#49 - AGI: Could The End Be Nigh? (With Rosie Campbell)
Increments
00:00
The Risk of Using AIs to Cause Harm
AIs have the capability to do these things when directed by a motivated human actor. But there's no obvious solution as to why how we can guarantee that a system wouldn't develop incorrect objectives or deceptive alignment. This has been shown in toy examples already as a problem. So I honestly can't tell what you're saying is science fiction or if it's totally out," he said. "I think most people probably couldn't visit your world where motivated motivated to cause harm"
Transcript
Play full episode
