4 ms·
If someone just decides to use the torrent and ignore those restrictions it might finally establish precident for if you can copyright model weights.
by sebzim4500 4y ago
If someone just decides to use the torrent and ignore those restrictions it might finally establish precident for if you can copyright model weights.
- dougmwne 4y agoBut even if you could copyright them, once you do some fine-tuning, they are not the same model weights!
- charcircuit 4y agothat would be a derivative work
- dougmwne 4y agoThat’s a legal unknown. And it’s also a technical unknown how you would even determine it was descended from the same model in a way that would hold up in court.
- sebzim4500 4y agoI would imagine that the weights of a finetuned model are highly correlated with the original weights. Having said that, simply permuting the neurons would make it way harder to match them up, I can't think of a straightforward way to reverse it.
- counttheforks 4y agoYou mean like how the model itself is a derivative work of tons of copyrighted content? If the original model can sidestep the issue of being trained on copyrighted content, then it should be fair game to train a new model off of a copyrighted model.
- charcircuit 4y agoYes, the model is a derivative work of its training data, but the difference is that a model is transformative. Llama is not a replacement for reading some book. On the other hand a fine tuned model built upon llama is much less transformative.