John Al Biles is well-renowned on the computer/music scene for his work with GenJam, the computer jazz improviser. Mr. Biles works at the Rochester Institute of Technology as a professor and the undergraduate program coordinator. We were certainly grateful that Al Biles could give us the opportunity to interview him.
1.) Why did you chose GAs to use for GenJam?
I began playing with GAs in graduate school in the late 70's. I had one of John Holland's students, Al Bethke, as an advisor for my MS thesis, and his course on "Adaptive Systems" was my first exposure to GAs. I think that GAs are a terrific approach to searching large, ill-defined spaces, in this case the space of "nice" melodic ideas. There is also an analogy to the "population" of licks that most jazz players have in their heads. These licks come and go over time in a manner similar to evolution; ideas that were cool in the past become overused or cliched, so I stop playing them. The real key to GenJam's success is the hierarchical chromosome representation I come up with for four-bar phrases. This representation works very well for training "stored" soloists, and it has been a perfect target representation for the interactive stuff I've added in the last year. My paper covering the real-time interactive evolution of what I play into what it will play will appear on my web site after the conference at which I will present it, in early October.
2.) Do you think you will see similar programs appearing over the next few years? Do you think AI will be used more and more in the music business?
Undoubtedly. There already is a market niche for systems that generate original (and therefore royalty-free) music for use in multimedia
production. The quality of the music produced is not awesome, but it's suitable for background "ear candy." There currently are not too many people useing GAs in music (maybe a dozen or so that I'm aware of), but there likely will be more in the near future. Recently Bill Gates even got a patent for something that smacks of algorithmic composition, although a doubt that the patent could be broadly enforcible.
3.) GenJam is a step towards dynamic music creation - how long do you think it'll be before computers will be able to autonomously creates compositions?
It's been done. Algorithmic composition dates back to the Illiac Suite in the 1950's. You said two things in your question, however, "dynamic" and "autonomous," and they are different. GenJam is dynamic in that it responds interactively, both to a mentor who is training it and to a performer who is playing with it. Most "autonomous" composition systems, like David Cope's EMI, are not as dynamic. They listen and learn from examples, but they compose from a stored base of compositional knowledge, which is not dynamic during the actual compositional process.
4.) What other applications do you see GA fit for?
GAs have been used for a wide variety of applications. Many are configuration problems like IC or chip layout, network topology and design,
and all kinds of messy optimization problems.
5.) Do you feel that computers will also start to branch out into other fine art fields - such as painting?
Actually, computer art is at least as well developed as computer music, and has received much more media attention. If by painting you mean actually wielding a brush on canvas, I'm not aware of anyone doing that robotically, but I'd be surprised if nobody is. I've seen robots playing bagpipes and drums, so moving a paintbrush shouldn't be that hard, as long as you're not trying to imitate Van Gogh or Renoire.
