I do mean "grounding" as seeking the primary source of information by hand as my sibling commenter said.
If an LLM says something about history, have it give you source documents so that you can verify. If it says Earth's gravitational constant is 98 [sic] m/s^2, have it generate a prediction of a timed 1m paperclip drop test, and go do it yourself.
It's perhaps excessive in the case of gravity, but for things that matter like medical stuff I tell LLMs to expose their step by step causal reasoning / inference, and then check whether I buy it myself. SotA models' biochemistry + neurochemistry is often pretty solid. Often their answers are based on a slightly hallucinated understanding if it's a spatial anatomy question. It's fun when you get some research paper citations and then upon reading them decide you don't agree with the underlying research (eg. fMRI studies often make broad / interesting claims based on small sample sizes or sample populations that are clearly biased).
To be clear, I don't think that the LLM's self explained reasoning in a single turn is how it actually arrived at a given conclusion, but I just want to expose and bind the final conclusion to a verifiable epistemic chain.
Yes, the trick is having a "ground-truth" KB it can link to, and trying to avoid citogensis from happening. I'm not sure if it's a technique that's every going to scale because of that.
I don’t think this is correct, as we’re talking about all online discourse being influenced by AI, and you don’t know what to trust anymore, rather than generating AI output yourself (which is what you are referring to).
In this context, it would mean building a knowledge base the model can link to that is considered authoritative. So maybe the LLM is has some slop in it's training, and that causes it to try and output some junk, but if it can't match it to it's verified ground-truth KB it doesn't end up outputting it.
The problem is figuring out what is the authoritative sources.
If an LLM says something about history, have it give you source documents so that you can verify. If it says Earth's gravitational constant is 98 [sic] m/s^2, have it generate a prediction of a timed 1m paperclip drop test, and go do it yourself.
It's perhaps excessive in the case of gravity, but for things that matter like medical stuff I tell LLMs to expose their step by step causal reasoning / inference, and then check whether I buy it myself. SotA models' biochemistry + neurochemistry is often pretty solid. Often their answers are based on a slightly hallucinated understanding if it's a spatial anatomy question. It's fun when you get some research paper citations and then upon reading them decide you don't agree with the underlying research (eg. fMRI studies often make broad / interesting claims based on small sample sizes or sample populations that are clearly biased).
To be clear, I don't think that the LLM's self explained reasoning in a single turn is how it actually arrived at a given conclusion, but I just want to expose and bind the final conclusion to a verifiable epistemic chain.