According to this tweet,
when gpt4 first finished training it didn’t actually work very well and the whole team thought it’s over, scaling is dead…until greg went into a cave for weeks and somehow magically made it work
So gpt-4 was kind of broken at first. Then greg spent a few weeks trying to fix it and then it somehow worked.
So why did it not work at first and how did they fix it?
I think this is an important question to the OSS community,
He’s just alluding to the fact that most enterprise customers are too stupid to use base models as they expect to be interacting with a human-like dialogue-driven agent or chatbot rather than a supercharged text completion engine. It’s a shame given that, used properly, the GPT-4 base model is far superior to the lobotomized version made generally available through the API.