Playgrounds for the ideas I work on. Click around. Break them. They're cheap.
A 1D Gaussian process learns an unknown objective. Three modes — manual, guided, autonomous — each a different amount of trust given to the agent proposing the next experiment. The agent narrates every move.