The Exchanges

Q So you, you wrote, uh, a, uh, a really interesting piece and you called, uh, Claude the Macintosh of AI. Do you want to talk about what you mean?

A Yeah, absolutely. So Claude is the, Uh, the, the chat box, uh, and large language model built by Anthropic. Anthropic is, um, created by some escapees from, uh, from OpenAI. Uh, and they came from a, uh, uh, the, the safety side of, of, of OpenAI. And Claude, um, has It's kind of, I've got a friendly personality. I mean, the way I used to characterize using Claude versus ChatGPT was that ChatGPT was a little bit like the really, really nerdy You know, honors physics student, uh, college who knew everything was, but was a little bit pedantic and unpleasant. Uh, and Claude was like the really scholarly history humanities student who knew an awful lot. It was just really charm. Uh, and so depending on what you're after, you go to the, to, to different character. And that was certainly the case before. Anthropic had released their latest model, which is called . And so GPT-IV was definitely, in a sense, better employed. But what I would found, find in the 10 to 20 hours a week that I work with these systems is that I would prefer to use. I would get less tired. I would want to interact a bit more. And it came back from, it was based on the The personality that the software has. And, and personality is a kind of interesting phrase to use. I mean, I think about it as the user friendliness. Uh, I think about the way in which when, with the Macintosh bus came out in, in 1984, uh, and i…

AI assessment note: “I think about it as the user friendliness.”

Q So do you think, uh, regulation of AI is a somewhat futile exercise from that perspective because it will just move too fast for any movement or body to just redo anything?

A No, I think we have to, um, uh, we have to have regulatory frameworks in place for, for industries. Uh, they, Are always, um, in general, you can have bad regulations, uh, that, that don't make sense, but the pharmaceutical industry is much bigger for being well regulated and for being completely unregulated. Uh, and, you know, I think we're starting to see the airline and the aviation industry, what happens when you change You know, regulatory quality. Um, so, so what should happen with, um, with AI I think is, um, is really, really complex because you have, you have three different, um, things going on. You've got In what way does AI behave like things we already regulate and therefore we should be thinking that these applications need the same treatment? The second is, are there particular applications that emerge or ways in which AI applications work that, um, Create harms or create risks that are not captured by things that we have already recognized as being problematic. And the converse could also be that we thought of something as being problematic, but we don't now because the technology has changed. And then the final bit is, is there something that is unique about AI? Uh, that means that we need to have some sort of stage framework. And the obvious example is About the safety of frontier models, which people are very, you know, some people are very concerned with, wh…

AI assessment note: “No, I think we have to, um, uh, we have to have regulatory frameworks”

Q Uh, I'm originally from France. I've been here, you know, many years and then more foreign US citizens were discussing, but, um, you know, uh, the, the, from a European perspective, the fact that a lot of AI is being built in the US, that's super meaningful. Yes. And, and China, like how, how, how do you think?

A Well, The, I think that the US has got a, has got a lead, um, and the, it's quite hard to read how far behind the Chinese are or arms. And so in recent things I've read, um, some people believe that the gap has been, has been narrowed. Um, I would find it surprising if the gap has been narrowed for a couple of reasons. The, you know, the first is that in moments of science like this that are driven by talent, and then you have to build instruments and tools to support the frontier, uh, those ideas and capabilities tend to agglomerate. And so there's a lot of advantages that the US has from All of the transfer wall for seeing, you know, Google, uh, the venture capital community, open AI, and Anthropic and, and other companies, and then the supply chain or the scale AI and, uh, you know, impossibles and so on. You do the training and that, that are all quite mature. They're, they are an ecosystem that has evolved within the dynamics of the US system. And, and so how do you race against Against that, especially as the tools themselves help you build better tools. So we're using GPT-IV to look at GPT-III to figure out better explainability, which we can then hopefully apply to bigger models later on. Um, so, so I think that the underlying The, the, the network dynamics that exist within that ecosystem would naturally favor the most mature and developed ecosystem, provided it mainta…

AI assessment note: “I would find it surprising if the gap has been narrowed for a couple of reasons.”

Q So in the same vein, like, one of the most, uh, obvious and talked about impact of the rise of this exponential technologies is its concerns around the labor market and, and what it What means for people in the jobs? What is your view on the impact of AI, robotics, automation, and so on and so forth on labor?

A Um, you know, it's funny when I, when I wrote the book and I started the research, uh, all, all of the analysts, well, not all of them, but many of them were talking about the robot jobs apocalypse. Uh, and of course, here we are after the pandemic and, uh, every employer is desperate to hire wages are rising really, really, uh, rapidly. Um, and robots still can't open doors. So the, you know, I think, I think the, the reality is, That the data seems to show is that companies who invest in robotics, automation, AI, um, grow. Uh, but it's companies that don't that shed the job. So where there are job losses, it's not because, um, Marvin the paranoid Android showed up one day in first mark and was more skeptical about everything than you and you lost your job.

AI assessment note: “companies who invest in robotics, automation, AI, um, grow. Uh, but it's companies that don't”

Q It's a really interesting trend that you highlight in the, in the book, uh, which is the return to local. So in a world where, you know, we all grew up in the, the world is flat, um, you know, type, type trend towards, you know, globalization will only compound and accelerate. This, this seems to be very different. So what, what, what is it and why is that happening?

