In this article, I discuss AlphaZero, an Artificial Intelligence (A.I.) system that learns from experience, without shortcuts or solutions given by others. I emphasize AlphaZero's ability to create, starting from its experience, temporary and relative evaluations, renouncing the seduction of absolute truth. There is never certainty but an empirical evaluation.

In this respect, AlphaZero is similar to the kind of meditation that I practice: Recollective Awareness Meditation. This form of meditation is not based on a set of a priori assumptions or dogmas about what is supposed to happen during meditation. Instead, in meditation, thoughts, emotions and mental associations have full freedom of expression. Some suggestions or intuitions which emerge during the session recur outside the meditation, provoking other insights, or more or less rational re-readings in relation to what is happening in that moment. Following the meditation session, the material can spontaneously recur and be further reworked in the process of journaling, which often leads to going deeper into the material.

Recollective Awareness Meditation has a unique kind of relationship with dharma. This relationship is not based on a passive reception of a doctrine imparted by others but a process of experimentation, which involves doubt and uncertainty. For this reason, Recollective Awareness Meditation is very compatible with secular Buddhism, which also emphasizes the need for creativity, flexibility, and the recognition of uncertainty.

These qualities of the mind are essential in responding to the challenges  and potential of A.I.  There is no doubt that A.I. has concerning aspects, particularly its potential for dehumanization. However, A.I. can also show human beings how to be more aware of their own distortions, including when they are called to make decisions that affect others. In this regard I directly refer to a splendid book I read recently, Noise: A Flaw in Human Judgment, by Kahneman, Sibony and Sunstein. The authors illustrate how the choices of judges, doctors and insurers, among others, are profoundly affected by unperceived background noise, and how algorithms can help correct these distortions.

The value of ‘space’ and freedom in chess and the dharma

Recently I picked up some books about chess after a long gap. Among them was an excellent book by Davorin Kuljašević, Beyond Material. When I started studying it a couple of years ago, I was struck by the similarity between the description of the function of space in chess given in this book and the description that Stephen Batchelor gives of the same concept with respect to dharma in one of his books (Living with the Devil perhaps).

In both, space is needed so that a person is able to move freely, without stumbling. Kuljašević  gives many examples in which the attachment to the material (the pre­ established and absolute value given to the pieces) is sacrificed to obtain a more important result: space, activity (of the pieces), time. These three words, which might seem abstract, become very concrete, both on the board and in meditation. The deepest moments of meditative absorption always have a particular spatial dimension.

In the preface of this book Kuljašević refers to AlphaZero: the first program based exclusively on A.I. that plays chess. In particular, he refers to the similarities between the contents of his book and the comments on AlphaZero's incredible playing style. In fact, AlphaZero's style is so free from prejudices linked to ways of understanding the dogmas of chess, including attachment to the material, that it seems like something otherworldly while at the same time provocative.

AlphaZero: a ‘Game Changer’

So, I picked up this another book again: Game changer: AlphaZero's groundbreaking chess strategies and the promise of Al, by Matthew Sadler and Natasha Regan, and in recent days watched some games that AlphaZero played against Stockfish 8. The latter is a classic chess program, at the time of the challenge the strongest chess program in the world. It was constructed using a complex combination of computation and instructions received from humans on how to evaluate a position that might occur on the chessboard. Stockfish thus brings with it rules established by others, exactly like a meditator who follows a meditation technique based on instructions given by others.

How does AlphaZero decide its moves? AlphaZero only knows the rules of the game. It has no instructions, no rules that have been inputted from outside. Before meeting Stockfish 8, the software spent what in human terms could be described as a series of existences, playing against itself, isolated from the world, as if in some kind of retreat. Everything it knows it has learned from experience, from its own mistakes and its own successes. This 'meditator' does not have any meditation technique. It only has the initial reality in front of him, and its process of knowledge starts from there, based exclusively on its own experience.

