
Anca Dragan on why Asimov's three laws of robotics need updating
The Robot Brains Podcast
00:00
Can We Detect That We're Seeing an Input?
Ri: Can we somehow detect that we're seeing an input where that's, that's out of this, ah, out of tion ilets, that's different? We're correlations that we've seen no longerhump a. And yes, what wat what i'd be very excited to see happening. That's the holy grail, is doing this for the reward function specification itself. Ri just kind of looking at the reward function, the objective, and saying, i think it make sense for these contexts because you've verified it and clearly thought about that.
Transcript
Play full episode