The rules of the major games that challenge Artificial Intelligence (AI) algorithms, such as Chess and Go, were created before the computers themselves, thinking about people’s competitions. Actually, they were created for a reality where the coldness and the brute force of the machines were not a competitive differential.
In addition, the total number of possible moves were created to allow matches to be played in relatively short times, due to the mental effort required, or even to enable quick matches. In this way, a match with experienced players can typically last 50 moves in Chess and 200 moves in Go. So even though AI developers for these games consider the high complexity of the decision tree involved, the fact is that the limited number of total moves makes it all easier these days, for systems with high computing potential like big IT players and with brute-force capability to analyze millions of total games in fractions of seconds.
Therefore, I believe that the moment we use games with old rules to measure the effectiveness of the state of the art of AI as a benchmark of cognitive potential comparison, our measurement is not right. Also, I think it is best to establish new rules that filter out exactly emotional issues and brute force, and at the same time really make the problem of solving these highly complex games more and more difficult, bringing competition to a new level of challenge, where we will really measure the evolution of machines that seek to think, as humans.
But as change rules, even older ones, is not easy, my proposal is to change only one rule, at least for the famous Chess and Go games, as a new golden rule for AI evaluation.
And this new rule I propose is very simple: to determine the number of moves per match where it would be possible for a player to redo a previous piece move after the last move of his opponent, i.e., to return a movement that considered worse than the probable line of reasoning of the adversary, or yet, simply was a typical mistake by emotional questions, only realizing later the gravity of its error.
Undoubtedly, this would be a controversial change, and to be honest totally improbable, given the history and tradition of these games. Be that as it may, here is a way to seek to balance the dispute between machine and man in the main points presented previously, besides allowing to increase infinitely the total number of moves per game, so that the real state of the art of technology can be evaluated of AI in this area, today and in the future, making a new story in this fascinating dispute.
By Rogerio Figurelli at 05/29/2020