>We also welcome submissions that present a theoretical reflection on the impact of new technologies on artistic practices and forms of expression.
Finally, the perfect forum to deliver my perspective of ML and AI current state of music and the outlook for the next 1, 3, and 5 years. There will be a lot of hurt feelings. It will be very constructive however and show the roadmap between artists and engineers that is yet to develop in earnest and can potentially be the next largest US cultural export.
To use a simile, the current level of ML/AI music is about on par with a precocious elementary student with a basic understanding of an Akai MPC. That’s why you can listen to 1,000 of them and maybe one resembles a not-bad song. Not bad is not good mind you. It just means it’s not physically offensive / unlistenable to humans.
One of the hardest lessons for non-creative people to learn about the how of what I do need to understand creativity is inherently rule breaking. Derivative art - like that hack warhol’s Prince portrait- have a limited impact overall. Coming back from crazy town with a handful of cool images nobody has seen before like Dali? Now we’re talking!
There is hope in the sector, but unfortunately for the tech realm, you’re doomed unless you reconfigure your development approach to have 60% of the brain pathways look like spaghetti designed by a coked up ambitious chimp with Visio. I’ll have to dig it up but I heard guitar sales in the US increased during the pandemic…same basic design since the log.
The biggest issue with AI art isn't the quality, we can already make generative art that is faithful enough to not be distinguishable. It's that authorship is an inherent part of art, and nobody is really interested in art that has no human creator in the same way nobody watches bots play chess, despite higher quality of play. Authenticity is central to art, what Walter Benjamin called the 'aura' of a work. (it's unique place in space and time, which automated art does not have).
Keeping this in mind the NFT craze over seemingly random pictures makes more sense, because it was essentially a mechanism to give a replaceable, fungible, mechanic work an aura of sorts, anchoring it. Rather than fidelity, this will be the central question of the tech/art intersection.
And how much training and guiding did that take? It's not like they hit "generate" and the first try made some good music. This definitely took a ton of human guidance and curation.
Well, it's not really AI any more, it's more like a Beethoven Phrase Generator operated by a committee of Beethoven Music Appreciators. It's a cool thing and I'm also bullish on AI as a creative tool, but a lot of current offerings are great at small scales but fall apart on the larger structural level.
If you have an OpenAI account, try getting GPT-3 to write a short murder mystery, for example. GPT-3 is kinda good at dyadic and triadic interactions and even some kinds of story generation, but doesn't get the structural requirement of multiplying contradictions until a tipping point resulting in convergence (though I think LLMs can be taught that).
Finally, the perfect forum to deliver my perspective of ML and AI current state of music and the outlook for the next 1, 3, and 5 years. There will be a lot of hurt feelings. It will be very constructive however and show the roadmap between artists and engineers that is yet to develop in earnest and can potentially be the next largest US cultural export.
To use a simile, the current level of ML/AI music is about on par with a precocious elementary student with a basic understanding of an Akai MPC. That’s why you can listen to 1,000 of them and maybe one resembles a not-bad song. Not bad is not good mind you. It just means it’s not physically offensive / unlistenable to humans.
One of the hardest lessons for non-creative people to learn about the how of what I do need to understand creativity is inherently rule breaking. Derivative art - like that hack warhol’s Prince portrait- have a limited impact overall. Coming back from crazy town with a handful of cool images nobody has seen before like Dali? Now we’re talking!
There is hope in the sector, but unfortunately for the tech realm, you’re doomed unless you reconfigure your development approach to have 60% of the brain pathways look like spaghetti designed by a coked up ambitious chimp with Visio. I’ll have to dig it up but I heard guitar sales in the US increased during the pandemic…same basic design since the log.