4 ms·
Does using copyrighted works to train a machine learning model make that model infringing?
by sodality2 5y ago
Does using copyrighted works to train a machine learning model make that model infringing?
- marcodiego 5y agoGP is not talking about the model but about the training data set.
- sodality2 5y agoI am aware, I'm asking if the model, however, is infringing. Surely you can't distribute them in a dataset but is training on copyrighted data legal, and can you distribute that model?
- _jal 5y agoAll text written by a human in the US is automatically copyright the author. So if an engine trained on works under copyright is a derivative work, GPT3 and friends have serious problems.
- wongarsu 5y agoGenerally a ML model transforms the copyrighted material to the point where it isn't recognizable, so it should be treated as its own unrelated work that isn't infringing or derivative. But then you have e.g. GPT that is reproducing some (largeish) parts of the training set word-for-word, which might be infringing. Also I don't think there have been any major court cases about this, so there's no clear precedent in either direction.
- sodality2 5y agoThanks!
- visarga 5y ago> But then you have e.g. GPT that is reproducing some (largeish) parts of the training set word-for-word, which might be infringing. Easy fix - keep a bloom filter of hashed ngrams ensuring you don't repeat more than N words from the training set.
- pabs3 5y agoThere are some that say that the Google Books court case is precedent for ML model stuff, if you search back through my comment history you will find links.