A In general, what we've been able to do with, by reducing the cost of communication, Through the car and the telephone and the internet is, is expand supply chains and distance. With these new exponential technologies, um, what we're, what we're able to deliver is, is locality. So, um, solar power, uh, wind power, um, grid, virtual power plants and batteries made up of the, the batteries and electric vehicles, um, all enable You know, you know, you know, we have a local rather than national provision of one of our key inputs, which is energy. So you don't need to spend 3.4 billion dollars on a 600 megawatt coal plant. Um, you can just, as they have in South Australia, And feed into the, into the grid. Um, and a similar thing is happening on the side of food. So of course your, um, uh, participants will be familiar with bioreactors and lab grown meats, but also high intensity urban farming, which is fundamentally a game of, uh, data, genomics, uh, robotics, um, and, and AI will for the first time since we actually had Um, cities, um, in from 9000 BC, allow cities to be self-sustaining in their food, and three-D printing, while it's still kind of weird and small, will also allow, take up some of the slack for the production of either finished manufacturers or parts. So I think what you see there is a set of headwinds against this idea of really, really deep, um, and extensive and …

AI assessment note: “With these new exponential technologies, um, what we're, what we're able to deliver is, is locality.”

Q Where, where does it all, um, leave us? Like how, how do you see the next, uh, five years and then the, the next 20, where do you think that's, that's all going?

A And there's a set of technologies that are really on that, um, on that cusp, uh, of, of being sort of amazingly interesting. And I think you see, uh, the synthetic biology arena that overlap between data-driven biology and machine learning. Starting to, to come to fruition, and what we're doing is we are, We're struggling on the scaling, so no one has really got the scaling right, but more and more people are getting the front end right, and, you know, the scaling will therefore follow, right? We'll figure out that it's all about having custom bioreactors for whether you're making a therapeutic versus a meat versus a material, and that will all start to fall into place, but I think that that is the sort of most ripe, really interesting technology, but then I also think that you're seeing Over that 20 year time frame, real breakthroughs in, um, you know, two super hard technologies that will rely on more ancillary technology, so quantum computing and nuclear fusion, um, that will need all sorts of breakthroughs in material science amongst other things, um, that we could expect to see. But the other thing to just think about is that, you know, in the decade to, um, between 2011 and 2021, Um, just look at Germany, because I spoke to the, the, the sort of top people at Deutsche Telekom recently. Um, the amount of data across Deutsche's, um, mobile network increased 300 fold over th…

AI assessment note: “Over that 20 year time frame, real breakthroughs in, um, you know, two super hard”

Q So I know you spend, um, time advising boards and corporate leadership teams. What do you tell them these days? What do you ask them and what do you guide them towards?

A I tell them to all of them to go and use It should actually be to applaud for a couple of weekends, solidly on their own, and to not stop until they've done like 10 or 12 hours. And I say, this is not a technology that you can learn through PowerPoint that your juniors have presented to you, and you tick off. This is like the internet in 93 or 94, which unless you actually used it and started to grok what it meant, You would not be able to explain why you would use it over X 400 or the telephone system or an ATM connection. So you have to go off and use it yourself to make sense of it. And, and, and one of the reasons for that is that If you view the current set of AI tools, large language models, as NLP, natural language processing on steroids, right, higher quality, faster, all you'll be doing is looking at point tasks that you can improve. So can we, um, look at the flow of Emails that come in and routes them more efficiently than we previously have been. Can we automate this particular task where, which is an edge edge case that we haven't spent the time putting in the nested if, if they're now statements in Python, but maybe an LLN can kind of deal, deal with it for us. That's either no one's going to build a great business that way around. Uh, what you need to do is have the experience that I had Which is, wait, I can now think so much better because I've got a cognitive …

AI assessment note: “I tell them to all of them to go and use”

Q One of the, um, characteristics of this, um, exponential age is the rise of, um, what you call the unlimited company, what, what others call the superstar companies. What are those and why are they growing so fast and becoming so dominant? And I guess why are they so different from the previous marketing companies?

A When I, um, started to learn about business in the early to mid nineties, the general rule was that companies could only get to a certain size because competition would eat away at them. Um, and that, uh, there were a few, there were a few little aspects, right? So first of all, was that one, one was that if you had a successful company, a new competitor could come in and compete away at the margins in, in some other, uh, way. The second is that Um, managerially, things would get too complicated. This was sort of Ronald Coase's insight back in the 19 thirties, that effectively at some point a company would get so big that it was so expensive to do things internally, it'd be cheaper to go to the market. Um, and the third was this idea that companies would have increasing, um, uh, increasing costs, right? So the, the 10,000th ton of steel that a car company bought would be more expensive than the 500th ton. And at some point it just wouldn't be profitable to, to source that. Um, and so there was this idea of a 40, 20, 10 market, right? The first, the top company, a 40% market share, the second would have 20 and the third would have 10. In these exponential markets, you don't see that, right? So Google is number one, and some of us can name Bing, and a handful of people use DuckDuckGo, but who's the fourth biggest search engine in the US? I mean, I don't know. And the reason for t…

AI assessment note: “In these exponential markets, you don't see that, right?”

