It’s difficult to introduce ‘AI’ into a conversation without people spinning off into philosophical rabbit holes, but if we focus on music/culture, things become a lot simpler. Turing machines are an extremely powerful technology. They have already had a huge impact on music and utilizing more ‘novel’ algorithms(markov chains, neural nets, genetic algos, blah blah blah) is just a continuation of that.
Like any other tool(drum, piano, synthesizer, DAW), it is ultimately the job of the composer to use it creatively to write music. The fact that autechre can produce what they do using mostly if/then conditionals is a testimony the importance of this.
The only paradigm change I think ML brings to the table is that emulation of past works because almost trivial. I have already heard eerily spot-on examples of fake Beatles songs trained on a corpus of Beatles tracks. Humans are already very guilty of this mimicry(see most sub-genres of music), it’s now just becoming more efficient. Perhaps this will push us, as species, to prioritize originality more than we currently do?
Not to beat up on ML, of course. Again, like any technology, there’s nothing stopping people from using it in novel ways. How long before someone trains a convolution networks on a mixture of Taylor Swift and Merzbow?
