He Led Safety Teams at OpenAI and Google DeepMind. He Puts Our Odds at a Coin Flip.

The chains of human extinction is about a coin flip in my view about a half and I can talk about why that is. The trajectory of super intelligence sort of leads in the direction of the machines control the world and the economy and and such. I don't know whether this is fixable. I think given enough time I'm

pretty optimistic that it is. If you look back from a 100 years in the future and say could we have done a thing that would have caused it to go well I think the answer will be yes. My name is Jeffrey Irving. I've worked at Google Brain, OpenAI, DeepMind, the UK AI Security Institute, and a new

nonprofit, Resolution. I led safety teams at OpenAI and DeepMind, and it was the chief scientist at the UK AI Security Institute and Resolution. I think the most important thing for the public to understand here about AI is that we're moving very rapidly towards potentially superhuman systems. The systems that are better than humans,

every human at a lot of very riskre the capabilities that could mean if they want to cause harm, we could all die. So we are on a very rapid uh trajectory currently where AI capabilities are improving very quickly. If this continues, I think it's quite likely we get to super intelligent models either

in a few years or 10 years uh potentially one or two years. Um and at the extreme this can mean that they will have the capabilities and then possibly the desire to kill everyone. Um and that is on top of other risks and related to other risks such as such as job loss um

and kind of nearer term kind of bad outcomes. So I think fairly soon if we get super intelligence they will be able to do everything humans can do better than humans all humans that includes initially sort of cognitive tasks then it will broaden once we have sufficient robots to all physical tasks as well. At