In reply to @patio11

@patio11 @robinhanson That’s not RSI though. That’s SI. The implementation of a new function giving the agent more capability does not itself improve the ability to add further capabilities. RSI would be if it could, without human input, train itself on its own games to improve future runs.