How does it solve the problem of dependencies without becoming bloated?

This might be it. Depending on how the sound is mixed the voice might cancel itself out since it’s usually similar between left and right and the instruments are not. So they will not experience the same interference

Well that’s kinda worrying

Generally the hope is that people naturally converge on one community and then the other would die out. Or maybe resurge if for some reason people don’t like the first

I like the point about LLMs interpolating data while humans extrapolate. I think that’s sums up a key difference in “learning”. It’s also an interesting point that we anthropomorphise ML models by using words such as learning or training, but I wonder if there are other better words to use. Fitting?