Reinforcement Learning: Recent Progress and Future Problems