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

GitHub Copi­lot inves­ti­ga­tion

githubcopilotinvestigation.com

Post reply on HN
Print view View original

661–670 of 1001 posts

Re: GitHub Copi­lot inves­ti­ga­tion

#661

Post by ShamelessC »

smaudet wrote: post #628
reflexe wrote: post #585

Earlier quoted context omitted.

Maybe the solution here is do adopt the approach from humans: if you independently produce someone's copyrighted work and then discover about it - you'll drop it. The same approach could be used here, they can add a check for the similarity between the output and the original training material, if it is above a threshold, they'll drop the suggestion (maybe they are already doing that).

That doesn't work, though - the problem is you have automated crime. When you do this, you can no longer handle this on a case by case basis - you have to resort to automating justice. And at this point, you are getting both attacked and supported by AI, and not really better off in a meaningful way. There's no good way of solving this issue, without general intelligence, and the problems that will bring (what reason…

AI research is basically just “the history of computer science”. Alan Turing’s Imitation Game being an apt example.

I don’t think AI research in a vacuum is as deeply unethical as you suggest. It’s about current societal context- people won’t like being worse at things; resources will be hoarded rather than distributed.

Re: GitHub Copi­lot inves­ti­ga­tion

#662

Post by ummonk »

Fordec wrote: post #311
ummonk wrote: post #280

Earlier quoted context omitted.

If they hire photographers to take photos of people in public and use them for training, there’s no law stopping them really. Your only real hope would be to always walk around in a burqa.

There are laws covering that use case. It just depends on the country. Assuming your countries laws is the law everywhere is a bit of a fallacy.

Good point

Re: GitHub Copi­lot inves­ti­ga­tion

#663

Post by cdrini »

remram wrote: post #539
oefrha wrote: post #530

Earlier quoted context omitted.

I use Copilot all the time and I’ve never once used it to generate a whole prepackaged function that’s more than maybe three lines. So no, I don’t benefit from its reproducing other people’s code at all. Tell me you don’t use Copilot without telling me about it.

> Tell me you don’t use Copilot without telling me about it. You don't accept arguments against the use of Copilot from people unless they... use it? That's a nifty way to ignore any and all criticism of Copilot, or indeed any discussion about any ethical issue ever.

I believe the argument is that you shouldn't accept arguments against the use of copilot from people unless they have tried to use it. In a realistic context. That seems reasonable to me. It's the bare minimum to make an informed opinion. I think the wording was perhaps poor, but I think your interpretation is a little reductive/disingenuous.

Re: GitHub Copi­lot inves­ti­ga­tion

#664

Post by iroh2727 »

naet wrote: post #640
schoen wrote: post #486

Here are a few thoughts I haven't formulated before: It seems clear enough to me that training AIs on copyrighted works is typically or commonly a fair use under existing law, because the AIs can and commonly do learn non-copyrightable elements and aspects of those works. It's very obvious from enormous numbers of examples that current AI systems are capable of learning much more abstract features of human culture (g…

I see this basic logic in almost AI ethics threads, and it starts with a big assumption: "humans learn from copyrighted source material without copyright violation". This then gets tenuously extended to "ai also learns, so it must not be in violation of copyright law". The first assumption is highly flawed though. Humans routinely do violate copyright law. Plagiarism is a huge problem in many sectors; un-cited direct…

+1. And let's not forget too that "AI", that is, ML models, are not "autonomous" in the way that humans are autonomous. Sure, we use the word "learn" to describe what they do, which is one word that we also use to describe what people do. But ML models are always wielded by people or corporations for particular purposes.

If a corporation was to directly publish some copy that appears plagiarized, we'd call that plagiarism. I don't see how adding a piece of code—one that's fully created, owned, and wielded by the corporation—as an intermediary changes anything. If anything, it looks like plagiarism-as-a-service, which seems worse (at least to my eyes).

Of course, this matter is a bit confusing. Because, for example, (1) it's not always plagiarism, (2) defining what exactly is plagiarism even in the purely non-technological realm is difficult (and likely somewhat subjective), and (3) there is a lot of corporate marketing which suggests this "AI" is "autonomous" (presumably to distract from who exactly is autonomous in this picture). And of course ML art is quite useful for many things. But I mean, so are artists.

Not long ago, a lot of Silicon Valley rhetoric was that the purpose of "technology" was to free up time so that people could be more incentivized to "do what people love to do" like, for example, artistic creation. But now it seems that rhetoric was just that: rhetoric, or what was needed to be believed/said at the time.

And now at our present time, when technological "progress" has been followed a bit further (that is, when we've developed our machinery a bit further under the incentives of our present economic system), much rhetoric has conveniently shifted to something else, something largely contradictory, but again precisely to what is needed to be believed/said to continue following the same incentive structure.

