Good idea. So you’re claiming that some sort of “randomized Solomonoff induction” can win all games, including uncomputable ones, that have bounded payoffs per step (and possibly all other games too) against all computable agents who are also allowed to randomize? If true, this looks like a new and important result. We need to pin it down formally.
Good idea. So you’re claiming that some sort of “randomized Solomonoff induction” can win all games, including uncomputable ones, that have bounded payoffs per step (and possibly all other games too) against all computable agents who are also allowed to randomize? If true, this looks like a new and important result. We need to pin it down formally.