Copilot sells code other people wrote • Hacker News Board

Re: Copilot sells code other people wrote

#751

Post by HeavyStorm » Fri, Jun 24, 2022, 1:16 AM UTC

visarga wrote: post #613

sirsinsalot wrote: post #578

Earlier quoted context omitted.

"A commons of knowledge is a public good." Yes but this copilot model takes that, adds value and doesn't itself join the public common good. Instead it takes it, and makes you pay to have it back in another form. If copilot were open source and the model released for the public good, being built of public data (in your scenario) we would have a very different conversation.

It costs money to run a huge language model with low latency, in the loop with you - charging 10$/month is reasonable. You need multiple GPUs to load even a single copy. Copilot is adding something extra to the original code - it selects the recommendation from the whole corpus, while keeping the surrounding context into consideration and adapting to your variable names. And in reality 99.9% of the generated code has…

Thank you for you this. I wouldn't never been able to articulate it better - people are just annoyed that someone is making m money and they aren't, without considering why that is.

Re: Copilot sells code other people wrote

#752

Post by HeavyStorm » Fri, Jun 24, 2022, 1:17 AM UTC

sirsinsalot wrote: post #658

andybak wrote: post #582

Earlier quoted context omitted.

And I really don't mind. I want every line of code I've ever written to be used as much as possible. I find "intellectual property" to be dubious to the core. I'm not confident enough in my feelings to be a zealot, but if I had to pick sides then I know which side I would pick.

If an AI "listened" to music and created new samples for musicians to use for a fee, do you not think the original musicians should be compensated? The value transfer is basically theft. It isn't about the usefulness of the service, or even that something similar is a good thing ... it is about the execution and what it says about fairness for those that worked to create the data it depends on to produce value.

If I listen to music and create samples to be sold, is it theft?

Re: Copilot sells code other people wrote

#753

Post by mullikine » Fri, Jun 24, 2022, 1:48 AM UTC

Traditional 'real' (as opposed to 'imaginary') programming is like writing in assembly code; It's outmoded because of generative models, in a way similar to 'C' outmoding assembly code. The most important thing, I think, is that free (libre) software developers are able to work with the language models directly, so that libre software is allowed to continue progressing into what I call Imaginary Programming. That's because with a generative internet all you really need is blockchain + prompting.

https://huggingface.co/spaces/mullikine/ilambda

Language models are able to 'steal' the linguistic meaning-making 'essence' of the software, by modelling:

- How the software is used (mimicing its function) - external meaning

- How functions are 'inspired' - internal meaning (reflection)

https://github.com/semiosis/imaginary-programming-thesis

The models themselves should be clear about where the data came from. However, this is only possible in a fair world which we do not live in. Compromise must be made to protect national interests.

Generative models are license blind and there's very little that could be done to prevent progress. Like what the invention of the camera has done for art.

Large language models including Codex are a transformative technology.

Bi-directional fair-use is probably the best result we can hope for.

So long as Microsoft and OpenAI are not selling back usage of the model to the open-source community, I think it's OK, though it's the bare minimum obligation.

Re: Copilot sells code other people wrote

#754

Post by williamcotton » Fri, Jun 24, 2022, 1:51 AM UTC

leereeves wrote: post #744

williamcotton wrote: post #738

Earlier quoted context omitted.

The fast inverse square makes for a poor example when it comes to notions of intellectual property and copyright because there is prior art. The Wikipedia page has a history. And imagine if Microsoft had been able to copyright the fast inverse square function before Carmack sat down to write Quake!

Prior art matters for patents, not copyrights. Carmack's code is still protected by copyright even if he didn't invent the algorithm. And copyright doesn't prevent someone else from implementing the same algorithm, only from copying the code. If Microsoft had been able to copyright the fast inverse square root function, Carmack could still have written his own version and even copyrighted that version himself.

Prior art has been used for copyright in plenty of court cases, and increasingly so!

