I'm not sure I completely agree w/ the comment (nor do I think it vindicates CoPilot), but I think it does provide insight into why CoPilot is violating copyright.
The only analogy I can see is that copying the code internally to use in CoPilot training could be a violation of copyright (like how backing up your own MP3s is a violation of copyright?), but the licenses on these public repositories probably already allow that...
https://docs.github.com/en/site-policy/github-terms/github-t...
this is separate from the license you specify in the repository and you can't revoke it without removing your code from github.com.
GitHub shows code to those who wish to see it. it is up to the person using that code to use it according to the license. when I buy a car, it is up to me to use that car according to the law. when I buy a gun, it is up to me to use that gun according to the law. etc.
And yet we (modified) the law to mandate speedometers and seatbelts to make you more aware of the speed and more secure against failure. We require car companies to perform thousands of crash tests to validate that the tool the give you is safe for when you inevitably push “according to the law” a little too far.
We mandate mirrors and backup cameras because we know that those who intend to follow the law closely still have blind spots and it’s in everyone’s best interest to mitigate and increase awareness.
> when I buy a gun, it is up to me to use that gun according to the law. etc.
And yet few laws have caused the US (and other nations) to question this principle quite like gun laws.
Gun laws are really both a perfect example and the worst example of why we’re having a debate around CodePilot. We both expect people to be responsible for their decisions (you need to verify legality of that code snippet before using) while also giving them the notion that they can toe the line as much as possible (why regulate the availability of dangerous tools, crime is illegal, users won’t make a mistake).
Guns are used to kill people despite it being illegal. That’s why people want gun control laws. And in a comparison I never expected I would make, perhaps people want AI to be regulated because it will be (is?) used to circumvent copyright.
Edit: I don’t know if I really have a side I stand on in this debate overall, but I think the argument for why it’s copyright violation today is pretty compelling. We wouldn’t make the progress we’ve made without this violation and perhaps the loss of copyright is a worthy sacrifice?
I still don't see how there is any footing for a copyright infringement claim here, given that users who put public code on github.com explicitly grant GitHub a license to use that code to provide services to other GitHub users.
that license grant is above and beyond what any specified license terms the repo itself grants to users of the code.
you literally grant GitHub the right to do this when you put your code on github.com.
Relevant snippet:
> you grant each User of GitHub a nonexclusive, worldwide license to use, display, and perform Your Content through the GitHub Service and to reproduce Your Content solely on GitHub as permitted through GitHub's functionality (for example, through forking). You may grant further rights if you adopt a license.
They key parts are the “through GitHub” portion. GitHub is being careful to not give people rights to your content beyond the right to view it through GitHub. Performance refers to multimedia like music and video assets (according to others parts I didn’t reproduce).
No one is gaining a license to use your code through the inclusion on GitHub.
Section D is the relevant section.
https://docs.github.com/en/site-policy/github-terms/github-t...
>We need the legal right to do things like host Your Content, publish it, and share it. You grant us and our legal successors the right to store, archive, parse, and display Your Content, and make incidental copies, as necessary to provide the Service, including improving the Service over time. This license includes the right to do things like copy it to our database and make backups; show it to you and other users; parse it into a search index or otherwise analyze it on our servers; share it with other users; and perform it, in case Your Content is something like music or video.
>This license does not grant GitHub the right to sell Your Content. It also does not grant GitHub the right to otherwise distribute or use
I would say they would have a pretty hard time to justify using the content for AI training (and selling) based on that license. Copilot didn't exist at the time when many agreed to that license, so an argument saying Copilot is part of the service would be difficult to pull off. Moreover they don't even provide copilot to people hosting on GitHub.
Note that MS themselves are not claiming that they are allowed to use the code due to their terms of service. They claim they can do it due to fair use.
if this is indeed how they charge for Copilot, and I don't know if it is or is not, then they will need to show that they have done their due diligence in making sure that code is not reproduced verbatim when a user requests that it not reproduce code verbatim.
I'm quite sure that GitHub can defend Copilot in court. That's part of the process of offering a new feature to customers; making sure that it is legal and defensible to do so.
All of the armchair attorneys here who think they know better than GitHub's attorneys when operation of the service puts GitHub's ass on the line is ... I wish I had 1 percent of that confidence. I would be a thousand times more confident than I am now.
I don't think you can control it. Machine Learning models do not create anything, they make a prediction of the expected outcome based on the training/validation data. Similar to how human beings are an outcome of their experiences, so are ML models. Ofc human beings are much more complex than a ML model.
> The latter is obviously a violation of copyright, full stop.
It's not obvious to me that (2) is a violation of copyright. Unlike patents, copyright violation is not as simple to prove. My understanding is that, at least in the US, independent creation is a valid defense against copyright infringement. For example if 2 people independently write the same story and can prove that they did, they can both hold copyright over that story.
The analogue to this does exist without AI, when creating something that looks like copyright infringement, clean room design (don't look at similar things) is often done to ensure that "independent creation" can be used as a valid defense in court. Given that, I think (1) is probably not safe to do at all if you can't prevent (2).
This is the theoretical case but I don't think I've ever seen that actually happen in practice.
Outputting copyrighted material is a violation of copyright, period. Whether that violation is enforceable depends on your means though.
> > The latter is obviously a violation of copyright, full stop.
> It's not obvious to me that (2) is a violation of copyright. Unlike patents, copyright violation is not as simple to prove. My understanding is that, at least in the US, independent creation is a valid defense against copyright infringement. For example if 2 people independently write the same story and can prove that they did, they can both hold copyright over that story.
> The analogue to this does exist without AI, when creating something that looks like copyright infringement, clean room design (don't look at similar things) is often done to ensure that "independent creation" can be used as a valid defense in court. Given that, I think (1) is probably not safe to do at all if you can't prevent (2).
I don't think the analogue holds, the AI does have direct view of the actual code. In the most paranoid clean room design you have two teams, one analyses the behaviour of some software and writes a specification (without view of the source code), the other then uses that spec to write the reimplementation.
Copilot turns that on its head, you ask to do something it then looks up the source code how to do it and gives that to you.
The only way to prevent all uses of your code is to keep it secret.
If anyone wants to say me using copilot violates their copyright, then sue me. But if you have no loss of reputation or revenue, and I have an innocent infringer defense - noone can stop me.
Someone recently said most statements on HN should automatically get "in the US" appended to them due to how US centric many of the views are. This is an excellent example. There are plenty of juristictions where "fair use" doesn't exist.
Statutory damages still apply.
> I have an innocent infringer defense
There's no such thing. It's a meaningless phrase, in a legal sense. Claiming that infringement was accidental or unintentional is not a defense. It has no effect on a determination of guilt or innocence. All it affects is the penalty.
Fair use is a defense, but a more limited one than you seem to believe. The usual formulation is "criticism, comment, news reporting, teaching, scholarship, or research" but none of those apply to Copilot. Fair-use claims are also not accepted by default, but only by demonstration that the four factors defining it are all applicable.
Also, since you've brought it up recently, you as the defendant in a copyright case don't get to choose jurisdiction. Usually the plaintiff does, either because it's explicitly defined in the same license that grants anyone rights at all or because it's a place where they do business. If you live in a different jurisdiction that might affect whether the plaintiff or court can collect any penalties, but not whether any are assessed. Having yourself declared persona non grata in multiple jurisdictions doesn't seem like a good long-term choice.
https://copyright.columbia.edu/basics/fair-use.html
Instead of "flooding the zone" with dozens of comments offering nothing but the same few (false) claims - strong echoes of a recently banned user BTW - I strongly suggest you actually read up on copyright and fair use. They're not whatever you want them to be. Courts are unimpressed by your towering intellect.
The latter is obviously a violation of copyright, full stop.
The former, to me, is obviously not a violation. If it were, that would massively tilt the playing field in favor of large corporations. It would become very hard to independently train your own models. Philosophically, I go by the principle that if it's (il)legal to do yourself, then it should be (il)legal to do the same thing with an AI's assistance.
The massive complicating factor is that nobody knows how to do (1) without also doing (2) as a side effect, because we don't understand how deep learning works well enough to control it.