Tuesday, January 26, 2016

Remembering Marvin Minsky

I just finished reading the BBC News report of the death of Marvin Minsky, who succumbed to a cerebral hemorrhage at the age of 88. Minsky served as adviser for both my senior and doctoral theses. The former amounted to a systematic implementation of some ideas he had been playing around with through which he could compose music with the aid of computer software. The latter was basically an extended analysis of that what implementation could and could not express and what means of expression were involved. As is the case with many doctoral students, it led to a few publications, one of which I particularly enjoyed, since it got me into the Journal of Music Theory, which had been a major source for much of my background research.

I suppose what I remember most about Minsky was that he was a "hands off" adviser. He would let me go off on my own with relatively little guidance, but he would then subject my findings to intense criticism. At the time he was running the Artificial Intelligence Laboratory at the Massachusetts Institute of Technology, which was getting its funding from the Department of Defense Advanced Research Projects Agency. He had other areas of research to supervise to keep those guys was happy. I was a side show. So I had the freedom to run it myself while enjoying the facilities of the Laboratory. My best move came when the composer Ezra Sims became my "unofficial" adviser (and composition teacher); and much of what went into my thesis was the result of not only Sims' coaching but also his own experiences in using my software.

However, even before my graduate studies, Minsky made a mark in the world of computer music that never really got duly acknowledged. What I built for my senior thesis was basically a symbol manipulation system, whose symbols amounted to a hybrid of the fundamental constructs of music notation and the basic operators of a programming language. The former was, in just about any imaginable way, a predecessors of the MIDI coding system. Had Marvin been more concerned about intellectual property instead of chipping away at enormous research challenges, he probably could have made a strong case that what finally showed up in concrete form in my senior thesis would have been prior art for what MIDI had been trying to do. (On the other hand my work with Sims came about through a shared interest in microtonality. One of the first things I did, to satisfy the sort of expressiveness that most interested Sims, was divide the octave into 72 equal internals. My attachment to that representation was why I ignored MIDI during its first years of application.)

On a broader scale I acquired from Minsky the basic precept that finding the right questions to ask is always more important than answering them. In that respect I was glad to see the BBC quoting him on the downside of companies like Google and Facebook getting into artificial intelligence. He sees those companies trying to make money by commercializing things that never worked very well in the early stages of research. (There used to be a joke among graduate students that a thesis was based on a demonstration project that could only be applied to one example and did not even work reliably in that limited case!)

I suppose that Minsky was more interested in intelligence as a process, rather than a product that might eventually be marketed. In this respect he was a direct descendant of Alan Turing, whose "Turing Test" has been bastardized by contemporary technology hacks as a product that can vie in a competition, while Turing himself was more interested in the sorts of processes through which his "Imitation Game" could be played successfully. If those processes worked for a human player, then there was a good chance they could be implemented on a machine.

Minsky's background also endowed him with an appreciation of intelligence having a subjective dimension, that is elements that had more to do with psychology than with mathematics. Mind you, he believed that such subjectivity might eventually be mapped into a more objective domain; but he still knew enough to appreciate the difference. On the other hand I am not sure he appreciated that there was also a social dimension to intelligence, which may explain why he was often so impatient, if not bored, in many of his conversations.

That same comment that objects to commercialization advocates, instead, "giving support to individuals who have new ideas." I suppose that is why he took me on as a doctoral student. However, back when I was still trying to make a living in scientific research, I found very few such individuals when interviewing job candidates. Graduate schools were now coming out a crop of students who were most proficient at following orders, passed from funding organizations down through thesis advisers. I suppose that is why I am now happier these days writing on my own nickel (or, to be fair, nickels coming in part from the Social Security Administration). Even in the world of music, it is getting harder to come across "new ideas;" but I find the odds of finding them at a conservatory are better than those of finding them at even the best of the universities!

1 comment: