Having used my Rehearsal Studio blog to write about the experience of writing both a senior thesis and a doctoral dissertation on the topic of computer music under the supervision of Marvin Minsky, who died on Sunday night of a cerebral brain hemorrhage at the age of 88, I felt I could catch up on other impressions of Minsky by reading the (updated) version of his obituary on the Web site for The New York Times. That article was written by Glenn Rifkin with contributions from John Markoff. What I felt was most important about that article was its recognition that the MIT Artificial Intelligence Project (later the Artificial Intelligence Laboratory) was a source of not only “steps toward artificial intelligence” (the title of a survey paper that Minsky wrote in the early days of the discipline) but also major advances in how we would think about and use computers in general.
Rifkin singled out examples such as Minsky’s work on a neural network learning machine and the role that Minsky’s lab would play in the emergence of large-scale computer networking, thus anticipating the Internet as we now know it. However, from my perspective as one trying to use the computer as part of my research into making music, what probably mattered most was how, at Minsky’s lab, programming computers was an interactive task mediated through symbol-manipulating systems. By the time I entered the Massachusetts Institute of Technology as a freshman, the use of computers had advanced well beyond the days of preparing configurations of binary bits, often at a bit-by-bit level. People would type symbols onto punched cards, which the computer would then use to encode those configurations of bits; and, at the other end of the process, the results of running that encoding could be revealed through a (frequently obscure) printout. At Minsky’s lab, on the other hand, we would type at teletypes; and, thanks to the TECO text editor, we could see what we were typing on a video screen and do things like correct typing errors. Similarly, we had an interactive “debugging” system (errors in a computer program were called “bugs;” and there is even a literal “origins story” behind that term) called (of course) DDT (which stood for “dynamic debugging tool”).
The point of all those details is that anyone working in Minsky’s lab was always thinking about symbols and how they could be interpreted. A major advance came with the development of the programming language LISP, for which symbols, in all their generality, were the major primitives, rather than numbers assuming that role. Thinking about the ways in which computers could impact music making thus became a matter of thinking about what sorts of symbol systems would best mediate what the computer was doing.
The good news was that those of us interested in the topic had a useful “primer” for our thoughts. Any notation of music was, de facto, a symbol system. My senior thesis, entitled “EUTERPE: A Computer Language for the Expression of Musical Ideas,” amounted to the invention of a symbol system that could accommodate both the basic primitives of music notation and the control primitives necessary for computer programming. That language had emerged from some of Minsky’s early experiments in getting a computer to compose music; and it provided a more systematic symbolic foundation for what he had been trying to do. (The expressiveness of that foundation would then become the basis of my doctoral research, resulting in a thesis entitled “A Parallel Processing Model of Musical Structures.”)
What Minsky had triggered, however, was the higher-level idea that the making of music had the same foundation as the making of any other computer program. This was a radical shift from the approach that Max Mathews was pursuing at the Bell Telephone Laboratories, which amounted to a “digital anticipation” of what we now know as the modular synthesizer. Our two views complemented each other. Mathews was looking at using programming techniques to synthesize auditory signals, while I was worried about symbolic constructs that would parallel not only what we could write in music notation but also what mind told us to do when we were playing our instruments.
After leaving with my doctoral degree, I would continue to think about expressive symbol systems, one of which amounted to a representation of what Heinrich Schenker had tried to communicate through his graphic diagrams. Meanwhile, I continue to work with symbol manipulating systems in just about everything that I do. (I am engaging with one as I type this.) What I have learned over the years, however, is that there is never an end to the questions that can be asked or the methods that can be devised for dealing with them. While I may be working with words instead of the symbolic constructs for computer programming available to me at the Artificial Intelligence Laboratory, just about everything I do is still symbol manipulation, just as it was for all of those early (and mostly anonymous) pioneers of music notation.