Part of Temporal Constructs Appendix, which contains details behind the research: dates, quotations, descriptions of instruments, and sources.
General definition
A recorded conversation between Jacco Gardner and David Rosenboom was held in October 2026. It is edited from audio. It traces Rosenboom’s route from analogue computers through the Buchla Touché and its software to the making of HMSL.
Relation to the practice
The interview is the professional doctorate’s primary account of HMSL. It explains why HMSL was built as it was, the working conditions that shaped it, and what its authors intended. It supplies the evidence for the first subsidiary question. It notes Rosenboom’s own recollection of the term intelligent instruments and of what he thinks contemporary AI can and cannot do.
Account of HMSL
Rosenboom’s route to the language began with an aim he had held from the 1960s. He wanted to build compositional procedures into the structure of the instruments themselves. He worked with analogue computers in live performance first, and the Buchla Touché gave that ambition a digital home. The hierarchy he designed for the Touché, and carried into HMSL, was chosen for generality.
Memory constrained the work, and he gives its cost in circuit boards of eight kilobytes. The brevity of the rule set is deliberate in his account. He built the models to apply as widely as possible to any kind of music. He treats the language as an artwork as well as a tool. He wrote the first version for the Touché as its regular user software, Foil. Foil stands for the Far Out Instrument Language, begun in 1979 and released in April 1980.
Rosenboom on intelligence
Rosenboom recalls that such instruments were called intelligent ones at the time. The developers stopped using the word intelligence when it became a bad idea to use. He reports Roger Penrose’s suggestion that artificial intelligence be called artificial cleverness. He argues that a system structured only on extant material cannot be creative in the sense an artist means. He calls the loss cognitive surrender, where a generated solution takes the place of the understanding of the person using it. He says the source of his own creativity is not accessible to him either, so he does not see it as something a model could be instructed to reach.
Cybernetic inheritance
At the University of Illinois, a centre of cybernetic research, he learned second-order cybernetics from Heinz von Foerster. He also learned the compositional algorithms of Lejaren Hiller’s generation. He built analogue computers for performance, including a vacuum-tube machine from before transistors. He describes voltage-controlled synthesisers as special-purpose analogue computers. They were designed for musicians’ access according to somebody’s idea of what that meant.
Verification
Read in full from the transcript supplied by the owner.
Preparation and checking
The questions were prepared in October 2026. They cover the origins of HMSL, its relation to MIDI and the cybernetic thinking it came from. They take in what has never been tried with it since then, and how its author reads the present moment against the vision of the 1980s. One further question, proposed by the agent, asks which musical ideas became conceivable only because the language existed.
That question gives a historical baseline against which programming with a model can be compared. The transcript was checked afterwards against sources. The corrections are part of the account: a wrong date for the death of Larry Polansky, a misspelled surname, and doubtful dates for the Buchla instrument the language began on. A chronology had also drifted.