5 ms·

The weights aren't created by humans so they aren't copyrightable - at least in theory.

monocasa 3y ago
That has nothing to do with whether it's open source or not. Software wasn't for sure copyrightable before 1976 in the US, but there was plenty of closed source code that simply only ever distributed the binaries.
- epistasis 3y ago
  It's open source in the sense that you can take the existing weights, apply any one of dozens of techniques for adapting the model to your data, and anybody you give your model to can do the same. This right isn't protected by copyright, it's protected by the lack of copyright, and how executables are directly modifiable. So the AGPL won't be possible, but it's a lot like permissive licenses, without attribution.
  - monocasa 3y ago
    That's like saying an executable that I never gave out the source to is "open source" because there's nothing stopping you from binary patching it. Also, the license to the model assumes copyright (Apache 2) and works by granting you permissions based on that copyright.
    - epistasis 3y ago
      It's very very different from patching a binary, because these models don't work like a simple compilation of understandable source code into incomprehensible weights. The source training data is even less comprehensible and useful than its encapsulation into weights. A set of weights is a pre-trained local minima in the model space is a both executable and modifiable. It's usually far more useful than the source training data, because the work has been done.
      - monocasa 3y ago
        > It's very very different from patching a binary, because these models don't work like a simple compilation of understandable source code into incomprehensible weights. The source training data is even less comprehensible and useful than its encapsulation into weights. Tons of binary patching works that way. For instance from the gameshark days it was relatively common for a patch that worked for unknown reasons, but who's discovery was tooling assisted and simply displayed a desired effect (and commonly a lot of undesired effects that weren't clear). > A set of weights is a pre-trained local minima in the model space is a both executable and modifiable. None of that changes whether it's open source or not. I guarantee you that mistral has some code (and a lot o data) laying around that created this model. > It's usually far more useful than the source training data, because the work has been done. For certain operations maybe. I guarantee that Mistral wouldn't constrain themselves to fine-tuning if they had a major change to make to the model.
        epistasis 3y ago
        No, patching binaries is not really similar at all, because it's viewed as an inferior method to having the source code and modifying that. With these large models, the weights really are quite useful objects in themselves, the very starting points of future modifications, and importantly the desired points of modifications Practioners in the field call these models "open source" and write open source software to run them and use them to power open source software.
        monocasa 3y ago
        > No, patching binaries is not really similar at all, because it's viewed as an inferior method to having the source code and modifying that. > With these large models, the weights really are quite useful objects in themselves, the very starting points of future modifications, and importantly the desired points of modifications For fine tuning. You go back to the training pipeline to make major changes. > Practioners in the field call these models "open source" Some do. In contrast to the accepted definitions of open source. > and write open source software to run them and use them to power open source software. There's plenty of open source code to run closed source binaries.
  - sroussey 3y ago
    It’s very likely protected by EU copyright.
Sayrus 3y ago
Copyright is not the only kind of Intellectual Property. While they may not be copyrightable (I think you may be refering to jugement related to having an AI as author?), they clearly are MistralAI IP. In the same way, data in a database are usually not copyrightable. They are still the company's property (excluding issues with PII and such).
- bhickey 3y ago
  Database rights don't exist in the United States. Their weights obviously aren't protected by trademark. So, what IP regime protects the weights?
  - Sayrus 3y ago
    Mistral AI is incorporated in Paris, France. IANAL, but I think these[1][2] may qualify to protect these weights. While there is a discussion to be had on sovereignty and international reaches of local laws (such as DMCA for instance), I think it's disingenuous to consider only US legal point of view. [1] https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32019L0790#d1e822-92-1 https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CEL... [2] https://www.legifrance.gouv.fr/loda/id/JORFTEXT000000573438/2024-01-31/ https://www.legifrance.gouv.fr/loda/id/JORFTEXT000000573438/...
    - bhickey 3y ago
      > Mistral AI is incorporated in Paris, France. No US court is going to invent new rights under US law because a company happens to be incorporated in a foreign jurisdiction. > While there is a discussion to be had on sovereignty and international reaches of local laws (such as DMCA for instance) The DMCA (and GDPR) do not have international reach. The DMCA only appears to apply internationally because it binds a whole series of American companies with an international presence. Meanwhile the GDPR protects European residents. A company dealing with these residents has crossed the border, so to speak. > I think it's disingenuous to consider only US legal point of view. Candidly, you're either confused about what disingenuous means or you're being disingenuous yourself.
shiandow 3y ago
That does raise the question how irreversibly you have to process something before it is no longer protected by copyright. Obviously zipping something is not enough, even lossy compression is (obviously) not enough. But then how does a language model differ from a lossy compression? Is it just the compression ratio? (are the weights even that much smaller than the data?) There are ways to train models that guarantee that the amount of information transferred per data point is limited, but to my knowledge those aren't used (and may be prohibitively expensive).
wongarsu 3y ago
When I take a picture with my phone, those pixels weren't created by a human either. Yet I still own their copyright. At the core the question is how much artistic input there is in creating LLM models. Are the choice of the model architecture, hyperparameters and training data artistic choices comparable to those by a photographer setting up a shot? Or are they more comparable to technical work that's only protectable by trade secrets, patents and trademarks?
- Q6T46nT668w6i3m 3y ago
  The parent is asking about weights and from that perspective, _stochastic approximation_ is essential.
- int_19h 3y ago
  Training a model is more akin to setting a video camera somewhere and having it automatically record or not according to some algorithm you devise. The code of the algorithm would be copyrightable, of course, but the recording is another matter, since the "creativity" aspect is kinda missing there.

