Follow

Not sure why should be license violation -- have you read the Terms of Service?

docs.github.com/en/github/site

It's written there: "... or otherwise analyze it on our servers ...".

Well, I'm not a lawyer, but it's kind of clear to me.

@qeef Github analysing your code is fine, but it's a random other person adding the results of that analysis (ie the generated code) in their own non-FLOSS (or differently licensed FLOSS) that's the issue

@ebel I'm afraid that you can understand "... or otherwise analyze it on our servers ..." the same as "... training our AI ...".

@qeef yeah, one could read that as meaning that.

But it doesn't say they can change the copyright licence.

@ebel Agree. But we are talking about Terms of Service, don't we?

docs.github.com/en/github/site

"D-4. License Grant to Us ... This license includes ... or otherwise analyze it on our servers ..."

@qeef yeah, Github has a right to train their AI on your code.

But the output of that AI? Isn't that still copyrighted work? A random other github user cannot include that in their source code.

@ebel What is problem IMO is that if you -- as human being -- learn something, even from the proprietary code, you are free to use the knowledge in your FOSS project. The same works in the other direction.

I do remember there was similar discussion in community regarding the but I haven't found the link just now.

@qeef There is a question about if you use OSM data to train a neural network, is the data that comes out of that ODbL? (IMO yes)

@ebel I tend to say yes, but I feel it's not clear enough.

Regarding the it's a bit clearer -- you agree by Terms of Service with the code analysis (translated to AI training.) This step IMO effectively kills the license of your project. The AI do not produce copy, but something new, which was learned with the license from the Terms of Service.

@qeef You're right that the big question depends on how much the original licence applies.

There are examples of Github Copilot reproducing large chunks of GPL'ed code verbatim, which is almost certainly still under the original licence.

@qeef As someone pointed out, Is MS willing release an AI that's based solely on the Windows kernel source code? (which might help windows device drivers)?

Has MS included the source code for _github.com_ in the neural net?

@ebel :D I don't think so. They will tunnel out as much open source projects as they can.

@ebel @qeef I guess that pushing things through a neural net sausage machine creates a legal grey area. A legal mixnet. If the result is close to being an exact copy or at least visibly a derivative then copyright would apply.

Sign in to participate in the conversation
En OSM Town | Mapstodon for OpenStreetMap

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!