How OpenAI made o1 "think" – Here is what we think and already know about o1 reinforcement learning

Transcript

OpenAI o1 makes our last video  thumbnail look a little dated, because it can finally count how many  ‘r’s are in “Strawberry”! This is actually why OpenAI o1 was codenamed  “Strawberry” during development. Oh, wait... Turns out, it doesn't  always count the ‘r’s correctly. and for some reason, it struggles  with counting the ‘t’s?! Aahh, come on, ... Read More