Comment by Hedepig

Slightly tangential, is there some kind of crowdsourced effort to build training data for fine tuning? Alpaca used the training data built from gpt-3.5, so there are terms of use restrictions