First of all, Go is not 'solved'. While the AI plays better than the best humans, it does not play perfectly (nor will any AI program in the forseeable future).

First of all, Go is not 'solved'. While the AI plays better than the best humans, it does not play perfectly (nor will any AI program in the forseeable future). If by 'solving' you mean, developing a strong AI program to play chess, I think the reason they haven't done it because chess is easier to solve than Go. The best computer chess engines are already a lot better than the best humans. Thus, there is not much to gain for Google by finding a new approach. There have been attempts to use similar approaches for chess, but they al seemed to perform worse than the traditional approach for chess. (Alpha-beta search with a hand-written evaluation function and null-move pruning). The differences between Go and Chess that might explain this are: - Chess has a smaller branching factor (the number of legal moves is around 40 in a typical position in chess, whereas it's over 100 in a typical position for Go) - A chess position is easier to evaluate. Simply counting the material already gives a decent evaluation in a lot of cases. This supplemented by hand-written rules gives a solid evaluation of a position - The average game of chess is shorter. A game of chess lasts between 20 and 80 moves, whereas a game of Go lasts around 200 games. This makes it a lot more feasible to extend the search far enough to oversee the consequences of each move in chess when compared to Go