Jul 26, 2026 · 2026 #27. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
AI And Its Enemies
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Andrew Keen
Hello, everybody. It's Sunday, July the 26th, 2026. It's time for our weekly tech roundup with my old friend, That Was The Week publisher, Keith Teare. Last Saturday, we did a show in which Keith asked who owns intelligence and talked about what he called the smart wealth of nations. In the meantime, Keith has used AI to author a new book, appropriately called Who Owns Intelligence? No wonder a lot of people are fearful about AI. Keith, what have you learned from using AI to write a book, Who Owns Intelligence? Appropriate title, perhaps, for an AI book.
You know, it was inspired by our conversation last week, the idea of doing it, which was... It's a fairly complicated discussion. The core idea is something called a human wealth fund, which is basically a fund created by AI companies for everybody, initiated by putting their stock into it, global on day one, and not governmental. And we talked a lot about other examples of similar funds, for example, Norway, Alaska, They're not that similar, but they have something in common with that idea. And I suddenly thought you can't really convey an idea like this without long form, because firstly, most people will instinctively think it's utopian and disagree with it. So you need to show evidence that humans have already done things like this. And then secondly, will it work economically? So you need to make a case for it, an economic and mathematical case. And then thirdly, they wouldn't believe that the companies would do it. So I thought the best form is a book. And I started by using AI to craft the chapters in my own logical narrative style and then write them one by one. And after about a week, I ended up with something that approximates what I think.
No wonder, Keith, AI has enemies. I'm an author. Many other authors are going to be really scared because coming up with a book, even if it's an AI book in a week full of ideas, structured, coherent, is terrifying for many authors. And it's an appropriate, perhaps, introduction to our conversation this week. For your editorial, you have... Chosen to talk about AI and its enemies. Who needs a kill switch? Of course, AI and its enemies brings to mind another famous book by Karl Popper in 1945, The Open Society and Its Enemies. So should your, Keith, should your effort to write this book from our conversation last week, should that explain why AI has so many enemies?
I don't think so because AI didn't do much more than take my ideas and turn them into something that approximates my ideas. So what really happened is I wrote a book in a week using AI as a tool. AI would never have been able to start, never mind finish a book without me. So I think authors should see it as, as long as they've done the work and they understand The content of a book that's in their head, they should see it as a tool to accelerate the production of what's in their head. So it's a good thing. It shouldn't lead to enemies. But I do think there will be a superficial narrative by people. For example, Substack this week introduced an AI detection tool, which I've tested. It's completely incapable of accurately detecting AI. But the spirit behind it was to stop AI unknowingly entering into the lexicon that Substack publishes. So yeah, that definitely would.
And we both publish on Substack. It's interesting. Would Substack then sniff out your AI, or at least maybe somewhat authored book, Who Owns Intelligence? Would you be able to publish this on Substack?
No, I think it's because I use AI to initially craft the structure of an editorial. And you start with its view of what it is I want to say. And then you go through it line by line and rewrite it. So you end up with something that AI has barely touched other than structure, I guess. But the tool says it's 100% AI, which...
Oh, you've got to be careful. They're going to get thrown off Substack. We won't be able to do this show anymore. You say in your editorial that... AI, when it comes to its enemies, and I'm quoting you here, the enemies of AI are mostly good people. That kind of reminds me of some of my best friends are Jews. Who are these enemies of AI, Keith? I mean, it's an unusual, I know you sort of borrowed perhaps from Popper's book, The Open Society and Its Enemies, but Um, does AI really have enemies? Or does it have people who are uncomfortable? An enemy is a hard, sharp word.
Yeah, and one of the pieces actually is from Casey Newton in this week's reader about Congress proposing an AI kill switch, which seems certainly highly science fictional, but it will seem.
I bet they're calling it the Kill Bill. Yeah, they should if they don't, shouldn't they? Now, interestingly, if you ask who's in favour, it includes the AI companies. And at the same time…
Not only Dario. Sam as well, Slippery Sam. Slippery Sam. But paradoxically, the other thing that happened this week is they all signed this, it seems like every week there's a new open letter. This week's open letter was in favor of, it was written by the CEO of Nvidia, and it was in favor of open source AI. Which, of course, is the very thing that the proponents of this bill are trying to drop.
