5 ms·

Julia Reda did weigh in: https://juliareda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/ https://juliareda.eu/2021/07/github-copilot-is-not-infrin

by onli 5y ago

Julia Reda did weigh in: https://juliareda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/ https://juliareda.eu/2021/07/github-copilot-is-not-infringin...

louthy 5y ago
That seems to entirely ignore that Copilot has 'generated' GPL protected code verbatim, even down to the comments. There appears to be exact copies of certain code segments within their model, and that is being exactly copied out. Doesn't that fall under copyright protection?
- onli 5y ago
  Where I live: It does fall under it, if the code is sufficient complex enough (e.g. there is evident originality) to be protected under copyright in the first place.
- jolux 5y ago
  She does say she doesn’t think most snippets meet the originality criteria.
  - onli 5y ago
    You are right: > Otherwise, copyright conflicts would constantly arise when two authors use the same trivial statement independently of each other, such as “Bucks beats Hawks and advance to the NBA finals”, or “i = i+1”. The short code snippets that Copilot reproduces from training data are unlikely to reach the threshold of originality. Precisely because copyright only protects original excerpts, press publishers in the EU have successfully lobbied for their own ancillary copyright that does not require originality as a precondition for protection.
    - louthy 5y ago
      Nobody is complaining about `i = i + 1`, it's examples like this [1], an entire function, including comments written by the original author, protected under GPL: [1] https://twitter.com/mitsuhiko/status/1410886329924194309 https://twitter.com/mitsuhiko/status/1410886329924194309
      - onli 5y ago
        There is still the question of whether such a small part of a bigger work is enough to be protected, but regardless. I think the article was written before that tweet and does not assume that 1:1 copies of longer code parts happen. As I wrote already to you, there is that whole faction of "even using GPL code for learning is wrong", and to my reading Redas article mainly addresses that. And it has the assumption that the verbatim copies will be very small.
MattGaiser 5y ago
This only covers the part about the training data and derivative work. At what point does exactly cloning code become not a derivative work?
- loopz 5y ago
  Obviously it is not derivative work anymore when huge international corporations let their "AI" copy the code verbatim./s Emperors new clothes and all that.

by onli 5y ago

Julia Reda did weigh in: https://juliareda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/ https://juliareda.eu/2021/07/github-copilot-is-not-infringin...

louthy 5y ago

That seems to entirely ignore that Copilot has 'generated' GPL protected code verbatim, even down to the comments. There appears to be exact copies of certain code segments within their model, and that is being exactly copied out. Doesn't that fall under copyright protection?

onli 5y ago
Where I live: It does fall under it, if the code is sufficient complex enough (e.g. there is evident originality) to be protected under copyright in the first place.
jolux 5y ago
She does say she doesn’t think most snippets meet the originality criteria.
- onli 5y ago
  You are right: > Otherwise, copyright conflicts would constantly arise when two authors use the same trivial statement independently of each other, such as “Bucks beats Hawks and advance to the NBA finals”, or “i = i+1”. The short code snippets that Copilot reproduces from training data are unlikely to reach the threshold of originality. Precisely because copyright only protects original excerpts, press publishers in the EU have successfully lobbied for their own ancillary copyright that does not require originality as a precondition for protection.
  - louthy 5y ago
    Nobody is complaining about `i = i + 1`, it's examples like this [1], an entire function, including comments written by the original author, protected under GPL: [1] https://twitter.com/mitsuhiko/status/1410886329924194309 https://twitter.com/mitsuhiko/status/1410886329924194309
    - onli 5y ago
      There is still the question of whether such a small part of a bigger work is enough to be protected, but regardless. I think the article was written before that tweet and does not assume that 1:1 copies of longer code parts happen. As I wrote already to you, there is that whole faction of "even using GPL code for learning is wrong", and to my reading Redas article mainly addresses that. And it has the assumption that the verbatim copies will be very small.

MattGaiser 5y ago

This only covers the part about the training data and derivative work. At what point does exactly cloning code become not a derivative work?

loopz 5y ago
Obviously it is not derivative work anymore when huge international corporations let their "AI" copy the code verbatim./s Emperors new clothes and all that.