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

471–480 of 734 posts

Re: All public GitHub code was used in training Copilot

#471

Post by TheSpiceIsLife »

jrm4 wrote: post #316
extra88 wrote: post #296

Earlier quoted context omitted.

> Is it Novel? Is it Non-Obvious? Those questions for patents are barely more clear-cut than copyright fair use tests, there is lots of room for disagreement. It's definitely true that a fair use defense against copyright infringement varies a lot by the field of work and norms can develop which are relevant to court cases. The music field is a mess, the "Blurred Lines" judgement was total bullshit. But the software…

But there's no reason not to either - I suppose my point is, don't take too much as gospel and think about everybody's best "end-goals" and push or pull with or against the law as needed.

There’s also an aspect of this that varies by size, budget, political clout, etc etc, of the individual or organisation.

The big guns like Microsoft, Google, Oracle, do this sort of thing as a matter of course in their business activities, they have the lawyers, the money, and the ear of members of parliaments, senators etc.

Whereas an individual or small business probably wants to conduct themselves within a more narrow set of adherences.

Re: All public GitHub code was used in training Copilot

#472

Post by encryptluks2 »

ostenning wrote: post #11
Tenoke wrote: post #8

Earlier quoted context omitted.

If you have public repos anywhere people can train on them just as much.

That's also my general sentiment. I assume anyone can do virtually anything with my public repos with little recourse from me. I wouldn't even know if they are indeed breaking my license agreements. Doesn't really help the situation though.

GitHub only recently allowed non-paid private repos. Previously these were only reserved for paid plans. Also, GitHub has a specific section for license files. GitHub actually believes these license files mean something, and states that they must be included with the repo so they are downloaded with it. Just because you can teach a script to ignore a LICENSE file, doesn't mean that it still doesn't apply. That is like saying that because you can teach a robot to ignore restricted airspace, that it is allowed to fly around an airport.

Re: All public GitHub code was used in training Copilot

#473

Post by qayxc »

mehdix wrote: post #440
qayxc wrote: post #309

Earlier quoted context omitted.

Why and how? I'm honestly interested in an answer here. What exactly is the difference between a machine learning patterns and techniques from looking at code and people doing it? Is every programer who ever gazed at GPL'ed code guilty of plagiarism and licensing violations because everything they write has to be considered derivative work now?

I can think of certain things here. As human beings we have limitations. We get tired of gazing at code, GPLE'ed or not. GitHub's clusters don't. It puts fair use of copyrighted content under question. The next concern I have, is what happens when Copilot produces certain code verbatim? I saw the other day on HN that it produced some Quake code verbatim. See https://news.ycombinator.com/item?id=27710287

> As human beings we have limitations.

That's a fair point. ML models don't seem memorise all the code they've seen either, it seems. Plus while the argument of human limitations applies to the vast majority of people, what about those with eidetic memory?

> what happens when Copilot produces certain code verbatim?

There are several options: suppress the result, annotate with a proper reference or mark the snipped as GPL'ed.

There are technical solutions to this question, but it's also important to ask to which degree this is necessary.

Is a search engine that returns code snippets regardless of license also a tool that needs to be discussed the same way? After all, code samples from StackOverflow or RosettaCode are copied on a regular basis and not every example provides a proper reference as to where it's been taken from.

So maybe a hint like "may contain results based on GPL'ed code" suffices? I don't know, but that's a question best deferred to software copyright law experts.

Re: All public GitHub code was used in training Copilot

#474

Post by svaha1728 »

nonfamous wrote: post #380
shakna wrote: post #301

Earlier quoted context omitted.

Transformed output _may_ fall under fair use. However - Copilot directly recites code. That is _very unlikely_ to fall under fair use. Redistributing the exact same code, in the same form, for the same purpose, probably means that Copilot, and thus the people responsible for it, are infringing.

