EP02 — Has Europe Still a Chance in the Global AI Race? — Stuart Russell

We are right now in the global AI race, and you called it once a prisoner's dilemma.

And maybe you can explain that quickly, what you meant by that, and also then tell us where actually Europe stands in this race right now and in this prisoner's dilemma.

So prisoner's dilemma

is

a... is a simple construct in game theory, which is a branch of economics.

And it's used to describe all kinds of situations where two players could cooperate and be better off, but each of them has an advantage to defect from that cooperation.

And so what happens is they both end up defecting, and then they're both worse off.

So it's called The Prisoner's Dilemma because it imagines two prisoners who are separately being interrogated by the police.

And each of them could refuse to cooperate with the police.

And then the police wouldn't have any evidence, and then both of them would get off with a light sentence.

But each of them also says, well, I could spill the beans, I could implicate my accomplice, and then the police would reward me with probation instead of a prison sentence.

And so each of them spills the beans on each other, and then each of them goes to prison for 10 years.

And so that...

That simple idea happens in arms races as well, right? So each side could say, well, why would we want to build up an enormous nuclear arsenal, which will probably end up getting used and then we'll all be dead, right? So that's the bad outcome for both parties.

But each of them, of course, is afraid that the other one is going to cheat or the other one is going to get ahead and then they'll be at a disadvantage.

So each side individually decides that they need to build up their nuclear arsenal, even though they know that if both of them do that, they're going to end up in a terrible situation.

So that's how arms races work.

And with AI, the arms race obviously is between the US and China, and to some extent also between the companies.

Each company is afraid that the other company will get ahead, and so they ignore safety.

They ignore the fact that we don't know how to control AI systems that are more intelligent than human beings.

So they're building or trying to build and spending trillions of dollars, literally, to build AI systems that are more intelligent than human beings, knowing full well that if they succeed, that could easily result in the extinction of the human race.

And they're

quite explicit

about that.

So yeah, I would say the lead changes.