在游戏中卡壳时,你是会继续反复地撞南墙,还是转而探索游戏地图的其它方向?也许是受古装戏影响,我们可能会四处寻找密道开关。但当前的人工智能可能会选择前一种,不达目的不罢休。加州大学伯克利分校最近发布的论文,则想要开阔人工智能的视野,让它在卡壳时退一步,转而探索更广阔的天地,对视野外的世界产生兴趣。论文被8月的国际机器学习大会接收。刚刚退役的Alpha Go,使用蒙特卡罗树搜索来决定下一 When stuck in the game, will you continue to repeatedly hit the south wall, or turn to explore the other directions of the game map? Perhaps by costume drama, we may be looking for the secret switch. However, the current artificial intelligence may choose the former, do not reach their goals without giving up. The recently published paper by the University of California, Berkeley, wants to broaden the AI ​​horizon so that it takes a step back when it is stuck and instead explores a broader world of interest to the world outside of the field of view. Papers were taken at the International Machine Learning Conference in August. Alpha Go just retired, using Monte Carlo search to determine the next one