Despite the years of self play OpenAI Five has gone through, humans still manage to find and successfully exploit its weaknesses. Adapting to opponents remains an unsolved meta-problem.
Focus is often on OpenAI Five's victories, but losses are what drive our progress. On Aug 9th, we were crushed in a best-of-3 against @teamsecret (#4 team in pro circuit)—but won game 1 as Secret adjusted to Five's playstyle. Last week lost to several non-pro teams. Tomorrow...?
1
1
7
Team Secret was just plain better than Five — even weak points were scenarios seen during self-play. Was really amazing to watch Secret in action.
Note that Five does not learn online — it's more like a creature of pure instinct. More like an insect than a human in that way.
Aug 21, 2018 · 8:59 PM UTC
2
2
20