Here’s an example: https://scholar.google.com/scholar_case?case=728470765881077...

It seems particularly apt to consider prior art for use in software IP if only for the similarities with the patentable invention of mechanical parts.

Re: Copilot sells code other people wrote

#755

Post by ShamelessC » Fri, Jun 24, 2022, 2:03 AM UTC

bayindirh wrote: post #654

Guid_NewGuid wrote: post #621

Earlier quoted context omitted.

But why is it immoral? All that code is still out there, if I had the time and the resources I could build a language model. Unlike commons in the real world (e.g. land, fresh water, etc) a code commons is purely additive. With the release of Copilot (which I don't intend to pay for or use) nothing has been destroyed, instead we'll get more code for less work where companies do pay for their developers to use it, som…

Because copilot is violating the terms I put for my code. My code is GPL. It cannot be put into projects with incompatible licenses. That’s my code, and I share it with strings attached. You can’t just copy my code and sell to other parties no strings attached. If that’s fine and dandy, Microsoft should also train Copilot on their source code repositories, so we can use that knowledge, too.

I guess I've just never had to work with GPL code before, but the complaints essentially only seem to be coming from coders who like this style of open source where you still get to make it kind of a pain in the ass to actually use your software.

I guess you have the right to do this, but it doesn't mesh at all with why I personally contribute (without any expectation of attribution), which is that (much like stack overflow), programmers mostly agreed awhile ago that it's just easier if we all share.

So much of what's wrong with the modern economy comes down to seeking rent on an idea that should just be public knowledge.

Sorry if my viewpoint towards your work is apathetic, but the whole field is already infested with academics who only understand citation as a useful metric. Further, the point remains that anyone with enough money could do this - not just Microsoft (Salesforce has released several models for python competitive with Copilot). Times are changing - maybe don't share code anymore? I imagine in ten-twenty years this whole conversation will seem pretty petty though when your entire program is trivially recreated from its GitHub description without ever needing to have seen it in the first place.

Re: Copilot sells code other people wrote

#756

Post by zarzavat » Fri, Jun 24, 2022, 3:36 AM UTC

grayfaced wrote: post #435

jimnotgym wrote: post #295

Earlier quoted context omitted.

I sometimes read people's open source code on github and use the ideas from that to develop my own ideas. In fact sometimes I copy and paste short passages and then rework them. I also employ a team of people who may do the same. Is that fair use, yes of course it is. Is co-pilot automating that fair use, I would say so.

Or alternately, "I sometimes listen to other people's songs and use those ideas to develop my own. In fact sometimes I copy and paste short melodies and then rework them." Courts have held that it doesn't apply to music, why do you think different rules apply to code?

Courts are definitely aware of the need to protect the creative process and that no song is truly “original” in all aspects. e.g. the Katy Perry case[0]

Songs are different from code, in that the “hook” that makes the money may be only a few seconds long. There are many creative choices that a songwriter/producer can fit into just a few seconds: the harmony, melody, rhythm, lyrics, timbre, effects, ...

Whereas for code, the space of creativity is limited by functional considerations. A creative choice is protected by copyright but not all choices that programmers make are creative. Often the choices are limited by the API/interface or by efficiency considerations and it turns out that there’s only one good way to do something.

A function may be very intricate, yes, while still containing almost no creative value (e.g. a Vulkan setup function). Music doesn’t have an equivalent to this - the placement of every note is a creative act.

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

Re: Copilot sells code other people wrote

#757

Post by kraf » Fri, Jun 24, 2022, 4:04 AM UTC

concordDance wrote: post #529

VoodooJuJu wrote: post #364

It is now proven that copilot returns code from codebases with non-permissive licenses [1]. I'm curious - what are the legal implications of this going forward? I've so many questions. 1. Will Microsoft ever face lawsuits for these license violations? 2. If so, who/how? Class-action? 3. Will copilot be forced to open-source in the future? Under which license? Some open source licenses are incompatible with others, bu…

There's also one more question: 5. Even if it is illegal, is it actually bad? No one can possibly sell code snippets, the transaction costs are many orders of magnitude greater than any reasonable price. In my opinion, at least in this case the benefits massively outweigh the costs and the law should not apply here.

