Copilot wouldn't be shut down or neutered because "a few vocal people" protested against it.
It would be because it's illegal and violates the licenses, desires, and intentions of the thousands of workers who wrote the code in its corpus.
You act like Microsoft is trying to do a public service and people are angry about it. The reality is that they're taking billions of hours of work and using it to build a product that only they control.
If they re-released Copilot as FOSS, a lot of the valid criticisms would evaporate.
They should in addition release all code generated as a combo of AGPL, GPL, MIT, etc. and put a comment on every usage. Users would then need to license their code accordingly.
For a commercial version, run it on Microsoft's internal code, the code they actually own!
> If they re-released Copilot as FOSS, a lot of the valid criticisms would evaporate.
That changes nothing at all. A FOSS license is, in a legal sense, no different then a proprietary one. If it’s illegal it’s illegal regardless of whether it’s FOSS.
If Copilot is deemed a derivative work of its inputs, it would resolve the issue that it does not comply with the AGPL because it currently does not provide source code to its users under an AGPL-compliant license.
> It would be because it's illegal and violates the licenses, desires, and intentions of the thousands of workers who wrote the code in its corpus.
I wonder how many people on HN would be on the side of the creators if we were talking about content created by Walt Disney and whether pirating was ethical?
I am 100% on the side of content creators. Regardless of who they are .
The courts tend to take a dim view of theft. Which is what this is.
The article clearly lays out that multiple requests for sound legal basis have gone unanswered . It simply doesn’t exist and Microsoft is operating on a forgiveness vs permission model.
Licensing is 100% about permissions. Clear and explicit enumeration of the permissions (or lack thereof ) for a work.
This class action lawsuit should surprise nobody. It’s a class that is sick and tired of being exploited.
Do not take my work that I contributed with explicit permissions and use it in a way I didn’t grant permission for. Full stop. It isn’t complicated.
If my code is licensed under terms , that’s the permission .
If you use my licensed work then yes, you do need to follow the terms of the license .
The issue of license / contract / copyright is messy. It doesn’t ever seem (in the USA anyway ) to be definitively answered / “solved.”
I chose AGPL v3 only on purpose.
Co pilot and users thereof (so now two levels removed ) utilizing code in whatever work are stealing my work (unless it’s AGPL v3 licensed ). The adding of intermediaries (and the most likely unknown and with no way to know) infringing is going to be very difficult to mitigate. It’s like truly unknowingly buying stolen property,
If I use it under fair use, there is nothing you can do about it.
The fourth factor of fair use is the effect on the value of your work. If I'm not affecting the works value because there is no market for it, because it has no commercial value, you are going to have a very hard time defeating this argument in court.
so much animosity over rights you gave GitHub when you put your code there. "Theft" gimme a break. you license your code to GitHub so they can show it to others. This is separate to the stated license in your code. Nowhere in that terms of service document are the means that the code is shown to users specified.
Give me a break. Did you even read the license? You read that it is restricted to "the service". That service cannot be copilot, because it didn't even exist when most people agreed to the license. Moreover the license explicitly states they cannot use the code to distribute in any other way or sell it. So if anything they are violating their own agreement.
Also MS themselves don't even claim that training is covered by their terms of service, they claim it is fair use.
? Did you read what I wrote, MS doesn't even claim copilot is covered by their terms.
They claim it is covered by fair use (some people also claimed there is code not hosted by GitHub in copilot, which would further confirm that they believe they are covered by something else)
Moreover terms have been largely unchanged for years AFAIK. If someone agreed to the license years ago, they can't have agreed to copilot use. Also copilot is not a service on their website, it is a separate service and they charge for it, also contradicting the terms.
If you think I'm mistaken I will gladly reevaluate what I said if you could kindly restate your position, by answering a few questions
What does separate service mean? What would Copilot look like if it was not separate?
Elaborate on "not a service on their website", as it is available and listed as a feature "Github Copilot" on their website.
Is the contradiction related to payment for the service, or just because you think it is separate?
Since I thought you were arguing that Githubs own terms prevent them from using the public repositories in Copilot, this is what I argued against.
If you think fair use is involved, then that's the end of the line. If MS claims fair use, then until a court says otherwise, it is. Anyone who thinks their copyright is being violated can get an injunction tomorrow.
they claim the training of their model is covered by fair use, but they did not say that was the justification they were using. They don't need to claim fair use.
It's pretty clear from the Terms of Use that they can use code hosted on github.com to provide any service they like, so long as it is a GitHub service. They don't need fair use, they already have the rights to do what they are doing.
There is a certain amount of irony that a likely possible outcome will be the strengthening of copyright laws and enforcement, at which point people will realize they won the battle but lost the war
But isn't it copyright upon which open source is founded? The license is based on the right of the owner of the copyright to authorize someone else to create a derivative work?
Without copyright, it would be perfectly legitimate and legal for someone to not follow a license (because it would bear no legal weight because of the lack of copyright).
I would argue that open source is best served by strong copyright protections that allow the people created the software to make sure that further changes to it are released back to the community. Weakening copyright law means that it is that much easier for big companies to co-opt some software and not need to release the changes back.
That's a bit of an exaggeration. There have only been two copyright term extensions since Mickey Mouse was created, and only one of those can even remotely be attributed to Disney lobbying.
Even for the term extension that is attributable to Disney, you can also attribute it to Germany and the formation of the EU. "Mickey Mouse Protection Act" is funny, yes, but also an oversimplification.
This is also why I fully expect Steamboat Willie to fall out of copyright protection in January 2024 - right on schedule. There's a few countries that have supra-EU copyright terms, but none of them are dealmakers. Nobody is demanding we match Mexico's life+100 terms, for example.
I don’t think anyone at all is here arguing that an AI trained on a massive corpus of movies that outputs snippets of film based on a prompt would be illegal. Such a thing would literally be the same as Midjourney with images. The fact that you can likely coax any AI to output snippets close to some of the source material is not likely to really matter and would be as if you recreated a copyrighted work using any other tools.
That many people who are criticizing Microsoft for using open source code and claiming “fair use” and saying it’s not fair to creators are the same ones that say pirating digital content is harmless.
Because issues aren't automatically symmetrical, and impact of actions isn't equal in both ways, especially when there's power difference between the sides of the conflict, e.g. it's not hypocritical to punish a bully punching their victim, but not the victim for punching the bully back.
Disney used its power to distort copyright laws in a self-serving manner. As an individual you don't have equal power to oppose them (you were supposed to have in a democracy, but lobbying is legal and corporations are people).
Disney is a huge corporation that won't even notice if you pirate a movie, which you may not even have been able to pay for anyway, because of their region-locked twisted maze of distribution and DRM.
OTOH you may be screwed if you're a creator making a living from your work, and a big corp can just take it without paying, launder it through "… in the style of $YOURNAME" query, and say they own it now, because unlike your copyright, their Terms Of Service apply.
Even people who think copyright shouldn't exist may rely on using copyright — against itself. You can't unilaterally say "I don't believe in copyright", because the law doesn't care, but if you license something as copyleft, then the law does care about your anti-copyright license.
There’s an assumption that hackers are automatically anarcho-communist, and unconditionally, wholeheartedly, vocally supportive of content piracy, which is not correct.
I consider GPL as meaning, "don't make the same app as me using this very source code". Cribbing one method out of quarter million line code-base, hardly seems to be redistributing the source code. Literally, it is, but is there no concept scale? Can we go to an extreme and force anyone with a `catch(Exception e)` line in their Java code to go and prove they did not take it from a GPL (or similarly licensed project)? I think this indicates there is a line, at some point it is enough code where you are recreating the functionality of the software - to me that is the thing that matters. I don't give a crap if you use my GPL code to learn from and use any parts of it to build whatever software, but I do care if you recreate the same software using my GPL code.
I would accept a claim of license violation if someone used copilot to autocomplete so many methods from one specific project that you have recreated that original project.
I still think it is a matter of scope. It can still be the case that a relatively small module is not cool to lift, but I think in this case we are still talking about such small subsets of functionality that it is completely divorced from the original software. Like, I could see it if a specific method were really key in some way to a unique application, a very novel solution to a difficult problem - but if that were the case, how can an AI possibly use that for a training model? In other words, the auto-suggestions of an AI are going to be common coding solutions to common coding problems that the AI has seen hundreds of thousands of times. That individual proprietary GPL, unique and novel solution is not really the stuff of an AI suggestion. In other words, the code that co-pilot is going to suggest is going to be non-unique, generic, and not really specific to the overall application at all.
Well... If you do a search you'll find that there are actually lots of lawsuits related to open source license violation, mostly around GPL license. Maybe educate yourself first
Your assertion thst copyright doesn’t matter and that people should “get used to it” is… pretty incorrect on the former (surely there are still pending copyright suits on earth circa 2022) and ill-conceived on the latter (I’ll just rob your house while you sleep and you should just get used to it)
> It would be because it's illegal and violates the licenses, desires, and intentions of the thousands of workers who wrote the code in its corpus.
