Reinforcement learning for stabilizing an inverted pendulum naturally leads to intermittent feedback control as in human quiet standing.