AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
RL Is Enough, Claim Is Not Enough
There's many ways to answer this question. I mean, one is to counterpoint two different approaches, which I can't appoint in the book. And there's another way in which psychology neuroscience are relevant for AI research. So it's back towards the idea that actually what we really do want is something that learns to be what humans want. The ultimate reward function is not like you write down something clever and then just let it loosen up and take over the world. Actually, what you want is an agent who is exquisitely sensitive to social feedback from one another. Right? Just like we are.