3 ms·

This seems to completely ignore the fact that we've seen Copilot regurgitating exact copies of existing code, and even with the incorrect license attached when

This seems to completely ignore the fact that we've seen Copilot regurgitating exact copies of existing code, and even with the incorrect license attached when it was asked for it. [0] [0] https://twitter.com/mitsuhiko/status/1410886329924194309 https://twitter.com/mitsuhiko/status/1410886329924194309

sfletcher 5y ago
The Google Books case cited here allowed Google to show exact snippets (extracts) from the copyrighted books, hard to see how this is any different.
- creshal 5y ago
  It also has no relevance for the discussion at hand. Yes, Github can display all of its content – that's kind of the point of it. But Copilot doesn't exist to show you random code snippets for the sole purpose of showing them. Using this copyrighted material to create derivative works is a completely different use case, and not covered at all by the Google Books ruling, or any other I'm aware of.
- mjburgess 5y ago
  You wouldn't be allowed to make derivative works of those books; i.e. copy/paste into your own work. Google isnt making new books, or enabling people to make derivative copies; it is merely previewing a book. Github search is a preview. Copilot is a copy/paste.
  - duckmysick 5y ago
    What about Google Books Ngram Viewer? Isn't that a derivative work based on copyrighted content? It's more than just a search or preview - it contains both novel information and snippets of existing content. Is linguistic corpus a special case? https://books.google.com/ngrams https://books.google.com/ngrams
    - pessimizer 5y ago
      https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,_Inc https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,.... The research value actually turned activity that would be infringing into activity that was not infringing. Take away things like the ngram viewer and Google Books infringes.
timdaub 5y ago
Agree. I've written a comment on her blog about this. Hoping she'll enable it. I've published an opinion piece on the subject matter myself: https://rugpullindex.com/blog#BuiltonStolenData https://rugpullindex.com/blog#BuiltonStolenData Edit: My comment was enabled: https://juliareda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/#comment-38373 https://juliareda.eu/2021/07/github-copilot-is-not-infringin...
bootlooped 5y ago
That function exists in hundreds, if not thousands, of GitHub repositories. The function is so well known it has it's own Wikipedia page. If there is a more famous function in computer science, I don't know what it is. The fact that a machine trained on GitHub repositories might reproduce such common code is not alarming or surprising to me. I think people are using this as an example and implying it's happening all over the place, but I've yet to see another example like it.
- user5994461 5y ago
  >>> If there is a more famous function in computer science Maybe fizzbuzz? Let's try to auto generate some fizzbuzz code, see what we get :D

sfletcher 5y ago

The Google Books case cited here allowed Google to show exact snippets (extracts) from the copyrighted books, hard to see how this is any different.

creshal 5y ago
It also has no relevance for the discussion at hand. Yes, Github can display all of its content – that's kind of the point of it. But Copilot doesn't exist to show you random code snippets for the sole purpose of showing them. Using this copyrighted material to create derivative works is a completely different use case, and not covered at all by the Google Books ruling, or any other I'm aware of.
mjburgess 5y ago
You wouldn't be allowed to make derivative works of those books; i.e. copy/paste into your own work. Google isnt making new books, or enabling people to make derivative copies; it is merely previewing a book. Github search is a preview. Copilot is a copy/paste.
- duckmysick 5y ago
  What about Google Books Ngram Viewer? Isn't that a derivative work based on copyrighted content? It's more than just a search or preview - it contains both novel information and snippets of existing content. Is linguistic corpus a special case? https://books.google.com/ngrams https://books.google.com/ngrams
  - pessimizer 5y ago
    https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,_Inc https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,.... The research value actually turned activity that would be infringing into activity that was not infringing. Take away things like the ngram viewer and Google Books infringes.

timdaub 5y ago

Agree. I've written a comment on her blog about this. Hoping she'll enable it. I've published an opinion piece on the subject matter myself: https://rugpullindex.com/blog#BuiltonStolenData https://rugpullindex.com/blog#BuiltonStolenData Edit: My comment was enabled: https://juliareda.eu/2021/07/github-copilot-is-not-infringing-your-copyright/#comment-38373 https://juliareda.eu/2021/07/github-copilot-is-not-infringin...

bootlooped 5y ago

That function exists in hundreds, if not thousands, of GitHub repositories. The function is so well known it has it's own Wikipedia page. If there is a more famous function in computer science, I don't know what it is. The fact that a machine trained on GitHub repositories might reproduce such common code is not alarming or surprising to me. I think people are using this as an example and implying it's happening all over the place, but I've yet to see another example like it.

user5994461 5y ago
>>> If there is a more famous function in computer science Maybe fizzbuzz? Let's try to auto generate some fizzbuzz code, see what we get :D