And I'm always so dubious. When people use the phrase open source, I've never understood what it meant. You know, Keith, in your editorial that we've been through this before, you've You've seen this hostility. I'm quoting you. I've seen this pattern before. You and I were both around in the mid-90s. You were even around in the 80s with the internet, or certainly the early 90s. I got involved in the mid-90s. Is this a repeat? I don't remember the 90s being full of people who wanted to kill the internet or switch it off. There were people who were Skeptical, dubious, but mostly they just saw it as an economic bubble, as not impacting society. This one's different, isn't it?
Well, of course it is different, but it's got some things in common. I have a very strong memory. I opened Siberia Cafe in 1994, which was the world's first internet cafe where you could go in and use a computer when most people had never seen or understood what the internet was. And we were packed. There was a lot of clamoring for the internet. At the same time, the BBC one interviewed me, and the only topic they wanted to talk about was the dangers of internet porn and internet addiction.
I wonder why they were asking you? Yeah, weird. Who would have thought? You were addicted to internet. Well, I'm not suggesting you were addicted to internet porn, but that's different. I mean, there was that kind of skepticism. But compared to today, where, as you say, there's a kill bill switch going through Congress, even if it's probably fictional or promotional marketing tools by senators and congressmen. But I mean, this is a different kind of skepticism. I mean, If AI has enemies, they're very different from the enemies of the internet. It just tended to be analog types.
I think the difference is one of... Uh, drama. I mean, the, the, as you say, I mean, the, the, there is a great deal of hostility to AI. People want to close down data centers. People want to nationalize AI companies like Anthropic and OpenAI. There was none of that in the, in the early to mid nineties.
Yeah. Look, I think the, probably the dominant, the dominant continued, continued anxiety is all around the scale of change. What will it mean for me is probably the- The authors, Keith.
Right. And so what will it mean for me? The internet, you know, I remember people saying libraries were closed down. It didn't mean a lot for the average person. And over time, everyone started to use it.
Yeah. You know, Bitcoin only exists because of the internet. And, you know, if you'd have asked then, what does it mean for me? And the answer would have been the end of currencies. You would have got more anxiety.
Understandable in that sense. I mean, the idea of living in a world without currency is terrible. Troubling, worrying, especially if you're not quite sure what the alternative is.
So that's the scale of change. Now with the AI, the answer that is on everyone's lips is it might mean the end of jobs, which of course goes to the core of livelihood.
Well, the end, not just of jobs, but of man himself. I mean, as smart machines, if you talk about AGI replacing human intelligence or AGI replicating, competing with Human intelligence and then maybe acquiring its own intelligence. We're talking about a species-defining or perhaps killing event. We'll come to that later, Keith. But let me come back to your phrase. The enemies of AI are mostly good people. So I take your point there. Authors and other types worried about their livelihoods. You say mostly good people. Who aren't good? Who are the enemies of AI that aren't good people?
I think the AI companies themselves that are trying to get to be regulated by government in order to prevent especially open source competitors having access to a free market.
They're at the heart of it. I think there's a scale. I think Dario is totally trying to do that. You've got to remember he paid $1.5 billion this week as a fine, a settlement, not a fine, for having stolen a lot of books. I can't remember how many it was, something like,
But he's, and I've said this often, he's an entrepreneur. He's devious. He strategizes. And being regulated is part of a strategy. So he's probably the most sincere in wanting government regulation in order to prevent competition.
So you kick away. So you're arguing that he's not a good person because he's trying to kick away the Kick away the ladder and maintain the moat between him and everyone else.
I think they're relatively good on the narrow question of open source. They contribute to it. There isn't really a whole lot of good when it comes to, if you add the word Chinese to the start of open source, the big question of the week is, why is all the good open source AI coming from China? Where is the American open source AI?
Yeah, some skeptics or cynics might say it's because it gets funded by the Chinese state to undermine private enterprises. But that's just not factually true. So why, in your view, why the Chinese better? Why do they have more competitive open source models?
