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

821–830 of 1001 posts

Re: GitHub Copi­lot inves­ti­ga­tion

#821

Post by i_like_apis »

BowBun wrote: post #794

I'm in favor of this. You can't ingest code that says "you cannot use this without attribution", put it through a bunch of if statements that strip the license, and then say it's "AI-generated". I don't care about most of our generic CRUD apps or the 15th rewrite of a sorting algorithm, but I do care about those smart enough to advance the field and come up with novel solutions. If we take away the incentive for attr…

Totally disagree. Training is fair use. It is akin to learning. Code licenses do not restrict you from reading or learning.

ML training needs to be fair use of copyrighted works, or most machine learning and AI projects will be impossible.

Re: GitHub Copi­lot inves­ti­ga­tion

#822

Post by i_like_apis »

nilshauk wrote: post #820

So happy to learn of this and I wish them best of luck in their efforts. And I'm surprised to find so many people klinging to Copilot. We shouldn't shed any tears for a megacorporation which shows such blatant disregard for the licensed works of people's labour. Yes, AI is here to stay but we should be able to build AI that respects copyright. Yes, it's easier to just steal data and call it fair use. Whether or not t…

Foolish take. If ML training is not fair use then all ML progress is dead in the water.

ML training is akin to reading or learning, and licenses do not apply to that.

You’re not thinking past “megacorp = bad”.

Re: GitHub Copi­lot inves­ti­ga­tion

#823

Post by moffkalast »

mkr-hn wrote: post #529

A sizable, possibly plurality cohort of fully adult tech people is young enough to not know about United States v. Microsoft Corp. This would explain a lot of comments I see on this topic. If you don't know Microsoft's history, a lot of what more informed people are worried about seems overblown. Copilot was Microsoft's first test of people's trust after the GitHub acquisition. It's going very, very, very poorly. The…

Microsoft forgot that we live in a society.

Bottom text provided by copilot.

Re: GitHub Copi­lot inves­ti­ga­tion

#824

Post by Roark66 »

Well I am "a fan" of Copilot and I do think AI is the future, but I think the author has a valid point.

I think the fair use violation he describes doesn't happen during training. I do think training AI on anything that is publicly accessible is fair use just as in an example of a person learning by reading/watching the same materials.

However, this fair use rule is being violated the moment the resulting AI starts suggesting verbatim copied code from licensed works without attribution.

So one could argue the source code is not being used in a transformative way but copilot is just more efficient method of retrieval of licensed code. This misses the fact copilot actually is capable of writing new code. I've used it as "an autocomplete on steroids". Letting it suggest maybe half a line, or 1 line of code at a time (or trivial stuff we automate even without copilot like getters/setters in java). But when actual licensed code is suggested yes, this is IMO a license violation.

Therefore one way of resolving this would be to pair copilot with a tool that scanned the resulting code for presence of licensed code then it woukd make a list of "credits" or references. Also there should be measures taken (perhaps during training) to penalise generation of verbatim (or extremely similar) code. Would this make copilot less of a useful tool? I'm not sure.

One thing that's not going to happen is putting tools like copilot back "in the bottle". We now have similar models anyone can download (faux pilot) and I as well as many others have found those tools to speed up mundane tasks a lot. This translates into monetary advantage for users. Therefore there is no way this will disappear, lawsuit or no lawsuit.

Re: GitHub Copi­lot inves­ti­ga­tion

#825

Post by bheadmaster »

i_like_apis wrote: post #821
BowBun wrote: post #794

I'm in favor of this. You can't ingest code that says "you cannot use this without attribution", put it through a bunch of if statements that strip the license, and then say it's "AI-generated". I don't care about most of our generic CRUD apps or the 15th rewrite of a sorting algorithm, but I do care about those smart enough to advance the field and come up with novel solutions. If we take away the incentive for attr…

Totally disagree. Training is fair use. It is akin to learning. Code licenses do not restrict you from reading or learning. ML training needs to be fair use of copyrighted works, or most machine learning and AI projects will be impossible.

You are assuming that "training" and "learning" in ML means exactly the same thing as "training" and "learning" in humans. It doesn't. The processes are completely different, with only an apparent resemblance.

Re: GitHub Copi­lot inves­ti­ga­tion

#826

Post by i_like_apis »

This is such a disingenuous use of the word “investigation”.

They are “investigating” whether they should start a law suit. So this is not an investigation, it’s somewhere between “due diligence” and a PR stunt.

I very much disagree with the idea of a law suit that seeks to establish ML training as not being fair use. It is an utterly foolish thing for them to wish for.

Re: GitHub Copi­lot inves­ti­ga­tion

#827

Post by woodson »

kjeksfjes wrote: post #809
rhdunn wrote: post #730

Earlier quoted context omitted.

It depends on what the AI is learning. If the AI is learning to repeat text (e.g. Copilot) or images (e.g. Dall-E), then that makes it possible to reproduce the copyrighted works, so I would agree that that case is not fair use. -- It would be akin to compressing and distributing those works. If the AI is learning patterns -- such as "muggle" being a noun that relates to Harry Potter, or that the lemma for "muggles"…

I'm not sure how Copilot works, but I don't believe Dall-E repeats images. From my understanding it creates visual concepts of words and uses them to create entirely new images. If Copilot works in the same way for code, I honestly don't see that there should be any copyright issues here.

It just so happens that, sometimes, parts of these entirely new images are exact copies of those used for training.

Re: GitHub Copi­lot inves­ti­ga­tion

#828

Post by Timshel »

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…

I agree it's like people are not aware that clean room design is a thing (https://en.wikipedia.org/wiki/Clean_room_design).

Re: GitHub Copi­lot inves­ti­ga­tion

#829

Post by b4je7d7wb »

i_like_apis wrote: post #821
BowBun wrote: post #794

I'm in favor of this. You can't ingest code that says "you cannot use this without attribution", put it through a bunch of if statements that strip the license, and then say it's "AI-generated". I don't care about most of our generic CRUD apps or the 15th rewrite of a sorting algorithm, but I do care about those smart enough to advance the field and come up with novel solutions. If we take away the incentive for attr…

Totally disagree. Training is fair use. It is akin to learning. Code licenses do not restrict you from reading or learning. ML training needs to be fair use of copyrighted works, or most machine learning and AI projects will be impossible.

Most AI projects being impossible sounds better to me than license violations.

Re: GitHub Copi­lot inves­ti­ga­tion

#830

Post by i_like_apis »

bheadmaster wrote: post #825
i_like_apis wrote: post #821

Earlier quoted context omitted.

Totally disagree. Training is fair use. It is akin to learning. Code licenses do not restrict you from reading or learning. ML training needs to be fair use of copyrighted works, or most machine learning and AI projects will be impossible.

You are assuming that "training" and "learning" in ML means exactly the same thing as "training" and "learning" in humans. It doesn't. The processes are completely different, with only an apparent resemblance.

No, they aren't completely different. Learning is learning.
Post reply on HN