Demis Hassabis on how to stop AI destroying everything

Your peers, they joke or fret about. building God. How does that sort of. language sit with you? No, I didn't like. it that that kind of language. I think. of this as a tool. That's how we're. trying to build it. Like a scientific. tool, like a telescope or a microscope. But I think that's what defines us as. humans is we're tool-making animals. And. uh I'm incredible. Like look at what we. built, modern civilization with our. general minds. Um and I think the other defining. feature of us is our insatiable. curiosity and our thirst for knowledge. And I think I'm just following in that. tradition, really, of of building ever.

more capable tools uh that allow us to. interrogate um uh the fabric of reality. And they are also ever more capable. tools that pose ever more serious. threats. Are you still worried about. destroying everything? There's. definitely um you know, I would say the. way I call it is there's a non-zero. chance that things could go quite badly. wrong if then if we if the technology's. not built in the right way. I'm very. more optimistic gi- given enough time. and enough care and focus, um I'm a big believer in human.

ingenuity. If the best minds work on. this, obviously not just us, but our. colleagues from from the other. companies, um we will build this in the. right way and in a safe way. I think. that's very possible to do. Uh and then. we'll get all the benefits. You know, curing diseases, helping with energy, all the other big challenges that face. um society today. And so I think AI can. be one of those. uh uh technologies that helps us, kind. of meta-technology that helps us with. all of the solve all the other. challenges that we have. So I think the.

the potential's enormous, um but they. but the risks are there, too. So I'm. optimistic that these things can be. solved and they will be, but um but. we've got to be cautious and use the. sort of precautionary principle here as. we approach um this AGI moment, which. will be, you know, one of the most. transformative moments in history. How. do you act in a cautious way when you're. also locked in a you know, competitive. pressure with other companies, with. other polities? There's multiple race. conditions going on, obviously between.

the companies, but also between nations. So it's it's and and and of course the. geopolitics of the world is more. fragmented than ever. So it seems hard. to see how you're going to get um that. international cooperation, but I think. we should still try. And what I'm. hopeful of is that as we get closer to. AGI, maybe it's only, you know, 5 years. away or something, um it'll become more. obvious to people, just as it has in the. last couple of years, how momentous AI's. going to be. Something I've known for 20. years, but I think it's becoming more. obvious to everyone. Also what the. attendant risks could be. And then I.

think at least, I hope, there'll be um. some sort of minimum standards or. something like that you could think of, which can be agreed on by at least the. leading labs. Uh and that will be. something to build off on top of.