No, ChatGPT doesn't know it's own sources. It's just a trained model. Once the model has been created it's fixed - it can be recreated unlimited times, but it will never tell you the sources for it's output.
Maybe if the network nodes have a source attached to them...
But thinking out loud...
That's not how the number-tokens work. It's at a word level... so "a list of us presidents" is broken down into individual number-tokens for each word, and you can't provide a source for each word.
---
I'm not sure how you combine Google and ChatGPT.
Chat is creative/combinatorial and Google is "just the facts".
ChatGPT and Google are going to have problems going forward. How do both of them determine if the information they find on the internet is from a meat-brain and not a metal-brain.
Question -> "creative" output -> Google -> Summary of links -> Comparison -> confidence level (or re-write) + links that were used for checking
Not so different than how we work in a high-level. I believe that openAI has published a paper called webGPT that has a workflow like this (although not sure its exactly the same)
Maybe if the network nodes have a source attached to them...
But thinking out loud...
That's not how the number-tokens work. It's at a word level... so "a list of us presidents" is broken down into individual number-tokens for each word, and you can't provide a source for each word.
---
I'm not sure how you combine Google and ChatGPT.
Chat is creative/combinatorial and Google is "just the facts".
ChatGPT and Google are going to have problems going forward. How do both of them determine if the information they find on the internet is from a meat-brain and not a metal-brain.
Happy to be proven wrong.