The approach to making a human-like AI for games such as chess involves collecting human games, supervising learning, and setting regularization parameters correctly. By doing so, researchers can construct AI systems that play like humans but in a very strong style, which can lead to a fun gaming opponent for humans.