All public GitHub code was used in training Copilot
1–10 of 734 posts
Re: All public GitHub code was used in training Copilot
#2A direct confirmation from GitHub itself. This is problematic because Copilot sometimes outputs code that was present in its training set.
https://fossbytes.com/github-copilot-generating-functional-a...
Re: All public GitHub code was used in training Copilot
#3The outrage-bait approach in this thread detracts from it. Yes, they trained it on everything. No, it's not clear if that's legal or not (probably is) or if that is much of a problem.
Re: All public GitHub code was used in training Copilot
#4To be fair, this could just be a mistaken interpretation from the support staffer that answered the question - they didn't sound sure ("apparently"). It certainly needs an official response from GitHub senior management but I wouldn't call the foul yet (not that it's even clear that it is a foul).
Re: All public GitHub code was used in training Copilot
#5OP’s rhetoric, and most discussion I see, asserts that training a model on copyrighted data is a copyright violation. Personally I don’t find this to be so obviously the case. Think back to when we were listening to AI generated pop music, for instance. I don’t recall any concern in HN comments about the copyright holders’ music being used for learning.
Re: All public GitHub code was used in training Copilot
#6Really hoping to see a max exodus from GitHub after this. Microsoft back to their old tactics like we all knew they would.
Re: All public GitHub code was used in training Copilot
#7To the people arguing it's "fair use" to use this for training an ML network. Where do you draw the line? What if you train an "ML network" with one or two inputs... so that they almost always "generate" exact copies of the inputs? Five inputs..? Ten? A thousand? A million?
Re: All public GitHub code was used in training Copilot
#8Really hoping to see a max exodus from GitHub after this. Microsoft back to their old tactics like we all knew they would.
If you have public repos anywhere people can train on them just as much.
Re: All public GitHub code was used in training Copilot
#9Really hoping to see a max exodus from GitHub after this. Microsoft back to their old tactics like we all knew they would.
Any suggestions for an alternative? One thing I like about github is that it 'seems' to be a defacto standard for portfolios & public works. It also has excellent integration into AWS and alike
Re: All public GitHub code was used in training Copilot
#10To the people arguing it's "fair use" to use this for training an ML network. Where do you draw the line? What if you train an "ML network" with one or two inputs... so that they almost always "generate" exact copies of the inputs? Five inputs..? Ten? A thousand? A million?
I can imagine a requirement of the sort 'generated code needs to match at most X% to snippets of the training data as shown over Y amount of sampling' but I am not sure if you can get a much better requirement than that.
Forbidding the training of AI on public code would definitely be a step too far though.
Edit: I'd also like if they provide a tool for checking if your code matches copyrighted code too close so you can confirm if you are violating anything or not when you use copilot.