Unlike Wikipedia, the references supplied by LLM output often point to material that doesn’t actually support the LLM output. This can be because the LLM misapplies it to the wrong context or because it’s incorrectly correlated and has nothing to do with what the LLM output making it unhelpful for determining the accuracy of the output.
Sure it is, if I look something up in an encyclopedia or training material written by a subject matter expert I have high trust in its content. If I ask an AI I have low trust in its content. The two are not equivalent.
If you are looking up something you don’t know about, how do you know the info provided us correct?
it supplies references. check them.
you could also say the same thing about wikipedia.
Unlike Wikipedia, the references supplied by LLM output often point to material that doesn’t actually support the LLM output. This can be because the LLM misapplies it to the wrong context or because it’s incorrectly correlated and has nothing to do with what the LLM output making it unhelpful for determining the accuracy of the output.
This is not an AI specific problem
Sure it is, if I look something up in an encyclopedia or training material written by a subject matter expert I have high trust in its content. If I ask an AI I have low trust in its content. The two are not equivalent.