GenBebop

Lee Spector
Cognitive Science, Hampshire College

GenBebop is a project in which genetic programming is used to produce interactive jazz music-making programs (called "constructed jazz musicians"). This page contains links to output from evolved programs and references to related publications.

Output from GenBebop

• Sounds related to [Spector & Alpern 1994]

  The constructed jazz musician that was used for these recordings produces a 4-measure "response" to a 4-measure "call" provided by a human. In these recordings the human input was via guitar and pitch-to-MIDI translation (using a Roland CP-40); the responses were produced on a Kurzweil K2000 using a trumpet sound. The recordings were edited to remove short processing lags between calls and responses. The "medley" is a sequence of illustrative call/response pairs glued together. The "music critic" that assessed fitness during evolution was based on rules derived from Jazz method books.
  

  • What Would Charlie Parker Think? (115KB) (one trade-4 interaction).
  • GenBebop Medley (565KB) (played at AAAI-94)

• Sounds related to [Spector& Alpern, 1995]

  The constructed jazz musician that was used for these recordings produces a 1-measure "response" to a 1-measure "call." In these recordings the calls were taken from a script and played using a piano sound. The responses were played using a trumpet sound. The "music critic" that assessed fitness during evolution was a 3-layer feed-forward neural network trained using back-propagation on a small set of Charlie Parker melodies.
  

  • Way Out to Lunch (5.5MB): This was evolved using the neural network critic only. It is unplayable by humans and it violates many musical conventions.
  • Hybrid (5.5MB): This was evolved using an additive hybrid of the neural network critic and the rule-based critic from [Spector & Alpern 1994]. The output is playable by humans. It is often repetitive, but it is sometimes interesting.

References

[Spector & Alpern 1995] Spector, L., and A. Alpern. 1995. Induction and Recapitulation of Deep Musical Structure. In Working Notes of the IJCAI-95 Workshop on Artificial Intelligence and Music. pp. 41-48. (postscript, 615k)

[Spector & Alpern 1994] Spector, L., and A. Alpern. 1994. Criticism, Culture, and the Automatic Generation of Artworks. In Proceedings of the Twelfth National Conference on Artificial Intelligence, AAAI-94, pp. 3-8. Menlo Park, CA and Cambridge, MA: AAAI Press/The MIT Press. (PDF, 4MB)

Page updated April 25, 2008