If you dig Perlin noise fields, the next stop is fBM (Fractional Brownian Motion). Inigo Quilez has a good article about it - https://iquilezles.org/articles/fbm/
Saying it gives "the minimum" is generous, it's pretty much useless out of the box. And did I say slow as well? I think pi is great if you're into spending your time managing your harness rather than using it. In that regard, it's more like neovim.
What you mention of training without human data seems to me an impossibility. Unless you're talking about going back to programming an AI via traditional methods rather than relying on machine learning (which might not be impossible, hard to prove it as such at least).
I don't think you can divorce intelligence from all biological aspects and just get computational power. It's an interesting question though..
It's damn time for streaming platforms to allow users to filter out AI music. Better than that, would be for the EU to regulate this and to demand an opt-out by default for every streaming provider.
If i can be forthright, it looks like any other llm slop website design. The grain effect, the extra long FAQ, the reveal animations, the bad combination of font sizes and contrast ratios.. you're better off ripping off a website that has been actually designed by someone who understands what they are doing.
What does this have to do with copyright? This is stealing pure and simple. If you walk out of a museum with one of its art pieces, you're not going to get arrested for copyright infringement.
The meaning of native in these discussions is "no web technologies", because Qt gets thrown around and that's as native as to macOS as Electron, just in a different manner.