Copilot makes source code much more open, if you think about it. It implements code reuse in a different way than classes and libraries. It offers its skills equally to everyone, skills learned from everyone.
As for the cost of the API - it's expensive to run large language models, I think the price is justified. But there are free models if you like to run your own.
Why should Copilot engineers, and the company that invests in it, not be rewarded for their incredible product and SaaS offering they spend resources on providing?
They built it using some publicly available resources. These resources are available conditionally, subject to licenses (such ad GPL). Which is fine.
The problem is that their product sometimes produces verbatim copies of licensed works, without attaching licensing information. This not only goes against the licenses under which the original authors made these works available. It can also put the product's users in danger of anything from bad publicity to a copyright lawsuit.
CoPilot is a very interesting research project. It's not yet an acceptably mature product though.
That's not theft. Was the original code deleted? No. Then it's copying. And not even that, is the model replicating the training set like Google search? No. Then it's some kind of derivative work. And especially for Github it's OK because user agreements allow MS to do it.
Is "imagining" the same with "copying"? Does copy-right cover learning-right? Can learning and practicing be restricted by the authors? Can visual styles, algorithms and facts be copyrighted? I say no to all of them.
Human's aren't even allowed to do this. It has to be done clean room - this is not learning, it's copying and has been proven so.
The only implementation that would be allowed is if they had to describe the code via a specification and Codepilot was able to generate it from scratch. It does not do that - it just reorganizes stuff it's already seen, which is a copyleft violation in humans, therefore it should be for a machine created by humans.
That's for patents not for copyrights. All you need to do is make it a little different, generate until satisfied for patent safe code. Another model can also do the patent checking.
I've heard multiple stories from people who can't even look at the Linux source before they do something in their company's kernel because the act of just looking at it compromises them legally. So which is it?
Clean room implementation only happens for very special codes, like codecs or efficient matrix multmul that took billions of trials to develop with AI. Not for 2-3 lines of code snippets that do one simple thing and are already covered online in hundreds of places.
One of the canonical examples of software reverse engineering is Phoenix Technologies producing a compatible BIOS for IBM PC. They did exactly what OP described - had one person look at (public but copyrighted) IBM source and produce an extremely detailed design document, then another team went and wrote a new BIOS from scratch following that document. The issue at hand was copyright, not patents.
Larceny and copyright infringement are two distinct legal categories for two distinct kinds of crimes. One involves a taking of property and the other involves the violation of a government granted monopoly
It’s not about “learning” at all, copilot spits out copyrighted licensed code verbatim directly copied from the source to your project in violation of specific and multitudinous repositories.
They are making rips of other peoples stuff, selling the contents of peoples “books/movies/songs” sans author attribution or album credits etc… to put it in terms you may be familiar with
Vinegar and salt on open wounds. Bad
Is there any nuance for scope?
Software I've licensed as GPL, I'm concerned about the working software being re-used and re-licensed for something commercial. For a given method out of tens-of-thousands, it's very not-germaine to the overall software to the point that I don't see it is a really relevant (but that is my opinion). Though, if someone likes the opening sentence of the encylopedia (or some other giant work), and an AI says, "this is a good opening sentence - does that really make for "ripping" off? Isn't the covered work the larger contents of the encylopedia, rather than an arbitrarily well written opening sentence? Isn't the big part of the work the ensemble?
I'm starting to wonder about these arguments, and whether we've gone into bad faith and hyperbole territory here. Are algorithms subject to copyright? Is it the case that if a GPL work uses a well known algorithm, that GPL work cannot be used as a reference? (Given that algorithms have very limited forms they can take, using an algorithm as a reference is really just copying it. Even translating pseudo-code to code, it's still the same thing).
Can you explain to me how something like using Eulers formula to solve a math problem would not be copyright infringement? A GPL project might use that formula somewhere, but then using that would be a copyright violation?
How about HTML source code, does putting a 'copyright' notice on the webpage make it invalid to then use any of the javascript, even if it has nothing special to do with the domain of the website?
"selling the contents of peoples “books/movies/songs”
Going to this analogy, I don't know if it is really the contents, but more like the first sentence, or even the first few words of that sentence rather than any recognizable subset of that work.
Like, if I have an app that does a spreadsheet, I don't care if you take an implementation of quick sort from my code as reference, but I do care if you use the same and main features of the spreadsheet app that I made.