Of course it's bad. Noone who put up their work as open source wants some huge company taking it and selling it to get even more competitive advantage and influence in the world. And that's without mentioning the people who put that into their license pretty much explicitly. Taking GPL code and getting away with it is a failure of our justice system, and that can't be made right with throwing pennies at developers.

Re: Copilot sells code other people wrote

#758

Post by zarzavat » Fri, Jun 24, 2022, 4:18 AM UTC

bayindirh wrote: post #467

zarzavat wrote: post #218

Earlier quoted context omitted.

The entire point of a fair use right is that you don’t need the copyright owner’s permission to be able to exercise it. Fair use allows you to do things that the copyright owner doesn’t like. Is fair use on a massive scale still fair use? Courts generally think so, otherwise Google would have been out of business a long time ago.

An automated system will devour all my code, which is under a case-tested copyleft license, and regenerate its parts in any place, without respecting the license terms, and call it "fair use". I have two questions: 1. Why have licenses, then? 2. What if I just use leaked sources of closed source software and call it fair use?

1. Why have licenses, then?

The default under copyright law is that any substantial copy is infringement.

A license is a legal document that grants someone permission to use a work that they otherwise would not have had.

However the law also gives its own permissions to use a work - it defines what is unlawful infringement and what is lawful fair use.

The code snippets that copilot generates look more like fair use than infringement. They are small, adapted to the destination context, and usually not direct copies of one source but more of an average of many different sources. And usually the programmer does not keep the suggestion that copilot suggests unmodified - the programmer does their own editing of the snippet afterwards to further tune it to the surrounding context.

2. What if I just use leaked sources of closed source software and call it fair use?

As pointed out upthread, if it the source code is leaked then there may be trade secret protections. The GPL specifically allows the code to be posted online, so by design it is not secret.

Re: Copilot sells code other people wrote

#759

Post by nmfisher » Fri, Jun 24, 2022, 4:21 AM UTC

Guid_NewGuid wrote: post #569

I find this whole topic very annoying, this is like the 3rd variation to reach the front page today. But it has made me realize why I instinctively dislike Free Software as a movement. Copyright and licensing are bad, actually. Stop getting worked up about the idea of using courts to punish theft. Stop getting into a frenzy of arousal about the police kicking down doors to drag Billy Gates to jail because 80 characte…

I totally agree, this reaction seems very hypocritical. If some rinky dink startup did exactly the same thing - as they are entitled to do under the licences of huge swathes of code on GitHub - hardly anyone would bat an eyelid. But just because it’s a Microsoft-owned company, it’s somehow verboten?

That seems totally inconsistent with decades of people clamouring for more openness/liberty when it comes to IP rights.

Re: Copilot sells code other people wrote

#760

Post by ece » Fri, Jun 24, 2022, 4:49 AM UTC

sirsinsalot wrote: post #643

jazzyjackson wrote: post #616

Earlier quoted context omitted.

If it was just published as a public good it would probably be as illegal as sci-hub I consider the $10/m as a donation to the microsoft legal defense fund to allow free access to accumulated knowledge.

To allow access to a service that grants you the accumulated knowledge's output in small bits. I'm all for a world where these tools help developers, but i'm not here for a system that isn't open. I want to own my tools. Copilot is a bit like musicians paying a monthly fee for access to a loop library. Except all the loops are rip-offs of other peoples hard work and there's no effort to compensate them. If I made an…

I'm all here for openness and tools you own, so there could be a FOSS implementation. Microsoft could just open it up and still charge the $10/mo for hosting the model, and I hope that happens.

Making the tool better without verbatim copying and making it more effective should be the priority, IMO. Trying to control it too much would be missing the point of the tool.