If you assigned a task to a junior dev, and he/she used some code from open source projects and Stack Overflow to develop a custom program for the task, would you say that this person is selling you other people's code? Is it common or expected for this type of use to be acknowledged?
Copilot sells code other people wrote
21–30 of 838 posts
Re: Copilot sells code other people wrote
#22Same deal for Dall-e if they ever productize it.
Re: Copilot sells code other people wrote
#23Is it smart enough to: - respect attribution - respect copyleft - respect proprietary licences - give the user appropriate hints about the above Or does it just copy code without doing any of this?
No, it doesn't do any of that. However, it does not "copy code" except in marginal use cases, the far more common scenario is that it will suggest you very basic code that is akin to a Stack Overflow reply.
I read a lot of open source code and might subconsciously absorb techniques and patterns that are common. When I write code I might be influenced by what I read, not line per line, but rather generally.
Is it like that?
Re: Copilot sells code other people wrote
#24Earlier quoted context omitted.
It is hard to see how verifying licenses is a solvable problem, when licensing for code dependencies can be transitive. For ex - if I copy code from a GPL codebase like Linux and create a Github repository with an MIT license.
You should be able to choose flavors of the model trained only on public-domain code which does not require attribution, for example. But that would mean Microsoft acknowledging license violations.
Sorry, to be clear, I meant even if a Github user asserts their code is public-domain/no-attribution/unlicensed, they could have lifted it off a codebase that doesn't allow it. It would be tricky for Github to establish the code was indeed original and hence their agreement with the user allows them to train their models on it.
Re: Copilot sells code other people wrote
#25That's like saying a plumber just sells parts that other people made
Re: Copilot sells code other people wrote
#26I get the feeling this entire debate would have been non-existent had this been a Jetbrains product instead. The whole thing is just bizarre when the vast majority of developers constantly look at OSS code daily and lift ideas/patterns/snippets from there regularly without once looking at whatever license is attached.
> the vast majority of developers constantly look at OSS code daily and lift ideas/patterns/snippets from there regularly
Perhaps in your circles, but that's certainly not something I've encountered over a 25 year carreer.
Re: Copilot sells code other people wrote
#27I get the feeling this entire debate would have been non-existent had this been a Jetbrains product instead. The whole thing is just bizarre when the vast majority of developers constantly look at OSS code daily and lift ideas/patterns/snippets from there regularly without once looking at whatever license is attached.
> I get the feeling this entire debate would have been non-existent had this been a Jetbrains product instead.
why so?
> The whole thing is just bizarre when the vast majority of developers constantly look at OSS code daily and lift ideas/patterns/snippets from there regularly without once looking at whatever license is attached.
well, yes, copying an idea or pattern is generally.. accepted, to be kosher. copy-pasting too, in small amounts (a function, a type). that said, i would (and have) attribute even a notional similarity when writing something open source.
i don’t think co-pilot even allows the user to find where the code came from.
Re: Copilot sells code other people wrote
#28Earlier quoted context omitted.
You should be able to choose flavors of the model trained only on public-domain code which does not require attribution, for example. But that would mean Microsoft acknowledging license violations.
Sorry, to be clear, I meant even if a Github user asserts their code is public-domain/no-attribution/unlicensed, they could have lifted it off a codebase that doesn't allow it. It would be tricky for Github to establish the code was indeed original and hence their agreement with the user allows them to train their models on it.
> they could have lifted it off a codebase that doesn't allow it
Ah. But then someone else is guilty of redistributing code without permission.
But you're suggesting, GitHub should implement something like ContentID but for code. Which should be cheaper (since code is cheap to analyze, while videos are much more bandwidth-intense). And this would kill two birds with one stone.
Re: Copilot sells code other people wrote
#29It’s the same problem with those ML models, the other day someone generated a children’s book using GPT3, turned out that there is a real children's book with the same name and a very similar content: The Very Lonely Firefly by Eric Carle.
Re: Copilot sells code other people wrote
#30Earlier quoted context omitted.
No, it doesn't do any of that. However, it does not "copy code" except in marginal use cases, the far more common scenario is that it will suggest you very basic code that is akin to a Stack Overflow reply.
I read a lot of open source code and might subconsciously absorb techniques and patterns that are common. When I write code I might be influenced by what I read, not line per line, but rather generally. Is it like that?
Kinda, but I think you are imagining something bigger than it is. At least in my experience, it works well for simple stuff like "iterate over x and extract y" or similar queries that I imagine are well represented in its training data. When you get to very specific functions, its answer will be less reliable and more likely to be a wonky rehash of the few examples it has for that case.