John Schulman
TalkRL: The Reinforcement Learning Podcast
00:00
Instruct GPT - A Language Model That's Fine-Tuned to Follow Instructions
Instruct GPT is a language model that's fine-tuned to follow instructions. You can play with it on the OpenAI website by writing some text and then press the button to generate a completion. The idea here was language models are pretty useful and you can sometimes get them to do what you want by prompting them just right.
Play episode from 18:03
Transcript
