"To know that what is impenetrable to us really exists, manifesting itself as the highest wisdom and the most radiant beauty, which our dull faculties can comprehend only in their most primitive forms -- this knowledge, this feeling, is at the center of true religiousness." -- Albert Einstein

Einstein is reported in a well-known story to have said of his physics career, that everything he had done, he had done in an effort to know the mind of God.

This was not hubris on Einstein's part, because it wasn't a theological interest which occupied him; rather it was one of design. ("I have to add the last point explicitly," Einstein once wrote to his friend M. Solovine, "lest you think that, weakened by age, I have fallen into the hands of priests.")

Our investigation of go presents us with a much easier task: We have only to know the mind of Man.

To create a computer program that can play go as well as a human being plays, we needn't go as far as having to know whether or not God plays dice with the universe. Knowing how actual people behave when playing on actual go boards should suffice!

Ideally, we would just like to have the computer play like a particular human. Now, just as there are computer programs that can compose music in the style of Bach, or some other arbitrarily-chosen composer (of whose body of work there exist sufficient samples by which to define that style), why are there not any programs which can play go in the style of Honinbo Shusaku, for example, or of Go Seigen, or of Takemiya?

The same sort of pattern-recognition techniques that work for music may be applied to go, but the difficulty of that task lies in the appropriate selection of features to be recognized in patterns of go behavior: What sorts of things do we measure? Style comprises what things, exactly?

Feature selection is the most crucial phase of pattern recognition.

Given the sheer number of potential patterns in go, especially allowing for the generalization go into boards of arbitrary size and dimensionality, and further, providing for simultaneous play by an arbitrary number of colors, we have essentially a one-to-one mapping of go events onto quark events, and vice-versa.

At this point go theory goes on to replace superstring theory, and we may be so emboldened as to attempt even a glimpse into the mind of God. Indeed, God does not play dice with the universe: He plays go. [Principal investigator is available for this research, if you'd like to fund it. MacArthur grants cheerfully accepted.]

In the meantime, we may at least coax the machine into emulating the mind of Man: Let it write a fugue in the style of Bach, or play a go game in the style of Shusaku. For, if the right features are chosen for measurment, a computer program is not only capable of playing go, but of playing go in the style of a particular professional.

Stay tuned.