Reward is not enough: can we liberate AI from the reinforcement learning paradigm?