3 ms·
Meta has opened sourced almost all of their models trained on this data.
by derekdahmer 1mo ago
Meta has opened sourced almost all of their models trained on this data.
- jacquesm 1mo agoBut not the training data.
- derekdahmer 1mo agoNot sure what your point is - the articles they trained on are copyrighted and not theirs to open source. The only they can legally open source is the weights, which they have done repeatedly. My point is both Aaron was trying to make the journals public and Meta made models that encapsulate data from the journals public.
- thomasjeff1 1mo agoWith added filters (censorship)