,
3 tweets,
1 min read
Read on Twitter
Within a day of @OpenAI 's release of the Dota 2 bots, *eight* different human non-professional teams were able to beat them. It took them a number of games to learn how the bots worked, and devise strategies to counter them. Humans never fail to surprise us with their ingenuity.
#OpenAIFive has been defeated 26 times now. One team defeated them 4 times, and two others twice. A professional team will probably be able to defeat them consistently after some practice.
To me, this confirms the hypothesis that model-free RL doesn't scale to complex problems. Even after restricting the game to 17 heroes, and providing 45k years worth of data, model-free RL learns fragile one-trick policies which beat humans only due to mechanical superiority.
