"When the game is played with inhuman speed and accuracy, abusing superior control is very likely to be the best and most effective ... way to play the game." Looks like a clear confounder here in DeepMind's latest game-playing algorithm:https://medium.com/@aleksipietikinen/an-analysis-on-how-deepminds-starcraft-2-ai-s-superhuman-speed-could-be-a-band-aid-fix-for-the-1702fb8344d6 …
-
To me it appears that beating a video game with deep RL only shows that with enough examples you can largely reverse engineer the code running it, like a sort of "messy" inductive programming. I don't see what you learn from that about reality and human intelligence.
as @dileeplearning’s DQN demos show, this is actually too generous. the contingencies for a given level have been hacked, but the reverse-engineering is so fragile it breaks if you change the details of the game. what’s been induced is a long way from a general program
-
-
Agree. Such systems 'solve' decision making by dense coverage of possible scenarios using millions of iterations. For general intelligence we need ad hoc learning and problem solving facing never seen before situations
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.