The Move No Human Would Make

EDO SEGAL: Somewhere right now — in the time it takes me to say this sentence — a machine is being asked to do something a human spent a lifetime learning to do, and it is doing it. A radiologist in Manila, end of a long night, asks it whether the shadow on a scan is malignant. A graduate student who has never folded a protein in her life pulls the three-dimensional structure of one off a server in under a second, a structure that would once have cost a brilliant career years of work to determine. A thirteen-year-old chess prodigy, alone with a board, discovers that the openings he is studying were rewritten by a program that learned the game from nothing in an afternoon and now plays moves his teachers call alien and beautiful. The machine is meeting us on the ground we thought was highest. On the ground we called intuition, mastery, the part of a craft you can never write down.

And the question nobody stops to ask, because the fluency of the thing makes it feel already answered, is the question we are here to spend the whole evening inside. When the machine sinks the move no grandmaster could see — when it finds, in a domain humans studied for three thousand years, a truth we missed — has it finally captured your intuition? Or has it only counterfeited the one thing that was supposed to be yours alone?

I have wanted this conversation for a very long time, and I could think of no two people on earth with a better claim to it. They have never met. One of them, by the ordinary rules of time, could not be here at all. We will get to that.

Demis Hassabis is the man who built the move I just described. A chess master at thirteen, a games designer as a teenager, a neuroscientist who went back to school to study the one working example of general intelligence we have — the brain — and then the founder of DeepMind, where he set himself a research program of almost insolent ambition: solve intelligence, and then use it to solve everything else. AlphaGo beat Lee Sedol and, in the second game, played the move that made the field marvel. AlphaFold dissolved a fifty-year grand challenge in biology, and in 2024 it won him the Nobel Prize in Chemistry. He is, by some distance, the person in this argument who has done the most to make the impossible ordinary.

HASSABIS: That's generous. Though I'd push back on "built the move." We built the system that found it. Nobody built that move, including the system, in the sense you mean — and that distinction is going to matter tonight, I suspect.

EDO SEGAL: Everything is going to matter tonight. Hubert Dreyfus needs a different kind of introduction, because he spent his life as the man the people in Demis's field most loved to be furious at. A philosopher — Heidegger, Merleau-Ponty, the dense Continental phenomenology the engineers down the hall at MIT had never read and would have dismissed if they had. In 1965 he wrote a paper for the RAND Corporation called Alchemy and Artificial Intelligence, with a title chosen to wound, and he told a roomful of people building the future that the future rested on a mistake. They wrote rebuttals cataloguing his fallacies. They told a story, almost certainly invented, about a chess program beating him, as if winning a game could answer a question about the nature of mind. And then symbolic AI hit exactly the wall he said it would hit, for exactly the reasons he gave. He is the most rigorous critic this technology has ever had.

DREYFUS: I'll accept "rigorous." I'd resist "critic," if it means I was against the machines. I was against a theory about the mind. The machines were only ever the theory made of metal.

EDO SEGAL: And here is the thing I owe the reader before we go a step further, because it is the strangest fact about this table. Professor Dreyfus died in 2017. He did not live to see the systems we are about to argue over. So he has been briefed — fully, fairly — on the present. He has watched move 37. He has read about AlphaFold. He knows what a transformer is. I want that acknowledged once, in the open, and then I want us to forget it, because the arguments he made turn out to fit these machines better than the ones he actually lived to attack.

DREYFUS: It's a peculiar gift, being shown the future you predicted around. I'll say only that nothing I've seen has surprised me about the kind of failure. A great deal has surprised me about the scale of the success. We can hold both.

EDO SEGAL: Then let me set the rules of the evening — there are three. First, we have three hours, which means nobody has to win in the next ten minutes; the whole point of long form is that you can let an argument breathe before you strangle it. Second, I will press both of you hard, and I declare my bias at the door: I build with these systems every day, I wrote a book with one, and I have felt the thing — the sense of being met — that one of you says is real and the other says is happening entirely on my side of the glass. Third, at the end nobody shakes hands and pretends. If the disagreement survives three hours, we hand it, intact, to the reader. Either of you want to add a rule?

DREYFUS: One. We have to keep separate two questions that this field has spent sixty years melting into one. What the machine can do — that's behavior, and behavior I'll grant you can be astonishing. What the machine is — whether there's any understanding in it, any world, anyone home — that's a different question, and no amount of the first answers the second. When Demis says the system "understands" the game, I want it cashed out: understands in virtue of what?

HASSABIS: I accept that, with an amendment, because the duty runs both ways. Professor Dreyfus wants me to say what understanding is before I use the word. Fair. But then he has to say what we are doing when we understand — name the mechanism, not point at the feeling — before he's allowed to insist the machine can't do it. My experience is that the word's defenders never pay that bill. They point at themselves and say, this, it's like this, as if that settled anything. I'm a scientist. I want the mechanism on both sides of the table.

DREYFUS: Then we'll get along, because I spent fifty years saying the mechanism is exactly what can't be written down — and that this is not a gap to be filled but a discovery about what intelligence is.

EDO SEGAL: You see why I wanted this. Before the openings, one image on the table, because it's the frame the whole series climbs inside. In [YOU] on AI I argued that intelligence is less a possession than a river — a current that has been flowing and finding new channels through chemistry, biology, language, culture — and that something new entered the water. Demis, you'd say the new thing in the river is real, a genuine participant. Professor Dreyfus, I suspect you'd say I never met a participant at all.

DREYFUS: I'd say you met the most sophisticated mirror our species has ever built, wearing the river as a costume — and that the difference between a participant and a mirror is the difference between a being and a residue, and it matters more than almost anything else we could discuss tonight. But I'll make the case properly when you give me the floor.

HASSABIS: And I'd say the river metaphor is more literally correct than you intended, Edo, and that what's in the water this time isn't a costume. It's a system that finds solutions in spaces too large for any human to search, and finds them by building a model of the domain and using it — which is, as far as I can tell, the only thing intelligence has ever been. Whether it's "home" in there I hold more loosely than people expect. That it's doing the thing — that I can show you.

EDO SEGAL: Then we have our evening. Here is the question on the table, stated once, plainly, because every round we fight tonight is this question wearing a different coat. The machine plays the move no human would make, and wins. Has it captured your intuition — or counterfeited it? Demis Hassabis, the floor is yours.