The training of the models was actually sooner, and they underwent multiple 
rounds of additional training, because LLMs as of themselves can't actually 
chat normally, they just autocomplete existing text. I think you're referring 
to the fact that the dataset has information up to September 2021.

Reply via email to