The Problem With the Turing Test

I think that we've been stuck in the same loop for the last 10, 15 years because of our fixation on these benchmarks and things like that. In terms of generalization, I'm not sure that we've made a whole lot of progress with some of these things. But at the same time, there's been a lot of progress on the stuff that I care about in computational cognitive science or something that we call inverse planning. The input problem is you give me a desire and a belief and I need to generate action. That forward problem is the problem of robotics,. Like you are giving the robot a utility, a goal, and a belief, which we just formalize aslike a