News for hackers. Stuff that matters.

A comment on GitHub Copilot is not infringing copyright.

Re:GitHub Copilot is not infringing copyright by MattIPv4 on Monday July 05, 2021 @08:05AM (#27737081)
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

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by timdaub on Monday July 05, 2021 @11:08AM (#27738726)
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

Edit: My comment was enabled: https://juliareda.eu/2021/07/github-copilot-is-not-infringin...

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by bootlooped on Monday July 05, 2021 @01:16PM (#27739959)
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.
[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by user5994461 on Monday July 05, 2021 @05:21PM (#27741939)
>>> 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

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by sfletcher on Monday July 05, 2021 @08:14AM (#27737143)
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.
[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by creshal on Monday July 05, 2021 @08:19AM (#27737186)
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.

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by mjburgess on Monday July 05, 2021 @08:21AM (#27737205)
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.

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by duckmysick on Monday July 05, 2021 @12:32PM (#27739606)
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

[ Parent | on HN ]
Re:GitHub Copilot is not infringing copyright by pessimizer on Monday July 05, 2021 @04:10PM (#27741422)
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.

[ Parent | on HN ]