It's a fairly simple answer. They don't have access to Walls of Nvidia GPU. So they have to make do with lesser hardware for training, which means they're forced to use what's called inference, which is using someone else's AI asking lots of questions and learning from the answers. And that's a cheaper exercise. And so they're producing AIs that have leveraged
So it's ironic then that the CEO of Nvidia has articulated, at least in PR speaker, support for open source technology if the reason why the Americans are trailing the Chinese on open source are his chips.
Yeah, obviously, Nvidia's interest is in making sure that everyone can buy Nvidia chips. And for that to happen, there has to be a very big market in building AI models.
So who are the bad guys, then? One of your pieces, you I don't think you're a big fan of Paul Krugman, but you often have his pieces. He interviews one of your old friends, Lena Kahn. Are Krugman and Kahn proponents of nationalizing AI? Would they believe in this kill switch, do you think?
You know, they don't really talk about... Their topic is... By the way, it's an interview between the two of them, so it's kind of meandering. But if you wanted to extract from it, Khan is really...
She definitely has the persona in her own mind of being a good person. So her core is consumer protection. You know, protecting the average person from things that would happen to them if big companies are allowed to just do whatever they want. So, you know, on the face of it, you think, okay, well, that's reasonable. But in doing that, you end up having to ask the question, who makes the rules and who decides what is dangerous and what is not dangerous? And that's where it ends up going wrong, because the definition of danger that she supports is almost parental and treating consumers as children.
As the father of three boys, Keith, you're anti-parental. And you make it very clear that while you acknowledge lots of problems with AI and who doesn't, the answer is not nationalization. You don't want the Lena Kahn's of the world determining AI policy. Is that fair?
Absolutely fair. I think government can't really Understand how fast AI is changing. If government was to get AI today, three months from now, that AI would have been replaced by a better AI. And governments are just not in good place to lead product change and react to product change. The only thing they can really do is slow it down in their attempts to understand it. In order to then control it. And it isn't so much ideological, it's practical. It just wouldn't work. And so there's no point starting there as a remedy to anything. If you do believe that AI is potentially dangerous, probably the best thing that happened this week is when the OpenAI models hacked hugging face in the laboratory. When the rails were taken off of the OpenAI models, and they were given a task that required them to do that. And what you found is that OpenAI understood what was happening, stopped it, reported it much faster than any government would be able to. So the right point of control really is the companies themselves, and they should be held to a high standard But it's the people, they're the ones you're criticizing, the Darius and the Sam
For wanting to be regulated, and I think you've got to put that into the economic part of their brain, whereas what they did in the lab at OpenAI when this hack happened was the technical part of their brain. When it comes to technology, they don't want regulation because it's inefficient and Wouldn't work. When it comes to economics, protecting them from competition, they do want regulation. So that is a real contradiction.
So tech is better than economics. But even you acknowledge, or certainly some of your correspondents this week acknowledge that not everyone's bad. You have a piece by Matthew Inglesias, very influential writer. You know, I think the data center backlash is an American phenomenon, firstly. There's nowhere else in the world where this is happening, which is already a
the party of progress and the people is the one most against data centers is strange. I'd love you to tell me why you think that is, but it definitely is factually true that that's where it comes from.
Yeah, yeah. You know, I haven't tracked that, Andrew, but I believe you. It makes sense because politicians ultimately go where people go because they want to be popular. So they tend to embrace bottoms up policy instead of standing for something and trying to Get people to agree with it.
Yeah, I mean, the thing with the data center backlash, which is interesting, is it goes against, Cora, you talked about economic principles, that clearly for communities, having a data center should be in their interest. They get huge amounts of investment and many jobs, but there are many of them, or certainly there are activists within all these communities that are against it. So it's turning down economic upside. I mean, they might argue in the long term there's no upside.
bottoms-up kind of local resistance, which is often micropolitics. It's like, you know, we don't want that piece of land to become a data center because it will spoil the view type of reaction. He separates that from statewide bans, which are coming from politicians, and says they're not the same thing. I think that's kind of right. And who's doing the statewide bans? It's almost all democratic states, which is just bizarre to me. What do you think is going on there?
Well, I think we can learn from history. One of the other pieces you connect with this week is a book review by Noah Smith, an older book. It was a book review a couple of years ago, but he's reposting it. Power and Progress by Akamoglu and Simon Johnson. Simon Johnson was actually on my show a couple of years ago talking about this book. And it talks about this thousand-year struggle over technology and prosperity. And it seems as if Johnson and Akamoglu are arguing that there's good and bad technological progress. It needs to be controlled by, shall we say, the middle classes. What's your take? Another book, another influential book. Actually, this book helped, I think, Akamoglu and Johnson win the Nobel Prize.
Yeah. Yeah, look, it's an interesting essay because it fails to solve the problem it sets. But the problem it sets is that power over technology has always been centralised and governmental. But technology itself coming from entrepreneurs is decentralized and not at all governmental. And, you know, that leads to the question, how is technology controlled?
Which in a way is the wrong question. I mean, I don't think technology is ever controlled. I think at least good technology that benefits people tends to have this spontaneous growth in it, organic growth, if you will. Like me doing the book. Nobody told me to do the book. I just knew I could.
I can't click on it. But, you know, anyone can... I can put into Claude after this. I can feed this whole post and it can create 60,000 words. It doesn't really make it a book. Anyway.
Well, a book is something that people read, a book like Akamoglu and Johnson's Power and Progress or Popper's The Open Society and Its Enemies. Otherwise, I mean, I guess you can call it a book.
Well, what you're looking at is the website created from the book. There's also a Word document that is just a book. Whether it's printed or digitized, doesn't really matter. It's still a book.
But still, I mean, a book seems to me, I mean, defining a book, obviously, I mean, you have 60,000 words, but someone's got to read it to make it a book. I mean, you can author it. But once people start reading your 60,000 words about who owns intelligence, I might make it, I might acknowledge in my mind, at least that it's a book at the moment.
We will see. Let's talk about... Last week, we talked about Tim O'Reilly's argument that Musk... It's sort of anti-Smithian, Adam Smith, in the sense that he represents a kind of monarchical view of economics. And I think that that's the essence of the Akamoglu-Johnson book, Power and Progress, that middle-class innovation or innovation from below is a good thing, and they're very wary of the kind of Muskean monarchism. One of the other interesting pieces that you linked to this week It was by Rina Chandran about why state-owned AI won't solve inequality. We clearly have an inequality problem when you have trillionaires like Musk. What can, if state-owned AI won't solve inequality, which may be true, what can solve it then? What does she say?
Trying to grapple with AI. Her main effort is to try to separate historical ownership over industry with ownership of AI and make them distinct and separate. So she talks about Bernie Sanders' 50% ownership idea.
Yeah. And she also recognizes that globally, state ownership of technology is not unusual. The Norwegian ownership of the oil maybe is the biggest example. And it's led to the world's biggest sovereign wealth fund in Norway that every citizen benefits from individually. So there's clearly precedent. And she's not wrong to ask the question, if AI is going to replace workers, which the power and technology book says society has an option there. It can opt not to replace workers. I actually think that's not true. I think technology will force an outcome that It's just logical without bad intent, which ultimately leads to the end of most jobs. I think that's very likely. She takes that on board and says, well, therefore, there has to be a new social contract. And ownership has become the focal point for that because instead of tax, which has a bad rap, obviously, and is seen as anti-individual, ownership is pro-individual. And in a world where individuals become more and more important, especially in their own heads, ownership seems a fantastic alignment to future needs. And that's why I like the Sanders thing. I don't like his form of it, by the way. I just like the idea of ownership being at the core.
I mentioned that you linked with Noah Smith's reposting of his review of Akamoglu and Johnson. He also has an interesting piece this week, which you linked to, about whether we asked too much of the American University. I think it's not surprising the American University is in all sorts of crises. It's the... It's the heart of the revolution, isn't it? It's where AI will do its greatest damage or will have its most profound impact, if indeed this is an intelligence revolution, given that the American University was the place to train and, I guess, monetize intelligence.
