Monday, May 9, 2020
I am ever on the lookout for people who have what one might call "spiritual insight" that I can learn from. In the past few years, I have discovered a number of Buddhist thinkers that fill this urge, most especially ones featured on Michael Taft's podcast Deconstructing Yourself. These are not the New Age hippie variety of "Buddhist" many come to expect; they are more like philosophers. As a general rule, they tend to be well-spoken, analytical close listeners. They would make excellent lawyers. In fact, they are so lawyerly precise with their language, and analyze their own thoughts and behaviors with such precision, that it at times sounds too neurotic even for *me*.
Anyways, I wanted to mention one of the podcast episodes that I found particularly good and insightful:
Deconstructing Yourself podcast interview
It is an interview with a guy named David Chapman, who happens to have trained in A.I. at M.I.T. several decades ago. He describes his philosophy as a type of "meta-rationality". Very roughly, as summarized by Taft early in that interview, meta-rationality takes the view that there is no single, simple "map" with which to view reality and chart its progression; one should use a variety of maps, instead. One might consider *that* philosophy as a single map; but it is a map "at a different level", at the level of thinking about thinking (metacognitive).
What this philosophy is militating against is the rationalist impulse, whereby one tries to take in a small amount of data about the world, and then simply by sitting in ones chair and applying long chains of logic, one deduces everything else one doesn't know (or a lot more of what one doesn't know; a lot more than Chapman thinks is possible). And such rationalists do exist in abundance. Chomsky has been called a rationalist, for example, in this critique by Peter Norvig:
Norvig blog entry
I think, at times, Chapman might take things too far. For example, based on what I've seen him write on Twitter, he seems to be doubtful about the prospects of certain machine learning methods to make major progress in AI. Perhaps, to him, these methods seem too much like a rationalist solution, where one attempts to deduce too much, assuming little.
Those issues aside, I do mostly agree with Chapman. The reach of each individual theory is going to be very limited; those that go a rather far are very rare.