For the purists, then, it seems to me that you just need a means for providing feedback (which you need anyway for self-learning), and then adjust the actions accordingly. For example, if you use BIAB to generate a solo, but then did it again a few times, the soloist algorithm could learn from that and steer away from what it originally produced. If you save the file, it could assume that what it finally provided was something you liked and could start from there the next time.

But people need feedback in order to learn. So that must be a prerequisite for a computer system as well. If the computer is only generating output without a feedback mechanism in place, there is no way for the program to "learn".

Just my simplistic view about this. It can get very complicated and very convoluted very quickly.


John

Laptop-HP Omen I7 Win11Pro 32GB 2x2TB, 1x4TB SSD
Desktop-ASUS-I7 Win10Pro 32GB 2x1.5TB, 2x2TB, 1x4TB SATA

BB2025/UMC404HD/Casios/Cakewalk/Reaper/Studio One/MixBus/Notion/Finale/Dorico/Noteworthy/NI/Halion/IK

http://www.sus4chord.com