I would also be interested to know how it would perform if you loaded the game with new coloring, or slightly changed the graphics of the units in a way that would still be easily distinguishable for human players.
-
-
-
My understanding is that AlphaStar does not use computer vision, but interacts with tables of data that represent what is visible.
- 3 more replies
New conversation -
-
-
Re: the third point, having played pretty much solely Terran I can tell you that as a human, my transfer to other "races" is around nil.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
From the DeepMind Blog post: "Even with a strong self-play system and a diverse league of main and exploiter agents, there would be almost no chance of a system developing successful strategies in such a complex environment without some prior knowledge."
-
"Learning human strategies, and ensuring that the agents keep exploring those strategies throughout self-play, was key to unlocking AlphaStar’s performance. To do this, we used imitation learning..."
- 2 more replies
New conversation -
-
-
Another question I would be interested in is how fast would it be able to adapt to changes to unit strength or the complete removal of certain unit types. Could it quickly find alternative strategies to compensate?
-
Bravo. That's the question that sums up every other question.
End of conversation
New conversation -
-
-
Hey
@DayneRathbone , could we set Gary up with John to flesh this out on http://www.letter.wiki ? I'd be keen to see the answers. - End of conversation
New conversation -
-
-
Not very convinced on the viability of reinforcement learning. Takes up too much training and power to make it work in a very constrained environment. Not sure if brute forcing is the right way to do it. The amalgamation of supervised, unsupervised and symbolism is the way to go.
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.