5 ms·

Meta's leaked model isn't open-source. I can found a business using Linux, that's open-source. The LLM piracy community are unpaid FB employees; it is not legal

Meta's leaked model isn't open-source. I can found a business using Linux, that's open-source. The LLM piracy community are unpaid FB employees; it is not legal for anyone but Meta to use the results of their labor. I know this might be hard news but it needs to be said... if you want to put your time into working on open source LLMs, you need to get behind something you have a real (and yes, open source) license for.

AnthonyMouse 4y ago
Most of the code isn't specific to a model. It happens that LLaMA is approximately the best LLM currently available to the public to run on their own hardware, so that's what people are doing. But as soon as anyone publishes a better one, people will use that, using largely the same code, and there is no reason it couldn't be open source. I'm also curious what the copyright status of these models even is, given the "algorithmic output isn't copyrightable" thing and that the models themselves are essentially the algorithmic output of a machine learning algorithm on third party data. What right does Meta have to impose restrictions on the use of that data against people who downloaded it from The Pirate Bay? Wouldn't it be the same model if someone just ran the same algorithm on the same public data? (Not that that isn't an impediment to people who don't want to risk the legal expenses of setting a precedent, which models explicitly in the public domain would resolve.)
- leereeves 4y ago
  > I'm also curious what the copyright status of these models even is That's my question as well. The models are clearly derivative works based on other people's copyrighted texts. Only a twisted court system would allow Google/OpenAI/Facebook to build models on other people's work and then forbid other people to build new models based on GOF's models.
  - AnthonyMouse 4y ago
    > That's my question as well. The models are clearly derivative works based on other people's copyrighted texts. That's not that clear either. (Sometimes it's more clear. If you ask the model to write fan fiction, and it does, and you want to claim that isn't a derivative work, good luck with that.) But the model itself is essentially a collection of data. "In Harry Potter and the Philosopher's Stone, Harry Potter is a wizard" is a fact about a work of fiction, not a work of fiction in itself. Facts generally aren't copyrightable. If you collect enough facts about something you could in principle reconstruct it, but that's not really something we've seen before and it's not obvious how to deal with it. That's going to create a practical problem if the models get good enough to e.g. emit the full text of the book on request, but the alternative is that it's illegal to make a model that knows everything there is to know about popular culture. Interesting times.
    - leereeves 4y ago
      > "...Harry Potter is a wizard" is a fact about a work of fiction, not a work of fiction in itself But LLMs aren't trained to learn facts like "Harry is a wizard", they're trained to reproduce specific expressions like "You're a wizard, Harry". That is, they're trained by prompting them with a selection from a (probably copyrighted) text and weights are adjusted to make it more likely they'll output the next word of the text. They're not a collection of general facts, they're a collection of estimates about which word follows which other words, and the order of words is the essence of copyright in text.
      - AnthonyMouse 4y ago
        A probability distribution isn't the order of words, it's a fact about the order of words. Pedants have been complaining about this kind of thing for years. If you generate random data, no one has a copyright on that. But if you XOR it with a copyrighted work, the result is indistinguishable from random data. No one could tell you which was generated randomly and which was derived from the copyrighted work. But XOR them back together again and you get the copyrighted work. Things like that get solved pragmatically, not mathematically. There is no basis for saying that one set of random bits is infringing and the other isn't, but if you're distributing them for the sole purpose of allowing people to reconstitute the copyrighted work, you're going to be in trouble. Now we have something with different practicalities. The purpose of training the model on existing works is so that it can e.g. answer questions about Harry Potter, which the majority wants to be possible and is the same class of thing that search engines need to be able to do. But the same model can then produce fan fiction as an emergent property, so what now?
math_dandy 4y ago
LLaMA leaked intentionally?
- happycube 4y ago
  De facto, yes. There was no way the weights wouldn't be posted everywhere once they went out to that many people.
- int_19h 4y ago
  There's a pull request in the official LLaMA repo that adds Magnet links for all the models to the README. Until these were uploaded to HuggingFace, this PR was the primary source for most people downloading the model. https://github.com/facebookresearch/llama/pull/73/files https://github.com/facebookresearch/llama/pull/73/files Two months later, Facebook hasn't merged the change, but they also haven't deleted it or tried to censor it in any way. I find that hard to explain unless the leak really was intentional; with pretty much any large company, this kind of thing would normally get killed on sight.
lerchmo 4y ago
this is a temporary state. Open source alternatives are already available and more are being trained.
mirekrusin 4y ago
Meh, you can experiment on it for personal use as much as you want and that's all what's needed in this short period of time before powerful, open base models start appearing like mushrooms at which point the whole thing is going to be moot.
dragonwriter 4y ago
> Meta’s leaked model isn’t open-source. Meta’s leaked model has been a factor in spurring open source development, whether or not it is open source; the article also discusses the practical market of effect of leaked-but-not-open things like Meta’s model combined with the impracticality of prosecuting the vast hordes of individuals using it, and particularly notes that the open-except-for-the-base-model work on top of it is a major benefit for Meta (who can use the work directly) that locks out everyone else in the commercial market (who cannot), and leaning into open source base models is a counter to that.
tysam_and 4y ago
I'm still not on board with calling it leaked...the weights were open for anyone to get and use as long as they agreed to use them academically. Basically, completely open source with a non-commercial license. I'm not sure why so many people keep saying it 'leaked'. It's just using open source weights not directly from the provider in a way that violates the software license.
- tdullien 4y ago
  I am still absolutely baffled that people think weights are copyrightable and hence licensable. There is no reason to believe they are, which means any restriction placed on the weights themselves is bullshit.
  - downWidOutaFite 4y ago
    It takes millions of dollars to generate the weights, shouldn't it have some legal protection?
    - dragonwriter 4y ago
      > It takes millions of dollars to generate the weights, shouldn’t it have some legal protection? It does, if you choose to keep them internally as a trade secret. It does, if you share it only with people you contract with not to disclose it. But, for copyright specifically, rather than “some legal protection” more generally, the “it takes millions of dollars” argument is a financial recasting of the “sweat of the brow” concept which has been definitively rejected by the courts.
  - SuoDuanDao 4y ago
    why wouldn't they be copyrightable? is it a discovered versus written thing?
    - dragonwriter 4y ago
      > why wouldn’t they be copyrightable? Why would they be? I mean, what is the specific argument that they fall within the scope of the definition of what is copyrightable under US law?
spullara 4y ago
I don't think we know where weights stand legally yet. They may end up being like databases, uncopyrightable.
- anticensor 4y ago
  Database rights are a thing in Europe. They are almost like copyright, but with shorter duration.
Blahah 4y ago
You are making an assumption that seems very strange to me - that the license matters for important use cases. It doesn't. Access to the technology is the only important factor, because nothing interesting about AI involves commercialising anything. It's a tool and now people have it. Whether they can make a company out of it is so far down the list of how it could make a difference that it doesn't even register.