> However - Copilot directly recites code. You make that statement as an absolute, but in the interests of clarity, all evidence so far shows that it directly recites code very rarely indeed. Even the Quake example had to be prompted by the specific variable names used in the original code. In practice, the output code is heavily influenced by your own context — the comments you include, the variable names you use, e…

So if a foreign company pilfers the source code to Windows, can they add it to a training set and then 'prompt' the machine learning algorithm to spit out a new 'copyright free' Windows, just by transforming the variable names?

Re: All public GitHub code was used in training Copilot

#475

Post by leventov »

habitue wrote: post #353

I am really confused by HN's response to copilot. It seems like before the twitter thread on it went viral, the only people who cared about programmers copying (verbatim!) short snippets of code like this would be lawyers and executives. Suddenly everyone is coming out of the woodworks as copyright maximalists? I know HN loves a good "well actually" and Microsoft is always suspect, but let's leave the idea of code la…

Perhaps people on HN start sensing that successors of Github Copilot will take their programming job. Rightly so.

Personally, I think that in the age of AI programming any notions of code licensing should be abolished. There is no copyright for genes in nature or memes in culture; similarly, these shouldn't be copyright for code.

Re: All public GitHub code was used in training Copilot

#476

Post by slownews45 »

loeg wrote: post #368
slownews45 wrote: post #349

Earlier quoted context omitted.

Developers have lost the plot here. The number of people browsing stack exchange and copying code is huge. The number of people who have read GPL'ed code to learn from (from the kernel to others) is huge. The number of people who learned from code they had to maintain -> huge. This idea that a snippet of a code is a work seems crazy to me. I thought we went through this with SCO already.

Stack exchange code is explicitly permissively licensed.

It is, but it has a GPL style permission.

ShareAlike — If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original.

The idea that programmers taking snippets from stackexchange or co-pilot etc meaning they have a derivative work seems like total insanity.

Re: All public GitHub code was used in training Copilot

#477

Post by data_ders »

BoxOfRain wrote: post #467
neonihil wrote: post #457

I think it’s pretty easy to defeat MS in court. We just need to bring the music industry into this! For example: Let’s train a network on Beatles music to generate new Beatles songs. I’m pretty sure music lawyers will find a way to prove that the trained network is violating the label’s copyright, as they always manage to do that. And then we just need to use the precedent and argue that music is the same thing as co…

There is an absolutely enormous archive of fan-taped Grateful Dead shows out there, someone with much more time and money than me needs to train a network on that!

username checks out lol

Re: All public GitHub code was used in training Copilot

#478

Post by partiallypro »

neonihil wrote: post #457

I think it’s pretty easy to defeat MS in court. We just need to bring the music industry into this! For example: Let’s train a network on Beatles music to generate new Beatles songs. I’m pretty sure music lawyers will find a way to prove that the trained network is violating the label’s copyright, as they always manage to do that. And then we just need to use the precedent and argue that music is the same thing as co…

That already exists though? SongSmith and other similar tools are used by musicians a lot.

Re: All public GitHub code was used in training Copilot

#479

Post by albertzeyer »

So, when a human reads public code on the Internet (no matter the licence), and gains knowledge, learns (updates the synaptic weights of the brain), and then makes (indirectly) use of that gained knowledge for further work, how is this different to this case?

Re: All public GitHub code was used in training Copilot

#480

Post by user5994461 »

TuringNYC wrote: post #165
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

IANAL but the serious answer -- i think -- is that you always use things at your own risk, even purchased tools, and are protected via indemnity agreements. If there is no indemnity agreement (is is the case here), you assume the risk. That said, if enough people are bitten by this, i'm not sure what happens -- does anyone know of a relevant case. One somewhat relevant case that caused mass pain was the SCO Linux Dis…

If you're thinking about the liability waiver found in many licenses and contracts and EULA and other, they are often void, depends on the jurisdiction.

The official answer from Github that they take all input on purpose doesn't play in their favor.

Post reply on HN