It creates the effect of real time computation and conversation. It’s kind of like the page turning animation that Apple used to have on iBooks, it was magical for no reason other than to delight.
Because that's how it actually works. It generates the text token by token and you're seeing it live. GPT-3 Lab does the same, you can see when it gets slower/faster.
3 comments
[ 1.7 ms ] story [ 9.7 ms ] thread