AnthonyMouse 4y ago

Most of the code isn't specific to a model. It happens that LLaMA is approximately the best LLM currently available to the public to run on their own hardware, so that's what people are doing. But as soon as anyone publishes a better one, people will use that, using largely the same code, and there is no reason it couldn't be open source. I'm also curious what the copyright status of these models even is, given the "algorithmic output isn't copyrightable" thing and that the models themselves are essentially the algorithmic output of a machine learning algorithm on third party data. What right does Meta have to impose restrictions on the use of that data against people who downloaded it from The Pirate Bay? Wouldn't it be the same model if someone just ran the same algorithm on the same public data? (Not that that isn't an impediment to people who don't want to risk the legal expenses of setting a precedent, which models explicitly in the public domain would resolve.)

leereeves 4y ago
> I'm also curious what the copyright status of these models even is That's my question as well. The models are clearly derivative works based on other people's copyrighted texts. Only a twisted court system would allow Google/OpenAI/Facebook to build models on other people's work and then forbid other people to build new models based on GOF's models.
- AnthonyMouse 4y ago
  > That's my question as well. The models are clearly derivative works based on other people's copyrighted texts. That's not that clear either. (Sometimes it's more clear. If you ask the model to write fan fiction, and it does, and you want to claim that isn't a derivative work, good luck with that.) But the model itself is essentially a collection of data. "In Harry Potter and the Philosopher's Stone, Harry Potter is a wizard" is a fact about a work of fiction, not a work of fiction in itself. Facts generally aren't copyrightable. If you collect enough facts about something you could in principle reconstruct it, but that's not really something we've seen before and it's not obvious how to deal with it. That's going to create a practical problem if the models get good enough to e.g. emit the full text of the book on request, but the alternative is that it's illegal to make a model that knows everything there is to know about popular culture. Interesting times.
  - leereeves 4y ago
    > "...Harry Potter is a wizard" is a fact about a work of fiction, not a work of fiction in itself But LLMs aren't trained to learn facts like "Harry is a wizard", they're trained to reproduce specific expressions like "You're a wizard, Harry". That is, they're trained by prompting them with a selection from a (probably copyrighted) text and weights are adjusted to make it more likely they'll output the next word of the text. They're not a collection of general facts, they're a collection of estimates about which word follows which other words, and the order of words is the essence of copyright in text.
    - AnthonyMouse 4y ago
      A probability distribution isn't the order of words, it's a fact about the order of words. Pedants have been complaining about this kind of thing for years. If you generate random data, no one has a copyright on that. But if you XOR it with a copyrighted work, the result is indistinguishable from random data. No one could tell you which was generated randomly and which was derived from the copyrighted work. But XOR them back together again and you get the copyrighted work. Things like that get solved pragmatically, not mathematically. There is no basis for saying that one set of random bits is infringing and the other isn't, but if you're distributing them for the sole purpose of allowing people to reconstitute the copyrighted work, you're going to be in trouble. Now we have something with different practicalities. The purpose of training the model on existing works is so that it can e.g. answer questions about Harry Potter, which the majority wants to be possible and is the same class of thing that search engines need to be able to do. But the same model can then produce fan fiction as an emergent property, so what now?