It’s not illegal, it’s at worst a fancy code search tool that Github has the right to show you the results via the license you grant them when you upload and make public code on Github which is way stronger than other search engines like Sourcegraph have to show public code.
It doesn’t mean you have the right to use any of the code it generates but Copilot itself isn’t illegal in any meaningful sense.
This is definitely not true. When your license requires you bundle said license with any reproductions of the code, and Copilot spits out said code sans license, they are breaking the law.
> 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.
I don’t think it’s accidental that this product is specifically Github Copilot.
But even then I think this is legal overkill. If you use the search box on Github they will display snippets of code from public repositories without the license. Same as what Sourcegraph does same as Copilot does. Nobody here is arguing ripgrep is violating the license by displaying matches without the corresponding license.
The violations are when that code is incorporated into your own codebase, which is happening in none of those examples. If you copy GPL'd code from GH search with a non-compatible license you are still in violation.
Yes but that’s your problem as a user of the tool. It doesn’t make Copilot itself illegal which is what the person I originally replied to was saying.
Yes if you use a tool to violate copyright it’s copyright infringement. If you prod Midjourney into outputting near exact Starry Night that’s on you too.
So far no one has made a compelling case that Copilot itself is violating copyright.
Code search may show snippets, but it's clearly not separated from the rest of its code base, including the license. It may be your problem as a user of that tool if you pirate snippets out of the search results without honoring its license, but GitHub at least didn't distribute it without its license. A human judge would surely determine that a search result page showing a snippet and linking back to the project wouldn't constitute distributing the code without attribution. Copilot is a different matter. There is no way to know where the code came from, whether it's novel or verbatim copy of someone's copyrighted work. Microsoft _is_ distributing code snippets here sans its license.
Codesearch at least provides users the ability to hunt down any licensing concerns. Unless Copilot start spitting out citations (this snippet was generated based on repositories x, y, z, here are links) the users of the tool have no way to verify if they are in any way in violation of the licenses.
The people protesting aren't a "vocal few"; We're the people who made copilot possible. We are frustrated that our work is being used to profit a massive corporation without any compensation and in a way that we at best did not intend to allow and at worst is in direct violation of the terms we set.
Nah, you're definitely the few. Its not a random sample, but an informal survey of my coworkers found no one who would care and generally positive sentiment.
The people who comment on something are disproportionately those who care a great deal.
My point was not that those critical of CoPilot are in a majority, it is that our perspective is important because our labour is what makes copilot possible.
What proportion of its capability is derived from the labor of people who don't like it? I get your point about feeling like an unwilling contributor while github/MS harvests revenue from people who like it. But there's an implication here of being in a critical majority, which I am not convinced is the case.
Perhaps a good compromise is to make it opt-out, if it’s not already. Though even this is just pandering to the developer’s ego. AI writing code is a massive boost in giving users power and thus freedom. Of course, we need to make AI itself FOSS, but I doubt a legal case could be made for that. A more productive path is to clone the model like SD did with Dall-e.
I tried copilot and found it an excellent way to inject subtle bugs into my code. It always had a seemingly plausible guess, that was never correct, and coding turned into a guessing game trying to spot the bugs it had injected and hoping I’d found them all.
If the tech becomes open (and all indicators point to it being open in the near future) then it will become impossible to shut down. This has already happened with Stable Diffusion and the related model leaks.
People's expectations have already been set by this technology, and they are only going to want more. Also, AI researchers are still publishing their work out in the open for anyone to reproduce.
If there was a Copilot model out in the wild like with Stable Diffusion then this ceases to be a valid question, regardless of the model's legality. All it takes is a single leak or decision by another entity to release their own code generation model.
> It's been a massive productivity improvement to our senior devs
I would hate to work at a place where advanced-but-untrustworthy autocomplete would, at all, impact the productivity of a senior engineer.
Not only does this indicate that your senior engineers' productivity is measured poorly (lines of code), but also that your senior engineers are paid to type, rather than to think.
It doesn't indicate either of those things. It's a great tool when you're working on complex code that just reaches the limit of your working memory/attention, and has often suggested good and clean improvements for me. You may dislike it but there are people it helps.
And if the code generated by copilot was attached to a license that you had to obey? Suddenly your propriety solution must be released as open source or rewritten, because copilot is effectively laundering open source code?
Life's a lot easier when you can just copy whoever did the hard work without crediting/paying/etc for it.
It's been a massive productivity improvement to our senior devs, and I got so used to it that it's an annoyance when Copilot doesn't respond.