Demis Hassabis knows AI’s risks. He’s building it anyway

谷歌最强AI大脑

Insider

2026-04-08

27 分钟
PDF

单集简介 ...

Demis Hassabis worries that artificial intelligence could be catastrophic for humanity. He also runs one of the world’s leading AI labs, Google DeepMind, pushing to build smarter, faster and more powerful systems as quickly as possible. How can he do both? Alex Hern, The Economist’s AI writer, visits the London headquarters of Google DeepMind to put this question to its boss. They discuss his journey from chess prodigy to AI pioneer. What breakthroughs does Mr Hassabis hope lie ahead? And does he fear the destruction the technology may cause?
更多

单集文稿 ...

  • Demis Hassabis doesn't like the limelight as much as some of his AI tech boss contemporaries.

  • But he wields greater resources and power than many of them.

  • 10 years ago, AlphaGo brought him to the public's attention.

  • His technology had managed to beat a world champion at Go, an incredibly complex strategy board game.

  • Then came a Nobel Prize for chemistry.

  • After another of his AI models, AlphaFold, predicted the structure of almost every known protein.

  • But did these achievements put DeepMind at the forefront of AI or distract it from the real opportunities?

  • And was he ready for what he has helped unleash?

  • I sat down with him at DeepMind's office to explore those questions.

  • Demis Hassabis, welcome to Inside Tech.

  • Thanks for having me.

  • So it's the 10-year anniversary of AlphaGo, the first AI to beat a world champion at Go,

  • which means it's about 10 years since I first met you in your offices

  • down the road here in London.

  • And it weighed you threw down the gauntlet and you said,

  • I think at the time you actually said 50% chance of winning, you told me.

  • So you stormed that.

  • You'd just shown that machine learning could play Atari games,

  • and you know, you were setting this challenge up.

  • Why was Go so important to you?