Re: GitHub Copi­lot inves­ti­ga­tion

#665

Post by zarzavat »

naet wrote: post #640
schoen wrote: post #486

Here are a few thoughts I haven't formulated before: It seems clear enough to me that training AIs on copyrighted works is typically or commonly a fair use under existing law, because the AIs can and commonly do learn non-copyrightable elements and aspects of those works. It's very obvious from enormous numbers of examples that current AI systems are capable of learning much more abstract features of human culture (g…

I see this basic logic in almost AI ethics threads, and it starts with a big assumption: "humans learn from copyrighted source material without copyright violation". This then gets tenuously extended to "ai also learns, so it must not be in violation of copyright law". The first assumption is highly flawed though. Humans routinely do violate copyright law. Plagiarism is a huge problem in many sectors; un-cited direct…

The law does not concern itself with trifles.[0]

Programmers tend to think of copyright as a Boolean valued function. Either something is infringement or it isn’t.

Judges think of copyright infringement as a real-valued function of many arguments corresponding to the circumstances of the parties (e.g. what actual damage was done?).

A human quoting a human without attribution, without any profit made or identifiable damage, returns an infringement value very close to zero. Such cases, if anyone is petty enough to bring them, are likely to get dismissed.

[0] https://en.m.wikipedia.org/wiki/De_minimis

Re: GitHub Copi­lot inves­ti­ga­tion

#666

Post by sooyoo »

iepathos wrote: post #554
schoen wrote: post #486

Here are a few thoughts I haven't formulated before: It seems clear enough to me that training AIs on copyrighted works is typically or commonly a fair use under existing law, because the AIs can and commonly do learn non-copyrightable elements and aspects of those works. It's very obvious from enormous numbers of examples that current AI systems are capable of learning much more abstract features of human culture (g…

Humans do violate copyright if they use copyrighted passages directly in their work and pass it off as their own without any attribution, which is what copilot has been show to sometimes do, though not always. Copilot will sometimes offer chunks of code that can be found verbatim in open source code bases and passes it off to users without attribution. I agree it is ok to learn from copyrighted work and reproduce new…

> Microsoft will likely need to add checks to prevent copilot from offering verbatim copies of code going forward to try to avoid copyright violations here.

Or they could integrate a way to find the produced output back in the corpus if it's sufficiently close and provide a reference/attribution. Basically whatever tool a copyright lawyer would use to track down original work.

And that's just the engineering solution. The AI researcher solution would be to extend AI learning algorithms to attach attribution metadata to the learned data so that the output could already come annotated with information about the source.

But the latter is much harder to do, so maybe the engineering solution would suffice.

Re: GitHub Copi­lot inves­ti­ga­tion

#667

Post by sdenton4 »

schoen wrote: post #486

Here are a few thoughts I haven't formulated before: It seems clear enough to me that training AIs on copyrighted works is typically or commonly a fair use under existing law, because the AIs can and commonly do learn non-copyrightable elements and aspects of those works. It's very obvious from enormous numbers of examples that current AI systems are capable of learning much more abstract features of human culture (g…

I really appreciated the argument in the book "The New Breed," which is that we should adapt ideas around the governance of animals to governance of ML: You can train your dog to attack random passersby, but if you do, you're a monster and ultimately responsible for the dog's actions.

Likewise, you can tell Copilot to crank out code specific algorithms written by specific people, but if you do so, you're still creating infringing code, same as if you'd taken the more direct route of ctrl-c+ctrl-v. The fact that you /can/ make the algorithm misbehave through adversarial input is irrelevant to the primary use cases which lead to boring non-infringing code completions.

Re: GitHub Copi­lot inves­ti­ga­tion

#668

Post by diffeomorphism »

Vetch wrote: post #642
iepathos wrote: post #554

Earlier quoted context omitted.

Humans do violate copyright if they use copyrighted passages directly in their work and pass it off as their own without any attribution, which is what copilot has been show to sometimes do, though not always. Copilot will sometimes offer chunks of code that can be found verbatim in open source code bases and passes it off to users without attribution. I agree it is ok to learn from copyrighted work and reproduce new…

> which is what copilot has been show to sometimes do In those cases it seems that humans are already copying code without also propagating licenses appropriately. LLMs are more likely to memorize things which occur a lot (and I'd bet rare things that are representative of some conceptual axis). The main examples presented so far, Davis and Carmack, have the property of having been copied a lot. The generative model…

That seems like a weak defense: "sure, we violated copyright but only because many other people do, too".

Kinda the same problem as YouTube. Lots of people copy movies on the high seas, but if you are as big as yt you cannot easily get away with it.

Re: GitHub Copi­lot inves­ti­ga­tion

#670

Post by thesz »

schoen wrote: post #486

Here are a few thoughts I haven't formulated before: It seems clear enough to me that training AIs on copyrighted works is typically or commonly a fair use under existing law, because the AIs can and commonly do learn non-copyrightable elements and aspects of those works. It's very obvious from enormous numbers of examples that current AI systems are capable of learning much more abstract features of human culture (g…

Your comment is full of good and deep analysis and it uncovers a deep flaw in the Copilot: Copilot helps create new code.

All the problems and confusions mentioned above is due to this concrete inherent property of the Copilot.

If Copilot was made to help rearrange [1] existing code to satisfy new or changed needs, there would be no need in such a deep and explanative analysis of yours.

[1] https://www.folklore.org/StoryView.py?story=Negative_2000_Li...

Code is a liability. Less code is less liability. New code is a new liability.

Even tools to create new code is a new and unknown liability, it seems.

Post reply on HN