I recently watched AlphaGo, a documentary on the recent defeat by the best human player of Go by a computer, AlphaGo.
The commentary the beginning points out this was ten years ahead of schedule (which of course doesn’t take any account of the actual history of predictions by computer science about when computers would begin to beat human players. The first prediction being on the order of a decade perhaps sixty years ago now).
Another commentator noted that after human players had z chance to study AlphaGo’s playing stylevtheyvexoected the human performance to improve against AlphaGo. This will take some time.
Another way of gauging AlphaGo’s performance, which simply isn’t suitable, for humans, is to look at how it performs once the board is expanded say from 19 X 19 to 31 X 31 and then higher, 51 X 51 and to see whether it’s performance stays steady – or degrades.
A further way is by expanding the dimension of the game, from a 2d board game; to a 3D board game and higher.
Q. Have these experiments been carried out?
A notable feature of Go is that it’s very game is such that these changes are easily imaginable – and perhaps not imaginable by AlphaGo.