this currently only uses intrinsic reinforcement, no external learning yet