I work (play?) a lot with creative AI. There is some excellent stuff being done with musical applications right now, but you have little to no control over the final product. For example, several sites will allow you to do prompt to music. A few of them provide good, clean tracks that are done well. However, your options are limited if you dislike something about the final render. But just a year ago, writing a text prompt to music yielded garbage.

My point is that AI technology sometimes increases by the day. The text-to-image creators a year ago yielded some laughable results: six fingers, limbs that bent the wrong way, warped faces, etc..It was impressive what it COULD do, but still the limitations rendered it useless for most applications of the technology in the real world. People like me might ooh and ah over it because we looked past what wasn't there and saw the possibilities. Fast forward a year, and now you can text-to-prompt, and if you don't like it, you can in-paint sections of the image, reroll the entire prompt, expand the canvas, set aspect ratios, etc. We even have text-to-video now that's imperceptible from actual video (in some cases), but the technology evolves rapidly. I think we'll see the same thing with music applications of the technology.

I would encourage PG to explore AI as well. Can you imagine how cool it would be if instead of searching for real tracks and them maybe fitting or maybe not, we could just type something like, "I'm looking for a guitar part in the style of 1960's R&B in the style of the Temptations, the Supremes, and Sam and Dave" and it will render several options for you in those exact styles. There would be no need to dedicate so much space to Real Tracks. Then, you could store them yourself and make personalized styles.