It’s a statistical model, the bar is lower than that. You just have to get it right more than wrong. The ChatGPT web app doesn’t expose this behavior but real a

It’s a statistical model, the bar is lower than that. You just have to get it right more than wrong. The ChatGPT web app doesn’t expose this behavior but real apps can and do (I use this approach) use the ability to generate multiple completions, use that to generate confidence scores, and update the response appropriately. Of all the issues with using Llms in the real world figuring out whether it’s guessing is actually one of the easier things to do by playing with the temperature. When the model is very confidently incorrect is when you have to worry.