Analyzing the legal implications of GitHub Copilot • Hacker News Board

Re: Analyzing the legal implications of GitHub Copilot

#171

Post by jozvolskyef » Sat, Jul 17, 2021, 1:39 PM UTC

_ph_ wrote: post #168

jozvolskyef wrote: post #128

Earlier quoted context omitted.

If a person who's never seen a goat looks at a million copyrighted images of goats and draws a goat, are they committing copyright infringement? What if an algorithm does the same? The result is 'a recomposition of learned fragments' in either case.

The point is, with a human you can assume that the person did learn how a goat looks and that a drawing isn't a mechanical translation of one image. Only in cases, where a single image is faithfully reproduced in the drawing, copyright questions might arise. But the base for this is: humans can learn and have an obvious creative process. With the AI it is not clear, that there is an abstract lerning beyond building c…

From what I understand about these models, having studied and applied machine learning but not deep learning, they all build abstractions (embeddings) akin to eigenfaces[1][2] where many samples of a feature are turned into an abstract feature. These abstract features are then manipulated by the model. Sure, just like with Fourier transform-based compression, anything can be reconstructed from primitive features, it is called overfitting and the challenge of training these models is building a learning process that achieves generalization rather than overfitting. It is not true that 'it is not clear that there is abstract learning', because without it, generalization is impossible, and without generalization the model wouldn't pass even the most trivial benchmarks.

[1]: https://media.geeksforgeeks.org/wp-content/uploads/202003171...

[2]: https://www.geeksforgeeks.org/ml-face-recognition-using-eige...

Re: Analyzing the legal implications of GitHub Copilot

#172

Post by UncleMeat » Sat, Jul 17, 2021, 3:57 PM UTC

pedrocr wrote: post #170

UncleMeat wrote: post #159

Earlier quoted context omitted.

> I'll take the Windows and Office source code, run it through a decompiler and then train a neural network on that output. At the risk of adding another uneducated voice onto this legal discussion - the law is not written like code. Judges can absolutely say "hey asshole, we know exactly what you are doing" even though engineers like to imagine that all situations that use the same technical processes will be handle…

What asshole thing would I exactly be doing? The scenario I described isn't a weird hypothetical to exercise the law as code. It would be a genuinely useful use of the Microsoft copyrights if that was allowed, just like Copilot is a genuinely useful use of the open-source copyrights. You just can't have one without the other.

You are starting with the goal of recreating windows source and will produce a ton of it, rather than using the tool for another purpose and accidentally obtaining a small amount copyrighted code. The starting intentions are obviously different.

Re: Analyzing the legal implications of GitHub Copilot

#173

Post by pedrocr » Sat, Jul 17, 2021, 6:12 PM UTC

UncleMeat wrote: post #172

pedrocr wrote: post #170

Earlier quoted context omitted.

What asshole thing would I exactly be doing? The scenario I described isn't a weird hypothetical to exercise the law as code. It would be a genuinely useful use of the Microsoft copyrights if that was allowed, just like Copilot is a genuinely useful use of the open-source copyrights. You just can't have one without the other.

You are starting with the goal of recreating windows source and will produce a ton of it, rather than using the tool for another purpose and accidentally obtaining a small amount copyrighted code. The starting intentions are obviously different.

My version of the tool doesn't even create original code as it doesn't have access to it. If the difference between the two uses is how much verbatim, obviously infringing code, gets created, then Copilot has much bigger issues.

The point of my hypothetical wasn't to whitewash stealing Microsoft code in some kind of legal quirk. It was to point out that if learning the structure of code with this kind of ML model is fair use then doing so with Microsoft copyrights is also useful for other purposes. And if Microsoft themselves think that is ok it would actually be a strong argument in this discussion. I suspect they don't and was pointing that out.