I really like your ideas for patch generation and search. If your goal is to make music, it's much more compelling to have finer-grained control over creating something than having entire chunks made for you, even if they do sound good. This is probably why Suno, Udio etc. are not interesting for most musicians, but even for AI products intended as tools, there are still varying levels of control. In the case of Magenta, I really respect their angle, but I wonder if their tools aren't low-level enough. E.g. the Ableton plugin is impressive and fun to play with, but there are so few parameters. There's no control over primitives like instrument family, articulation, etc., and that's frustrating if you have an idea in mind. Maybe that's part of why it hasn't taken off? And yeah, I think a language interface would be a game changer, because you get more control with a low learning curve.
In defense of musical hardware, and since you bring up Kraftwerk, I saw them last year and they were incredibly boring to watch compared to the stuff they used to do (like seriously https://www.youtube.com/watch?v=UZiTr4hwHXI), even if they were supposed to be robots or whatever. And Tod Machover did some cool stuff with the Hyperinstruments back then, and contemporary classical folks are always doing electronic augmentation in some flavor. I think there's room for AI tools there too!
I really like your ideas for patch generation and search. If your goal is to make music, it's much more compelling to have finer-grained control over creating something than having entire chunks made for you, even if they do sound good. This is probably why Suno, Udio etc. are not interesting for most musicians, but even for AI products intended as tools, there are still varying levels of control. In the case of Magenta, I really respect their angle, but I wonder if their tools aren't low-level enough. E.g. the Ableton plugin is impressive and fun to play with, but there are so few parameters. There's no control over primitives like instrument family, articulation, etc., and that's frustrating if you have an idea in mind. Maybe that's part of why it hasn't taken off? And yeah, I think a language interface would be a game changer, because you get more control with a low learning curve.
In defense of musical hardware, and since you bring up Kraftwerk, I saw them last year and they were incredibly boring to watch compared to the stuff they used to do (like seriously https://www.youtube.com/watch?v=UZiTr4hwHXI), even if they were supposed to be robots or whatever. And Tod Machover did some cool stuff with the Hyperinstruments back then, and contemporary classical folks are always doing electronic augmentation in some flavor. I think there's room for AI tools there too!