You know how in chatgpt with any response from any of the models, you can give feedback to each prompt? Like giving it a thumbs up, or a thumbs down and giving it a reason why. It would be really cool if you could also give each step/thought that the o1 models tells you it did some feedback too!.
Here’s 2 super niche examples out of many many:
Like if you asked o1 to look through an set of your own gpt memories(by copying and pasting it to be safe) for redundancies, and it all of a sudden starts “engaging with” the ideas in the memory, like that’s useless and all those steps get a thumbs down, or if you gave it a complicated puzzle and its going through ways of solving it, you can give it thumbs up for all the reasonable(even if leads to dead-end) or correct steps it shows.