I think there are a few reasons that mean the network effect of music services (YT music, Spotify, etc) are hard to beat:

• Available of commerical music - a service could obviously enable Piracy but this does make things more complicated
• Exposure for smaller bands - there’s a reason that even C3 content tends to be posted to the fediverse using its YouTube upload instead of the native upload
• Larger training data for recommendations - YT/Spotify can compare your entire listening history to every other user they have and use this to generate playlists of songs you might like, this is computationally expensive, relatively complex algorithmically (especially because it can’t be as simple as I doscribed at scale) & benefits from a large userbase.

This is something I’d like to work on when the world isn’t as on fire, I’m less focused on the backend at the moment and think:

• improvement could first be made to frontend apps like spottube (effective a Spotify privacy app), while piracy is easier (you can’t takedown software for enabling piracy as easily as you can services)
• Scrobbing/Playlist integration across multiple services to get more data even from people still locked in to Spotify
• Playlist generation with open algorithms - honestly this is something the fediverse is pretty weak on, I think Lemmy is really the only app I’ve seen grapple with this and it’s still very much functionally closed (while the code is open users can’t tweak their feeds)

None of this really helps artists though, but by breaking the cartel of music services, playlist generation could more easily include sources that do pay artists.

Everyone complaining about matrix are just complaining about front end issues, they are absolutely fixable.

That’s not ironic, people didn’t confuse Reddit for Digg, it wasn’t a even Digg clone nor did it pretend to be one.

I don’t think you can get sued for using a trademark as long as your use doesn’t confuse people into thinking you are the original.

Do you have examples?

Because most of what you are listing is stuff that has been using ML for years (possibly decades when it comes to meteorology) and just slapped “AI” on as a buzzword.

What advances?

LLMs don’t have anything to do with abstract ideas, they quite literally produce derivative content based on their training data & prompt.

The same can be said of the approach described in the article, the “GPLv4” would be useless unless the resulting weights are considered a derivative product.

A paint manufacturer can’t claim copyright on paintings made using that paint.

Seems like the easiest fix is to consider the produce of LLMs to be derivative products of the training data.

No need for a new license, if you’re training code on GPL code the code produced by LLMs is GPL.