Series 8 of 16
Training Data
What went in, who put it there, and who was asked.
Choose how to learn
Episode 1 · comingWhat is it trained on?Where does the text actually come from?
Episode 2 · comingPretraining and fine-tuningWhat is the difference between learning language and learning manners?
Episode 3 · comingReinforcement learning from human feedbackWho taught it what a good answer looks like?
Episode 4 · comingData labellingWho does this work, and under what conditions?
Episode 5 · comingCopyright and consentWas anyone asked?
Episode 6 · comingSynthetic dataWhat happens when models train on model output?
Next seriesModel Limitations