Moxzi Alpha 9 brings GPU learning and persistent local actors to Motoko 2.0

An actor already has an identity, private state, and a way to talk to other actors. What happens when part of that state is a model it can train?