That has nothing to do with whether it's open source or not. Software wasn't for sure copyrightable before 1976 in the US, but there was plenty of closed source code that simply only ever distributed the binaries.

epistasis 3y ago
It's open source in the sense that you can take the existing weights, apply any one of dozens of techniques for adapting the model to your data, and anybody you give your model to can do the same. This right isn't protected by copyright, it's protected by the lack of copyright, and how executables are directly modifiable. So the AGPL won't be possible, but it's a lot like permissive licenses, without attribution.
- monocasa 3y ago
  That's like saying an executable that I never gave out the source to is "open source" because there's nothing stopping you from binary patching it. Also, the license to the model assumes copyright (Apache 2) and works by granting you permissions based on that copyright.
  - epistasis 3y ago
    It's very very different from patching a binary, because these models don't work like a simple compilation of understandable source code into incomprehensible weights. The source training data is even less comprehensible and useful than its encapsulation into weights. A set of weights is a pre-trained local minima in the model space is a both executable and modifiable. It's usually far more useful than the source training data, because the work has been done.
    - monocasa 3y ago
      > It's very very different from patching a binary, because these models don't work like a simple compilation of understandable source code into incomprehensible weights. The source training data is even less comprehensible and useful than its encapsulation into weights. Tons of binary patching works that way. For instance from the gameshark days it was relatively common for a patch that worked for unknown reasons, but who's discovery was tooling assisted and simply displayed a desired effect (and commonly a lot of undesired effects that weren't clear). > A set of weights is a pre-trained local minima in the model space is a both executable and modifiable. None of that changes whether it's open source or not. I guarantee you that mistral has some code (and a lot o data) laying around that created this model. > It's usually far more useful than the source training data, because the work has been done. For certain operations maybe. I guarantee that Mistral wouldn't constrain themselves to fine-tuning if they had a major change to make to the model.
      - epistasis 3y ago
        No, patching binaries is not really similar at all, because it's viewed as an inferior method to having the source code and modifying that. With these large models, the weights really are quite useful objects in themselves, the very starting points of future modifications, and importantly the desired points of modifications Practioners in the field call these models "open source" and write open source software to run them and use them to power open source software.
        monocasa 3y ago
        > No, patching binaries is not really similar at all, because it's viewed as an inferior method to having the source code and modifying that. > With these large models, the weights really are quite useful objects in themselves, the very starting points of future modifications, and importantly the desired points of modifications For fine tuning. You go back to the training pipeline to make major changes. > Practioners in the field call these models "open source" Some do. In contrast to the accepted definitions of open source. > and write open source software to run them and use them to power open source software. There's plenty of open source code to run closed source binaries.
- sroussey 3y ago
  It’s very likely protected by EU copyright.

Sayrus 3y ago

Copyright is not the only kind of Intellectual Property. While they may not be copyrightable (I think you may be refering to jugement related to having an AI as author?), they clearly are MistralAI IP. In the same way, data in a database are usually not copyrightable. They are still the company's property (excluding issues with PII and such).

bhickey 3y ago
Database rights don't exist in the United States. Their weights obviously aren't protected by trademark. So, what IP regime protects the weights?
- Sayrus 3y ago
  Mistral AI is incorporated in Paris, France. IANAL, but I think these[1][2] may qualify to protect these weights. While there is a discussion to be had on sovereignty and international reaches of local laws (such as DMCA for instance), I think it's disingenuous to consider only US legal point of view. [1] https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32019L0790#d1e822-92-1 https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CEL... [2] https://www.legifrance.gouv.fr/loda/id/JORFTEXT000000573438/2024-01-31/ https://www.legifrance.gouv.fr/loda/id/JORFTEXT000000573438/...
  - bhickey 3y ago
    > Mistral AI is incorporated in Paris, France. No US court is going to invent new rights under US law because a company happens to be incorporated in a foreign jurisdiction. > While there is a discussion to be had on sovereignty and international reaches of local laws (such as DMCA for instance) The DMCA (and GDPR) do not have international reach. The DMCA only appears to apply internationally because it binds a whole series of American companies with an international presence. Meanwhile the GDPR protects European residents. A company dealing with these residents has crossed the border, so to speak. > I think it's disingenuous to consider only US legal point of view. Candidly, you're either confused about what disingenuous means or you're being disingenuous yourself.

shiandow 3y ago

That does raise the question how irreversibly you have to process something before it is no longer protected by copyright. Obviously zipping something is not enough, even lossy compression is (obviously) not enough. But then how does a language model differ from a lossy compression? Is it just the compression ratio? (are the weights even that much smaller than the data?) There are ways to train models that guarantee that the amount of information transferred per data point is limited, but to my knowledge those aren't used (and may be prohibitively expensive).

wongarsu 3y ago

When I take a picture with my phone, those pixels weren't created by a human either. Yet I still own their copyright. At the core the question is how much artistic input there is in creating LLM models. Are the choice of the model architecture, hyperparameters and training data artistic choices comparable to those by a photographer setting up a shot? Or are they more comparable to technical work that's only protectable by trade secrets, patents and trademarks?

Q6T46nT668w6i3m 3y ago
The parent is asking about weights and from that perspective, _stochastic approximation_ is essential.
int_19h 3y ago
Training a model is more akin to setting a video camera somewhere and having it automatically record or not according to some algorithm you devise. The code of the algorithm would be copyrightable, of course, but the recording is another matter, since the "creativity" aspect is kinda missing there.