HN Explorer

Tell HN: GitHub Copilot scraping OS projects is good

7 points by fedeb95 ↗ HN
I know this is an unpopular opinion, but most OS software just sits there (including a lot of my projects). If I can exchange them to GitHub for free storage that's fine. Yes they should have asked. Yes I would have agreed.

8 comments

[ 2.6 ms ] story [ 33.3 ms ] thread
smoldesu
It would be good if Copilot was also released as Open Source, since it's a work derived from the contents of GPL/AGPL codebases. Until they clean up their training data, it won't be easy to call it "good" (and I say this as an AI/Microsoft/Github apologist)
saghm
If they made the concession that it was a derived work, would it even be able to license it as (A)GPL? I imagine it's scanning repos that have conflicting licenses. Off the top of my head, I remember ZFS's license not being considered compatible with GPL (hence it not being included by default by most Linux distros), but I'm sure there must be others.
gus_massa
It depends on the license.

For a project with BSD0 or MIT-0 that require no attribution, it is fine. Just use those licenses in your projects. (Note: Other people may repackage your code too.)

For BSD3 or MIT, that are license that require attribution, it's not clear that copilot can use the code, because they require attribution. Anyway, I think that most people that uses BSD or MIT does not care, but I'm not a lawyer.

But people that use GPL or AGPL usually cares about the "virality", i.e. all derived code MUST have the same license. I think they care a lot about using their code in a not GPL or AGPL licensed project. Many of them are angry.

fedeb95
Yes, I get it's more complicated. Just my perspective. The "should have asked" part includes "and respected negative answers"
gus_massa
In most interesting cases, asking is very complicated, because you must ask all the authors, not only the owner of the github account where the project is hosted.
e_i_pi_2
I don't think it's enough to say they should ask, or they their SLA technically allows it. The letter of the license might allow it but the spirit seems to say that copilot should also be free to use. If they weren't charging for it I don't think we'd see the same backlash. I've heard it described as "intellectual property laundering" and I think that's pretty fair
c7b
The only thing that bugs me about Copilot is that Microsoft apparently got an exclusive (!) license to the underlying "Open"AI models [0]. I fully welcome the technical progress those models represent, even if they should result in lower earnings potential for developers or artists (which I'm not convinced of, btw, given that they amount to productivity increases), but I think researchers, especially "Open"AI, should have prioritized open source models over corporate black boxes.

[0] https://blogs.microsoft.com/blog/2020/09/22/microsoft-teams-...

fedeb95
That I second. AI such powerful (will they hit a limit? Who knows?) should be public