Neither Deep Blue nor the superior chess playing engines running on PC hardware that came afterwards used machine learning. They relied on basic heuristics as humans do to numerically evaluate a position (material, activity, etc.) with the advantage that they could evaluate many more possible move sequences than a human. They also had opening books to avoid losing games from the outset and endgame tablebases to identify forced wins/draws with a small number of pieces on the board. Only recently did Google come out with a chess playing program that is actually ML-based, and it beat the top rated 'traditional' chess engine.

My point being, despite chess being considered a game that requires deep thinking, the use of actual AI in chess is very very new. As far as I'm aware, Kasparov had nothing to do with it let alone a deep understanding of it. He wasn't even involved in the earlier development of computer chess playing programs as they rose to the grandmaster level and eventually beyond his own level (super grandmaster). He along with many others had confidently predicted that machines would never beat humans in chess. So yes, I'm quite reluctant to believe that he has any kind of vision on this topic.