. Or we do it when you've got to prove that you're a human and you label... You have to click on the pictures. You have to click on the buses or whatever. The capture stuff, yeah. We're training AIs. I was going to say, can't AI do that? Because that's meant to prove that you're human. Well, they keep getting harder. You see how the captures, it gets harder and harder because the AI gets smarter and smarter. But then it turns out I'm not human because I can't do it. Right? That's right. Because I'm like, which one? I'm on my phone.

25:50I'm like, which one's a bicycle? I can't sit so small. So yeah. And the layers of them, it gets, you have to do 10 of them to get in to log into stuff. Yeah. This is something I want to talk to you about without sounding crazy. Does the AI get so smart that it's smarter than us? Do we hit a point of like a singularity? Am I going to live forever? Is what I'm really interested in. He just wants to invest forever. Those are different questions. yeah i mean you know that's this is the outcome i want what i want is in the next 20 years for it to just figure out everything and you know to cure death and then i can live forever because well it's really interesting we evaluate the ai by how good it is compared to humans at doing certain things um which is slightly weird because we'd really want the technology to be good at things that we're not good at doing and so previous machines were good at lifting heavy weights because we're limited in how much we can lift i'm not it's all that instagram that you watch oh man i wanted that one um so it will get it will get better than at some things than we are and that's all the cognitive skills and you know what's new about that computers are already better at sums than i am i don't think it's anywhere near being as good as we are at things that are hard to write in computer code, nowhere near.

27:12And that's all of the haptic, haptic stuff. Oh, taking my job now. I'm going to have a job soon. Which is movement and how, you know, being able to tell how close you are to a surface. So that's merging AI with robotics. That will come, probably not quickly, but I'm looking forward to the washing machine equivalent that will do my folding the clothes up for me and all of that. Oh, pairing your socks. Yeah. Yeah. Or finding your socks. Yeah. I think they all go missing in the washing machine. I think there's a secret hole where they all just sneak. That's the price we pay for the washing machine.

27:46You feed it socks. I'm taking 10 % of your socks every week. But then there are all kinds of emotional things as well. And if you think of an example like nursing, there are things that would be great that the AI would do. So it would be great if the nurse doesn't have to stand at a computer terminal with their back to the patient typing things in. the ai can record that and sum it up that'd be fantastic um there are things that you want to happen quickly uh blood tests really boring pretty straightforward uh can we speed that up or you know lots of routine things can we speed those up that should free the time of the medical staff to really pay attention to you or to um give you much better quality care to empathize with you all of that.

28:34To be more human. To be more human. I love that. So all of that should be freed up by the technology. It's never going to replace us as humans. I say this all the time in what I do with content. The advantage is to lean to being more human. Everyone is like, how can I make a million pieces of content using AI? And I'm like, that's a mistake because you'll just get generic bad content. Ask ChatGPT to crack a joke. It's pretty awful. And I say to students, you can use ChatGPT, but that'll get you a b if you want to be a good student you've got to use the stuff in your head some things it's really good like on i have client calls on zoom and i just click ai i pay for it it's like ai assistant and it just like summarizes the whole call and then emails it to everyone so if you forget something normally i used to just write all the notes while i'm on the zoom now i can just drive sit back talk to them and then it just summarizes everything makes a few mistakes but you know it kind of notes everything important so it's pretty good i do both um so you get the summary and um it doesn't always pick out the things that i would pick out so i've got a notebook as well and i write down the ones that i want to remember exactly yeah i hope it has potential for fact checking people talk about misinformation and how they think it will create it i hope it's the other way where we have these tools where we can kind of check the validity of what we're seeing through the ai well the kind of people who do that work that fact checking work i think are definitely using the tools to do that um but you know that's quite a lot of skill people like bellingcat They're very skilled.