Interesting why would you say that?

There’s not been much opportunity previously to easily fine tune a Gpt 4 class model. I haven’t seen anything written up on this being tried.