3 ms·
See? It is good to catch grifters abusing the GPL license and holding them accountable. We can all do the same and file complaints like this one [0] about GitH
by rvz 4y ago
See? It is good to catch grifters abusing the GPL license and holding them accountable.
We can all do the same and file complaints like this one [0] about GitHub, OpenAI and Microsoft with their Copilot product and model being trained on AGPL-3.0 code and a mixture of incompatible licenses and also outputs sensitive info of developers.
Give that Truth Social was caught using Mastodon's source code and violated the AGPL license by giving no attribution and no copy of the source code which they later complied, now we are starting to see this with AI projects violating licenses of open source projects; the same should also apply to GitHub, OpenAI and Microsoft with their Copilot product.
[0] https://githubcopilotlitigation.com/ https://githubcopilotlitigation.com/
- kmeisthax 4y agoDid Truth Social ever actually comply with AGPLv3? I thought they just pointed vaguely at the original Mastodon source repository.
- Dylan16807 4y ago> a mixture of incompatible licenses As far as copilot is concerned, those repos might as well be closed source. They are not accepting the license, and don't care about it. Their goal/claim is that the language model will not retain copyrightable sequences.
- jimlongton 4y ago> Their goal/claim is that the language model will not retain copyrightable sequences. In case people are not aware, we know for a fact this is demonstrably false. It may not even be possible to achieve this with existing models. People have replicated entire sections of GPL code (with comments) using CoPilot. This is true of other models, like image generation models which can return clearly recognisable images which were present in their training sets.
- Dylan16807 4y agoThe interesting question is how hard that is to fix in future versions.
- jeroenhd 4y agoAs far as I know, an AI spitting out training data has clearly been overfitted on its input. However, I think this may be intentional, as an AI that doesn't spit out the occasional bit of training data may not be delivering output of a high enough quality.