Q So why, why does it matter, uh, what personally it adds? Is it a question of, uh, whoever has the best personality will make more money? Or is that a question of like broad adoption of the technology by humanity?

A You know, I, I think it's about, uh, you know, user, user friendliness and about, uh, helping us achieve what we need to achieve. So I have, I have this discussion with people quite a lot because I have normalized everything and, and always Have since I was a little child, I'm just one of these people who, you know, puts personality into the tea, into a teapot, right? Or into, into a car. Uh, and, and, and so I'm not falling, but I also know that it's only a car and it's a teapot and it's not a person, even if I'm given it a name. And, and so I think that, that anthropomorphizing, uh, these, these products has to be done with a degree of care. But when I put my, you know, my user interface hat on, Having something that has a little bit of a personality absolutely helps you get your work done. Uh, and, and so when we think back to these computers in, in sci-fi, uh, the ones that, that respond in very pert ways, push a lot of competitive mode back onto the person. If it just says affirmative, negative, affirmative, negative, yeah, 200 light years. It, it, All of the work has been done by the person. Whereas once you start to make the interface a little bit more intuitive, the way that we start to think, Then the software is taking on some of that load. So I think it is about building better software. Now, if all you're doing is using the large language model to extract entities, …

AI assessment note: “I think it's about, uh, you know, user, user friendliness and about”

Q So the, the next logical step in thinking through sovereign AI is, uh, perhaps sad label warfare, right, as a big part of the geopolitics. Does AI change warfare, like, or is warfare changing as part of that exponential age?

A I think, I don't know how to interpret that, um, open AI So Paul Nakasone, who was Obama's top cybersecurity and cyber offensive guy on their board. And there are so many conversations happening with NatSec on these, these topics, because clearly it's a really important defensive and offensive tool. You asked how it, how it changes, um, things. It, it changes things because we've got this sort of ladder of escalation where still in our head, the worst possible thing a nation can do to another nation is to physically lob a kinetic projectile at it. Uh, and so all of the work that happens in that sort of gray space of information operations or cyber attacks, uh, has not really been seen as, um, Has already been seen as, as, as conflict. Uh, and, and, and, you know, in my book, I write that what AI does is it reduces the cost of, of aggression because it makes information operations much cheaper. You can flood the zone with realistic sounding tweets. By the way, you have a wonderful Twitter feed. I think you're one of the funniest French capitalists.

AI assessment note: “what AI does is it reduces the cost of, of aggression”

Q Yes, it was something like that. Had, um, so, so, uh, you know, great on the over, Overarching thesis. Has any part of what you wrote moved slower than you'd have thought?

A Well, I think that, uh, there were a couple of, um, key questions around, um, the relative power of the, the big technology companies and That power was part of the point that I was making was that you couldn't use the traditional lenses of, you know, monopoly to look at companies like Amazon or Apple and so on. You'd have to use these new lenses because they behave rather differently than the products that are cheap to use, and also their impact on our public space and political space. So, so what's happened since then is that there has been A new set of technology competitors in the sense of open AI sort of showing up and changing the game a little bit, but also the Um, especially through the EU, the willingness for governments to start to be much more proactive and ex-ante about the way in which they look at these companies. And of course there is, um, Lena Khan, I did actually write about her in the, in the book is now the head of the FTC and she's taking a different, a different stance. So that, that's, that's one dynamic. I think the other thing that has, um, uh, has been Perhaps we've got a bit more evidence is What really was the impact of social media on the fragmentation of the public space? And in the three and a half, four years since the book, I finished working on the book, it takes time to be published. There's more evidence from the digital humanities, the digit…

AI assessment note: “what's happened since then is that there has been A new set of technology competitors”

Q Any sense for why there's such a discomfort between Asia and the West, or is that just always being bad at explaining it?

A No, I think it's, I think it's that, uh, Asia has been getting richer rapidly, and connected to getting richer is more technology, more IT, and whether it's causal or correlated, it doesn't really matter. You just noticed that, ah, you know, technology came as you got richer, and technology created new opportunities to build businesses, and so you're able to be entrepreneurial, and there's a higher degree of, of optimism, whereas I think, and speaking of Europe, you know, Europe is, um, Uh, hasn't had a growth story, uh, in the society in Western Europe, and I did sort of specifically mention Western Europe, because I can't remember the data for, for Eastern Europe, it's a former, um, also quite countries. Um, Western Europe, there hasn't been that, that great story. And I think culturally we, we've got tied up on non-growth arguments that are, that are quite far away from the reality of science, science and technology. So part of the, You know, part of the, one of the things I think prescriptions and I think politicians need to, um, to take on, and it will be easier in the UK, but we have an election on July the fourth, um, that in, in the US is a way of putting science and science discussion at the center of whatever public debate that happens to, to be on whatever public interactions there end up being. And you have to do that against this very fragmented media, media landsc…

AI assessment note: “Asia has been getting richer rapidly... Western Europe, there hasn't been that, that great story”