math_dandy 4y ago

LLaMA leaked intentionally?

happycube 4y ago
De facto, yes. There was no way the weights wouldn't be posted everywhere once they went out to that many people.
int_19h 4y ago
There's a pull request in the official LLaMA repo that adds Magnet links for all the models to the README. Until these were uploaded to HuggingFace, this PR was the primary source for most people downloading the model. https://github.com/facebookresearch/llama/pull/73/files https://github.com/facebookresearch/llama/pull/73/files Two months later, Facebook hasn't merged the change, but they also haven't deleted it or tried to censor it in any way. I find that hard to explain unless the leak really was intentional; with pretty much any large company, this kind of thing would normally get killed on sight.

lerchmo 4y ago

this is a temporary state. Open source alternatives are already available and more are being trained.

mirekrusin 4y ago

Meh, you can experiment on it for personal use as much as you want and that's all what's needed in this short period of time before powerful, open base models start appearing like mushrooms at which point the whole thing is going to be moot.

dragonwriter 4y ago

> Meta’s leaked model isn’t open-source. Meta’s leaked model has been a factor in spurring open source development, whether or not it is open source; the article also discusses the practical market of effect of leaked-but-not-open things like Meta’s model combined with the impracticality of prosecuting the vast hordes of individuals using it, and particularly notes that the open-except-for-the-base-model work on top of it is a major benefit for Meta (who can use the work directly) that locks out everyone else in the commercial market (who cannot), and leaning into open source base models is a counter to that.

tysam_and 4y ago

I'm still not on board with calling it leaked...the weights were open for anyone to get and use as long as they agreed to use them academically. Basically, completely open source with a non-commercial license. I'm not sure why so many people keep saying it 'leaked'. It's just using open source weights not directly from the provider in a way that violates the software license.

tdullien 4y ago
I am still absolutely baffled that people think weights are copyrightable and hence licensable. There is no reason to believe they are, which means any restriction placed on the weights themselves is bullshit.
- downWidOutaFite 4y ago
  It takes millions of dollars to generate the weights, shouldn't it have some legal protection?
  - dragonwriter 4y ago
    > It takes millions of dollars to generate the weights, shouldn’t it have some legal protection? It does, if you choose to keep them internally as a trade secret. It does, if you share it only with people you contract with not to disclose it. But, for copyright specifically, rather than “some legal protection” more generally, the “it takes millions of dollars” argument is a financial recasting of the “sweat of the brow” concept which has been definitively rejected by the courts.
- SuoDuanDao 4y ago
  why wouldn't they be copyrightable? is it a discovered versus written thing?
  - dragonwriter 4y ago
    > why wouldn’t they be copyrightable? Why would they be? I mean, what is the specific argument that they fall within the scope of the definition of what is copyrightable under US law?

spullara 4y ago

I don't think we know where weights stand legally yet. They may end up being like databases, uncopyrightable.

anticensor 4y ago
Database rights are a thing in Europe. They are almost like copyright, but with shorter duration.

Blahah 4y ago

You are making an assumption that seems very strange to me - that the license matters for important use cases. It doesn't. Access to the technology is the only important factor, because nothing interesting about AI involves commercialising anything. It's a tool and now people have it. Whether they can make a company out of it is so far down the list of how it could make a difference that it doesn't even register.