Most likely, Zahavy believed, if a chess application could in some way have access to all These distinct versions of AlphaZero, properly trained on These different positions, then that range could spark a chance to method new issues productively. Perhaps it could generalize, in other words, fixing not just the Penrose puzzles, but any broader chess