Yes. It's interesting, you know, this week a bunch of universities, I don't know if I made this week's newsletter, it maybe end up being in next week's, but a bunch of universities have cancelled a contract for their AI detection tools because they were incapable of accurately detecting AI, the same as the Substack one, by the way. But they cancelled it because it wasn't accurate, not because they wished it would be accurate. And so at the core of the question is, what is learning and how we're learning?
By the way, I suspect you don't really believe what you're saying. So this is going to be insulting. Maybe an hour and a half. It's going to be insulting, but the real you doesn't believe it. So it's not really insulting. But the minute you doubt that a human being could get onto paper some very complex ideas over 80,000 words using AI without somehow undermining their credibility, that's wrong.
Well, let's take that example. Let's say that Karl Popper had OpenAI or Claude back in 1945. Could he have written, I don't know how long it took, I think he spent most of the Second World War writing the book, which is enormously influential. There are two volumes to it. Could he have written it in half an hour or an hour and a half?
It doesn't seem, that's like saying before the car, could somebody go from Manchester to London in two and a half hours? Well, we know the answer to that. But thank God someone invented the car and now they can. And thank God somebody invented AI, so you now could. I don't understand how it undermines the human agency. Surely it's the opposite. It gives human agency a path to be disseminated that wasn't available before.
So then let's come back to the Smith point, which in American universities, one of their remits, I guess, is training people to write books, which have always historically at least had great currency in our culture. What's the point of an American university in our AI age?
Yeah, look, we talked last week about bottling intelligence. What universities are doing is bottling learning or teaching. And they charge a lot for it. They're metering intelligence to those who can afford to pay for it, mainly in America. In the rest of the world, not so true. Universities are free in much of Europe, for example. So that business model seems to be past its sell-by date, I think. I don't think you can hire 50 professors, put them on a campus that costs a lot of money to build, charge people $60,000 to $80,000 a year to attend, give them a piece of paper at the end, and now they're educated in quotes. That will be probably the worst way to learn in the not too distant future because it's so inefficient and costly on both time, by the way, and money. It just doesn't seem to make sense.
I wonder if there's any coincidence that just as we live in an age where Congress at least is proposing an AI kill switch, whether it's, as I said, it seems to me just marketing fluff, but we will see. We also live in an age where the American state or the current MAGA state is interfering in universities, telling people what they can and can't teach, sometimes even what they can and can't read. Is this connected, do you think? Is it coincidental that in the age of AI, of all this uncertainty, the state is becoming more and more reactive? I mean, you're not alone in saying that government-owned AI is a terrible idea. I mean, I agree with you 100%.
wrote a piece saying government-owned AI is a terrible idea that you include. But what is it about the state? Are we falling back on it because it's our only certainty?
Well, I think the state, the modern state, is the outcome of democracy. And in a way, we should cling to it because it's what we collectively created to manage and organize society in all of its complex needs. In a way that individuals can't. And the legal system, going back to habeas corpus, the legal system provides a framework for getting things done. The law courts provide a framework for judging whether you did it the right way or not, and give everyone recourse.
There is a debate and discussion leading to laws and acts in the legislation. The executive branch executes all that. The state depicted properly is a fantastic outcome of modern times. And the nation state is, compared to anything that came before, fantastic. What's wrong is that it's become Both corrupt and flawed in various ways. It's really no longer an expression of the people. And therein lies what needs to be unpacked. What should a modern state look like? And then I think it gets even further complicated by globalization driven by technology, where things that pop up in America yesterday will pop up in, let's say, Singapore tomorrow. There's really no borders when it comes to technology. That leads to a globalizing policy need that is absent.
So what exactly are you calling for? One of the other pieces you connect with is from the San Francisco Standard about AI's new political donor class, which of course is made up of people like Dario and Sam and everybody else, Musk, is already outspending Big Tech's last one. Would you call for profound campaign finance laws so that this new AI plutocracy can't buy elections and politicians?
