Hacker News Reader

Comment by echelon | Hacker News Reader

Most of Meta's models have not been released as open source. Llama was a fluke, and it helps to commoditize your compliment when you're not the market leader.

There is no good or open AI company of scale yet, and there may never be.

A few that contribute to the commons are Deep Seek and Black Forest Labs. But they don't have the same breadth and budget as the hyperscalers.

View on HN

Llama is not open source. It is at best weights available. The license explicitly limits what kind of things you are allowed to use the outputs of the models for.

Which, given what it was trained on, is utterly ridiculous.

Yup, but that being said, Llama is GPLv3 weather Meta likes it or not. Same as ChatGPT and all the others. ALL of them can perfectly reproduce GPLv3 licensed works and data, making them derivative work, and the license is quite clear on that matter. In fact up until recently you could get chatGPT to info dump all sorts of things with that argument, but now when you try you will hit a network error, and afterwards it seems something breaks and it goes back to parroting a script on how it's under a proprietary license.

This is interesting but it has not been proven in court, right?

related stuff has, the core part being that if your model reproduces parts or all of a licensed work, it needs to comply with the license / copyright. Otherwise why aren't pirates just making 'models' that generate protected material, or music, and completely bypass all laws?

I know because I wanted to, as a form of protest/performance art, train a model to a few Disney movies and publicly distribute, but legal advice was this would put me directly into hot water not just because of who im pissing off (which i knew and was comfortable with) but also the fact there was precedent (i.e. news papers suing LLM providers).

It would be an open and shut case that would leave me in financial ruin.

The reason openAI hasn't been struck with this yet is, who has the time? and there isn't much to learn from all that either. Most open source tooling out competes openAI's offering as is, so the community wouldn't really win beyond punishing someone.

I don't see how this follows at all. Github isn't GPL3 just because it stores and gives you back gpl3 code

read the license, and look up what derivative work means. If you're still unclear after that I'm happy to walk you through it.

Is that easier to enforce than having AI only trained in a legal way (=obeying licenses / copyright law)?

Yes. Having training obey copyright is a big coordination problem that requires copyright holders to group together to sue meta (and prove they broke copyright, which is not something proven before for LLM).

Whereas meta suing you into radioactive rubble is straightforward.

That's not true, the llama that's open source is pretty much exactly what's used internally

> There is no good or open AI company of scale yet, and there may never be.

Deepseek, Baidu.

Reply on news.ycombinator.com