The software built the values that guide it by choosing them. Instead of being numerical (absolute), they are statistical (relative). Above all, they are values that are constantly called into question in one's own experience. Stockfish 8, on the other hand, is forced to follow the instructions it has received; it cannot question them.

When one compares the computing capabilities of Stockfish 8 with those of AlphaZero, the former is much more powerful than the latter. Its strictly cognitive capabilities, if we can define them that way, are decidedly superior. To give you an idea of the gaming strength of these programs, it must be said that no human being is capable of even challenging them. A world champion is less likely to beat one of these programs than I - an amateur player - would be to beat the world champion.

After rewatching some of these matches recently, the impression that these games make on a chess player cannot be defined in words. It is better to use a metaphor: I imagine myself as a music lover in the year 1720, waiting to listen to a Vivaldi concert, only to find myself listening instead to Stravinsky. The shock is almost disturbing. In one of these games AlphaZero repeatedly sacrifices pieces, without humanly being able to hope that all this will be able to return in the future.

However, we notice that its game becomes more and more airy: its pieces are increasingly free to act. Stockfish 8 confidently follows the rules it received, but gradually and inexorably its pieces have less and less space, until it reaches definitive paralysis.

Stockfish 8 is crushed. The impression I had looking at the chessboard is similar to what I might have had by looking at a work of art. There is no point objecting that the game is played by a machine: this is undoubtedly pure genius.

In another game, something tragic happens. Stockfish 8 continues to accept the material offered by its opponent. Its bold optimism seems tinged with human flaws: greed, conceit. The moment comes when Stockfish 8 realises the abyss in front of it. Despite its enormous calculation skills, it misses the catastrophe that is about to befall its king.

To avoid checkmate, it must give up more than what it has earned up to that moment. The reasons for this turnabout cannot be seen on the board; they can only be deduced from the analysis of what was never played. Stockfish 8 will lose that game as well, ironically because of the excessive amount of pieces it had to give up, without compensation.

Lessons for the dharma and life

What I can learn from these games as a chess player adds to what they offer me as a starting point for an analogy with meditation, with the process of self-knowledge and with the tools used to achieve it. I see myself at the chessboard and feel resistance to going against a preconceived principle.

On the other hand, there is intuition, the result of a view of the position read through a more genuine and personal experience. I see myself sitting, meditating, and feeling the resistance to abandoning an attitude I always believed to be right. However, there I feel an intuition - I can do without a constraint that I place on myself and can trust my intuition.

In both cases, there is no certainty that intuition will lead to the desired result. In fantasy, I transform the challenge between AlphaZero and Stockfish 8 into a dialectical confrontation between two dharma scholars: Stockfish 8 is the head of an esteemed and recognised school, it has incomparable erudition and superfine logical abilities. AlphaZero is a nobody, its knowledge is the result of purely direct experience. Stockfish 8 expounds a series of theories and concepts, all recognised and approved by tradition. AlphaZero highlights its relativity and offers the possibility of understanding its limits through direct experience. The winner is AlphaZero.

In reality, AlphaZero is the winner in this challenge with Stockfish 8 and becomes the strongest chess program that exists in the world  despite never having received any instructions on how to play, and despite the fact that its calculation capabilities are inferior to those of its opponent. Its strength lies in its experience, that is, in its constantly reconsidered past. It appears to have achieved some kind of awakening, but perhaps this is just an illusion given that, as the only example of a chess player, whether human or artificial, it has never been subjected to any form of indoctrination. For this reason, it never had to unlearn how to play chess.

Pondering the title of Jason Siff 's book, Unlearning Meditation, I think about the knowledge and freedom that I progressively gain as I leave behind or reconsider, from time to time, ideas, goals, and expectations from my past.

---

Stefano La Fontana lives in Orsenigo in the north of Italy. He is an educator, psychotherapist, jazz pianist and composer, and a chess player who is interested in Buddhism and Sufism. Stefano is a recollective awareness meditation practitioner.