“You can construct reinforcement learning environments where you say, okay, given that body of experimental evidence, and given the choices, what was the next choice that the scientists made”

“Pre-registering experimental predictions by using for example Fatebook helps me notice when I should be surprised and update my model of the world.”

“It was about three, I mean, it was my whole life, and then about three full time years of just experimental work in the studio, doing things like the 100 Days Project”

“The agent ... create a new category and then file it ... hold it in a holding category ... you have to stop, hold the information, and then in the next chat, ask me what I want to do about it”

“… very good RL controllers that are probably trained from motion imitation, and then fine-tuned with good RL that suits the robot's body… for the specific task of running really fast.”