I was interested by what is defined as more complex when it comes to DeepMind learning games.
Googles Ai has already mastered Chess, Go and poker. In February it was set the challenge of Hanabi. This is difficult because of the co-operative element. Poker is bluffing but still solitary, Hanabi has a hints system.
Geoff Engelstein is much more eloquent than I could ever be, and his “Ludology” podcast
episode on this was excellent!
Otherwise here is an article:
And here is the full paper: