Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

All public GitHub code was used in training Copilot

twitter.com

Post reply on HN
Print view View original

331–340 of 734 posts

Re: All public GitHub code was used in training Copilot

#331

Post by elliekelly »

unanswered wrote: post #248
_Understated_ wrote: post #227

Earlier quoted context omitted.

I understand the concept of fair use (I think) but I can't see how it applies to Copilot. Google didn't create new books from the contents of existing ones (whether you agree that they should have been allowed to store the books or not) but Copilot is creating new code/apps from existing ones. Edit: I guess my understanding of fair use was wrong. I stand corrected.

That you apparently think fair use is something you just think about real hard in order to see how you feel about it in a given situation demonstrates that you do not understand the concept of fair use. There are rules.

I don’t disagree with your point but was it necessary to make it in such a snarky way?

Re: All public GitHub code was used in training Copilot

#332

Post by swiley »

AaronFriel wrote: post #193
keonix wrote: post #185

If the training set contains verbatim (A)GPL code does this mean that Copilot also should be distributed by Microsoft under GPL? Because without it Copilot (as it is distributed by Microsoft) couldn't be built, wouldn't it make it a derivative work of GPL'd code (and obviously every other license)? I see a lot of people comparing human learning to machine learning in the comments, but there is a huge difference - we…

No, see Authors Guild v. Google. Even without a license or permission, fair use permits the mass scanning of books, the storage of the content of those books, and rendering verbatim snippets of those books. The Google Books site is not a derivative work of the millions of authors they copied from, and if they did copy any coincidentally GPL, AGPL, or creative commons copyleft work, the fair use exception applies befo…

Just for reference, the hackernews source is public.

Re: All public GitHub code was used in training Copilot

#333

Post by johndough »

MayeulC wrote: post #115
einarfd wrote: post #34

Earlier quoted context omitted.

> Out of curiosity, how do we define license violation in that case? I, as a human being, have trained by reading code, much of which is covered by licenses that are somehow not compatible with code I'm writing. Am I violating licenses? That depends, if you end up writing copies of the code you've studied then yes. You are on thin ice. Plagarization is definitely something that you can do with computer code. There ha…

> swap(var1, var2) Well, there's also the xor way to be pedantic :) var1 = var1 ^ var2 var2 = var2 ^ var1 var1 = var1 ^ var2 But yeah, not too much wiggle room there.

Another variation (assuming no overflows):

    var1 += var2;
    var2 = var1 - var2;
    var1 -= var2;
And another:

    var1 ^= var2 ^= var1 ^= var2;
Assembly even has an instruction for it:

    xchg eax, ecx

Re: All public GitHub code was used in training Copilot

#334

Post by otterley »

tyre wrote: post #262
otterley wrote: post #234

Earlier quoted context omitted.

If this issue is eventually litigated, we will see. The law in the Second Circuit (where the final judgment was rendered before the case was eventually settled) may well be different than the law in a different circuit. If there is a split in the circuit courts, then the Supreme Court may have to weigh in on this issue. When fair use is an issue, the courts look at the facts in context each time. These are obviously…

This could either be: 1. a fascinating Supreme Court opinion. 2. a frustrating ruling because SCOTUS doesn't understand software and code. 3. the type of anti-anticlimactically(?) narrow ruling typical of the Roberts court. While our Congresspersons can't seem to wrap their minds around technology/social media, I think SCOTUS would understand this one enough to avoid (2).

Fair use cases tend to produce narrowly-written law because the outcomes hinge on how the court judges the facts against the list of factors codified in the Copyright Act (17 U.S.C. section 107). The courts don't really have breathing room to use a different test. I don't recall any cases in which the courts have set binding guidelines for interpretation of these factors.

Re: All public GitHub code was used in training Copilot

#335

Post by Luker88 »

Like it or not, it seems like:

* most people here are unhappy

* most laywers will say it's fine (it very probably passed MS ones)

I can understand that. Copyright was not created with AI/ML in mind, even as a random stray thought. Those were not even words at the time.

So the question is: If we change the law and require trained algorithms to only work on licenses that permit this, and to output the "minimum common license" somehow, what are the repercussion on other application of copyright?

Because the consensus here seems to be that this looks a lot like a de-licensor with extra steps

Re: All public GitHub code was used in training Copilot

#336

Post by shakna »

AaronFriel wrote: post #330
shakna wrote: post #286

Earlier quoted context omitted.

> Is there any evidence of Copilot producing substantial (100s of lines) verbatim copies of copyrighted works? It doesn't need to be substantial. In Google v. Oracle a 9-line function was found to be infringing.

If I recall correctly, the nine line question wasn't decided by the supreme court, but the API question was. The Supreme Court did hold that the 11,500 lines of API code copied verbatim constituted fair use. https://www.supremecourt.gov/opinions/20pdf/18-956_d18f.pdf

> The Supreme Court did hold that the 11,500 lines of API code copied verbatim constituted fair use.

Yes, because it was _transformative_, in a clear way. Because an API is only an interface. Which makes that part of that decision largely irrelevant to the topic at hand.

> Google’s limited copying of the API is a transformative use. Google copied only what was needed to allow programmers to work in a different compu-ting environment without discarding a portion of a familiar program-ming language. Google’s purpose was to create a different task-related system for a different computing environment (smartphones) and tocreate a platform—the Android platform—that would help achieve and popularize that objective.

> If I recall correctly, the nine line question wasn't decided by the supreme court, but the API question was.

It was already decided earlier, and Google did not contest it, choosing instead to negotiate a zero payment settlement with Oracle over the rangeCheck function. There was no need for the Supreme Court to hear it.

Re: All public GitHub code was used in training Copilot

#337

Post by elliekelly »

rambambram wrote: post #188

I'm a programmer and also studied law for some time. These stories make me - once more - realize the old adage: "Possession is nine tenths of the law." Don't host that code in the cloud (or a better term, someone else's dirty bucket). What happened to developers hosting stuff on their own website!?

Anecdata: I’m a lawyer and programmer and my clients (large financial institutions) are increasingly insisting on hosting as much on-site as possible. It costs more, it can make it difficult to select vendors/service providers, and it’s not without business continuity risks which they take steps to mitigate.

But I think more and more companies, particularly those in highly regulated industries, are deciding that the benefit of controlling the data — access, security, privacy, and understanding who, exactly, it’s being shared with — outweighs the risks of someone else having that control.

Re: All public GitHub code was used in training Copilot

#338

Post by treesprite82 »

wpietri wrote: post #127
_Understated_ wrote: post #92

Ok, my curiosity has been fired here... I have conjured up two scenarios here: Let's say I use copilot to generate a bunch of code for an app, something substantial, and it regurgitates a load of bits and pieces from many sources it got from GitHub, I'd assume there won't be any attribution in it... it will be as if Copilot made the code itself (I know it sort of does but lets not split hairs!). I'm guessing the prev…

I think you're right. Especially given that Copilot can reproduce significant blocks of code: https://twitter.com/mitsuhiko/status/1410886329924194309 Famous code: https://en.wikipedia.org/wiki/Fast_inverse_square_root#Overv...

I see this held up as an example a lot, but the fast inverse square root algorithm didn't originate from Quake and is in hundreds of repositories - many with permissive licenses like WTFPL and many including the same comments.

GitHub claims they didn't find any "recitations" that appeared fewer than 10 times in the training data. That doesn't mean it's a completely solved issue (some code may be repeated in many repositories but always GPL, and there are limitations to how they detect recitations), but from rare cases of generating already-common solutions people seem to be concluding that all it does it copy paste.

Re: All public GitHub code was used in training Copilot

#339

Post by BuildTheRobots »

jefftk wrote: post #275
rambambram wrote: post #188

I'm a programmer and also studied law for some time. These stories make me - once more - realize the old adage: "Possession is nine tenths of the law." Don't host that code in the cloud (or a better term, someone else's dirty bucket). What happened to developers hosting stuff on their own website!?

It doesn't matter where the code is hosted, just that it is publicly accessible. If developers hosted code on their own sites, someone could still scrape them and use that to train models. (The question of whether this is sufficiently transformative to count as fair use is still wide open)

> It doesn't matter where the code is hosted, just that it is publicly accessible. If developers hosted code on their own sites, someone could still scrape them and use that to train models.

I'd suggest it makes it more interesting. If it's self hosted, then the hoster can choose to impose restrictions on server aceess, including no automated scraping, rather than trying to impose licensing on the code itself.

Re: All public GitHub code was used in training Copilot

#340

Post by akerro »

sailingparrot wrote: post #304
agilob wrote: post #134

This means that also all illegally leaked codes from Apple, CDPR, Intel, NSA and Microsoft leaks are used in the models? iBoot, Witcher 3? Gwent? NSA backdoors? Does the copilot still learn from new repos? Can I post github enterprise code publicly to let it learn from it? Serious answers only please

No, it's not trained on all public code as the title suggests, it's trained on all GitHub public code (so public repos hosted on GH), none of the things you enumerate are hosted on GH.

Just found Intel leaks and Gwent on github without any effort. Intel has a few repositories in different formats, plain copy of .svn directory or converted to git. TF2/Portal leak is there as well. All but 2 I found were made by throwaway accounts.
Post reply on HN