Full threadwetmore·Some more discussion on the difficulty of creating reward functions: https://www.alexirpan.com/2018/02/14/rl-hard.html#reinforcem...View on HN