Dom

If you or I (hobbyists) are going to end up with 3 reasonably well integrated and instrumented songs in one week, most of the heavy lifting will have been done by this AI program.
This last week empirically exceeds your "completed song" output of the prior 2 or 3 decades.

It's like strapping on one of those jetpacks and then thinking that YOU can actually fly. The jetpack is flying. You are guiding it and attached.
If you get one hundred feet in the air and then remove the jetpack, then you will find out what can fly and what cant.

I am not saying your AI is a bad thing. I plan to use. Probably a lot.
I am saying that your output in the last few days, posted here on JPF, is microcosm of what is coming.

As I stand, I am using a form of AI with TOONTRACK on keys, organ, drums, base. It starts with a "BANDMATE" feature where I put in a few bars of guitar rhythm and the software generates compatible drum, etc styles. Then I change and tweak from there. I sort through styles and pick per my sensibilities. But I cant play piano, organ or drums. And only root notes on the bass.

I audio engineer in my DAW.
I dont get a studio quality sound at my house, but I get good enough to fool most listeners and have never been turned down by any major outlets.

Back when I only had my very limited voice and guitar playing ability which is not great but competent (vis-a-vis most pop music), the instrumentation and hourly studio time ...that I spend now... would cost 25-40k per song using studios and studio musicians. Their product would be better but not 25-40k worth of better....for a hobbyist.

The point...
I know where my limitations as a songwriter, musician stop and the machines take over.
I am becoming more and more of a "producer."
If I was producing 3 songs in a week, that would be 95% the software and I would be a producer.