I haven't really turned my mind to that. I have the kernel of an idea, and the kernel of the idea is that smartphones and voting It means that voting can be almost instant and the results almost instant. Of course, not everyone has a smartphone. So if you don't have one, you should be able to go to the library and place a vote, but electronically. And I think that leads to democracy becoming deeper and richer for those who want to participate. It would be massively irritating if there's too many votes because most people don't want to spend their life thinking about The collective, they need their individual life as well. But it really comes down to that 16th, 17th, 18th century question about the relationship between the individual and the state gets reset on a new platform, which includes direct democracy as one of the options. I think the hyper-local democracy becomes possible as well, where The most important agency might not be the nation state, it might be your city.
I'm not going to tease you, Keith, about any new books, although I have to admit I'm slightly skeptical of plebiscitary democracy given Brexit, for example, but maybe that's the subject for another show or another book. Finally, it just
I didn't make that public. I've always argued that poor people are lucky to be around, that they certainly shouldn't have the vote. Finally, how much do you think you should earn, Keith, to get the vote in Palo Alto? A couple hundred grand?
Well, I'm just curious. So finally, my interview of the week, and it's a surprising one, given this is supposed to be a tech show. It was titled The Seduction of the American Mind by a young woman, Emily Akin, who's a book reviewer on the New York Times Book Review. She has a new book out called The Frenchman on My Life in Theory, which is an excellent autobiography of her experience with French Theory, guys like Michel Foucault. But she notes, and I think this is a good way to end this show, she ends by suggesting that whilst a lot of people have been very skeptical of post-structuralists or so-called post-structuralists or post-modernists like Michel Foucault, that he was actually right in a way. In his book, The Order of Things, which was written in 1966, so 20 years after The Open Society and Its Enemies. Another book, another influential book. He argues that... And I'm quoting him here. This is the last sentence in the whole book. He said, he's talking about man. Before the end of the 18th century, man did not exist any more than the potency of life, the fecundity of labor, or the historical density of language. He is quite a recent creature, while the demiurge of knowledge fabricated with its own hands less than 200 years ago. But he has grown old so quickly that this has become only too easy to imagine that he has been waiting for thousands of years in the darkness for that moment of illumination in which he would be finally known. And he talks about us disappearing. So I think Akin's argument in our conversation is that Foucault's prediction of the end of man in books like The Order of Things are actually being realized in our AI age, which is kind of ironic. I don't know whether you're familiar with Foucault.
No, I think the opposite's true. I do think postmodernism understood something that especially ideological Marxists didn't. Ideological Marxists tend to conceptualize in terms of the collective,
You know, modernism was all about social engineering at big scale, abstract. And that leads to Stalinist-like social construction. And the postmodernists understood the individual as it emerged in somewhat mundane ways. I mean, look at the Pantone color system that lets you choose your color scheme. Is intensely individual, but very satisfying for people who care about colors. So the rise of the individual is real. I actually think...
Yeah, weirdly Marx himself was very aware of the individual. He talked about the total human being, and he thought the individual was the point of history, the full individual, the true individual was the point of history. But ideological Marxist never really understood that, and so shame. The postmodernists did, but they didn't have any concept of how the individual lives in a society that enables them to be individual. So there was no social framework that they lived within. They only talked about the individual as if there was no context. And I think the future, and AI is definitely an accelerant here, is it lets you realize your individuality. Far from taking it away, my, in quotes, book is a realization of me. It really is. The chapters are mine. The arguments are mine. The putting it down in one place is me. The actual writing was not done by me, but it couldn't have been written without me. So it's a realization of my individual self, not an alienation of it. And I think the future is really all about the individual and society flourishing together.
Well, there you have it. Foucault believed maybe the end of man. Keith Teare believes that, and I'm not sure if Foucault was thinking about AI in particular, but Keith Teare believes that the AI revolution will represent the beginning of man or certainly his renaissance, his or her renaissance. So we will see, Keith. I'm sure this won't be the last time we talk about this. How many books are you going to write this week before our next show?
Oh, well, we will come back. This is supposed to be a family show, Keith, but we will talk again next week. Keep well, keep writing, keep thinking, and we will talk again next week about books, AI, and the end of man. Thank you so much. Bye.