Goroshi used his talk this week to ponder the meaning of Artificial Intelligence from a Buddhist point of view. This week saw the release of yet another advanced “language model” which allows a computer to simulate human conversation, writing, and even art work. Some of these new systems could even conceivably pass a Turing Test. The Turing test simply states that if a person cannot tell if they are talking to a computer or human, then we must at least entertain the possibility that the computer is conscious in some way. Attached below are some general articles on AI “chat bots” for background.
Roshi pondered whether a super advanced AI could achieve enlightenment. Can we consider that the machine would have already transcended the aggregates (form, sensation, perceptions, mind, consciousness). Our machine can also scan the vast trove of Zen literature and hold the combined wisdom of all the sages. Is this enlightenment? We could ask the AI system about dogs and Buddha nature and we would get back pretty sophisticated answers. Could we even use a system to build a Zen master? Imagine if we could put a Zen master in everyone’s pocket. What are we to make of this? According to this link, this is already being done in Japan.
Can we build an enlightened computer? The first step for us is to understand how the AI works. At a very simple level, the AI can use the internet to scan every available document on a topic. It then builds word associations. (e.g. it has seen many times that the word “dog” and the word “mu” are associated.) This is very similar to how autocomplete works in modern email programs. If you ask the computer a question on a topic or form a sentence, if knows, on average, what an answer is or what text comes next because it has looked at billions of documents. At one level, what we have is a super genius academic student who has read every book in the library and has perfect memory.
Roshi then asked us if such a person is enlightened? Here he was clear – this person is an expert on Buddhism and Zen but this person has no more chance of being enlightened than anyone else. We need only to remember the story of the sixth patriarch of Zen who reached enlightenment as an illiterate by hearing the Diamond sutra being chanted. This realization was instant! It was “magic!” It cannot be calculated. Something in the Diamond sutra unlocked Huineng. This is different than breaking the sutra apart and building mathematical and statistical word associations.
Someone in the group asked why even bother considering AI when thinking about our own practice. Roshi thought that pondering this subject was useful as it gave us another skillful means to explore what it means to practice? Does a machine have karma? Does a machine suffer? Does a machine feel? What about us? Do we really suffer? What is suffering? Trying to ascribe these phenomena to a machine gets us to think more deeply about these things.
A machine can simulate wisdom and realization, but it is just that – a simulation. We need only to consider the Zen proverb that “Painted rice cakes do not satisfy hunger.” In the same way, collecting the literature of Zen and putting a clever program on top of it is may not be our definition of realization or enlightenment.
However, things are never so cut and dried in Zen. Does a dog have Buddha nature? – Mu. Does a computer have enlightenment? Why would you answer this question differently than the dog? Are we falling back into picking and choosing? Roshi wrapped up by reminding us that the real question really has nothing to do with computers and AI, rather it has to do with coming face to face with the question we are trying to solve in Zen practice. We are right back to the very point we started at.