[Conasg]

The biggest problem as I see it is that they monetized it, on a massive scale. If this had been open source like they promised, it would seem more like fair use. But to take huge amounts of copyrighted material without consent and effectively sell access to that material? I struggle to see how that could ever be legal.

[Centigonal]

I agree with your main point, but I want to clarify that OpenAI never promised open source. This is a pernicious misunderstanding that OpenAI has been all too happy to leave uncorrected. Open source was never part of the mission, and I can't find any place where OpenAI leadership has implied that it is or will be.

From 2016:

> They’d raised a billion dollars and hired an impressive team of thirty researchers—but what for? “There are twenty to thirty people in the field, including Nick Bostrom and the Wikipedia article,” Amodei said, “who are saying that the goal of OpenAI is to build a friendly A.I. and then release its source code into the world.”

> “We don’t plan to release all of our source code,” Altman said. “But let’s please not try to correct that. That usually only makes it worse.”

Source: https://www.newyorker.com/magazine/2016/10/10/sam-altmans-ma...

[Wolfbeta]

> I can't find any place where OpenAI leadership has implied that it is or will be.

have you tried saying the name out loud?

[Centigonal]

The name is very misleading!

Was that a deliberate misdirection? Not sure. Apparently, Elon Musk takes credit for the name and wanted the company's products to be open source. However, Musk has said a lot of things about companies he is involved with that later turn out to be untrue.

[persolb]

Is three times a pattern yet? I love my FSD, and I though Autopilot was great (I got both when they were much cheaper)... but neither name matches the functionality.

[sublinear]

> Musk has said a lot of things about companies he is involved with that later turn out to be untrue.

Funny how putting forth lots of contrarian red tape after and because someone has made all the claims and promises has a way of doing that.

[7e]

Sam Altman seems like a snake from that quote.

[intrasight]

The human brain is trained on copyright material. The human brain can also be monetized. And the human brain is not open source. I don't see why artificial brains should be treated differently.

[Terr_]

Others have mentioned the "if the AI has those rights then it's also slavery" angle, but I think there's also a much less-dramatic argument...

Imagine that Acme Music hires large sweatshops of workers, who are tasked with listening to trending (copyrighted) songs and then go through grueling practice to sing and play near-perfect performances of those songs, delivered to any customer who asks for one.

That business model is still copyright infringement! It remains so no matter how much you de-automate by having human-brains replace machines at the same tedious tasks.

P.S.: Kind of like the inverse of patents, where "$task, but this time with a computer" is not automatically a new patent-able thing.

[williamcotton]

I work in copyright law.

Style is not subject to copyright, only expression. If there are unique lyrics and unique melodies under two bars then there is no copyright infringement.

A song that sounds like other pop songs is basically the very nature of pop music.

Pick a genre, for example reggae, and you’ll find a lot of perfectly legal songs that are almost indistinguishable aside from the musicians who are playing the music.

[Terr_]

> Style is not subject to copyright, only expression.

In this case you can assume Acme Music is making purposeful performances/impersonations of the original songs/artists, because the goal of the scenario is to show that "but I used a human brain to do it" isn't a magical get-out-of-infringement-free card.

Illustrating the boundaries of fair-use or transformative works is... Well, a somewhat-separate issue that would need a different kind of scenario.

[williamcotton]

Unless I’m shown concrete examples of what constitutes this kind of purposeful impersonation I’m unsure what is being discussed. I take your explanation to mean that any company that decided to start writing, producing and distributing reggae music would fall under this description.

Again, I am honestly confused by what is meant in this thread!

[Terr_]

A paraphrased recap:

1. Conasg: Taking huge amounts of copyrighted material without consent and selling access to it seems illegal.

2. Intrasight: Humans learn from copyrighted material all the time, and indirectly profit from it, so why can't the AI?

3. Me: Some of the stuff the software is being used to do would still copyright infringement even if you had teams of humans (inefficiently) doing it instead.

[pas]

getting inspired by visual art is more ok than its audio counterpart. mostly because audio is lower-dimensional somehow (melodies, riffs, sampling, beats, things are either recognizable or not at all.. well, not entirely, but in practice there are seemingly not that many things in this grey area, even when bands cite inspiration, the influence is sometimes not noticeable), of course maybe it's just a yet to be explored niche.

whereas visually copying style somehow is more possible and ok

[williamcotton]

As a musician, I can tell you that I can instantly hear a chord progression in any song and find loads of prior art.

But we’re really talking about style. Think reggae, isn’t it immediately obvious what style of music you’re listening to?

[pas]

Indeed, but and to me it seems there are "fewer" styles currently in music than in visual arts. (Even though I spend a lot more time listening to music than looking at visual art, even including movies/series.)

Of course this is a super bad n=1 qualitative study with very bad response coding, but maybe this is simply because I haven't spent enough time with musical style copying LLMs (if there are any), and only spent about a minute listening to that forever death metal youtube AI channel.

[Centigonal]

Human brains aren't birthed by companies who enslave them for exclusive use by their clientele. LLMs and human minds are not comparable.

[jacobdejean]

Your brain is trained on things that you were licensed to see. Unless, of course, you downloaded pdfs in college instead of buying the textbooks.

[johnnyanmac]

sure. and if you ask me as a human brain to copy code I worked with at a previous (non open-source) company line-by-line, I can get myself and current company in trouble. This usually isn't an issue because human brains aren't usually as good at regurgitating exact copies, but it's still pretty much the foundation of copyright.

So why should an artificial brain be treated differently, especially when it's been shown how you can prompt one in a way to almost regurgitate the exact copyright info (similar to how you can prompt a human brain)?

[JohnFen]

> I don't see why artificial brains should be treated differently.

Perhaps because they're tools, not people.

[cyanydeez]

slavery is illegal.

[intrasight]

Nobody mentioned slavery