suggest use katago games to increase self play games

[l1t1]

as contributions are too few

[Vandertic]

Sorry, but there would be no much purpose in that approach. We do not want to reproduce a weaker version of KataGo. We want to make a different experiment altogether, in particular by sticking very hard to the "zero" philosophy. KataGo works perfectly and is much stronger than SAI, so there is basically no need for our experiment to become stronger quickly. Instead, we will focus on seeing if our approach will eventually give different answers from other AIs here and there. Will SAI find different josekis? Different fuseki strategies? Will it be able to play the perfect endgame? Will it be able to teach us new principles?

Every AI trained alone is a new and different view on the world of go, and every one might find new stuff or help understanding the game better.

Anyhow, we are working under the hood and performing experiments we are not yet talking about. We hope we will be able to show some new stuff and maybe some stronger nets in the future. We have not abandoned SAI and we do not intend to do so.

Thank you as always for your interest and your support!

[l1t1]

Thank you as always and look forward your new stuff.