Also, a quote from Utah agreeing with your point about common surname v. U.S. "There are thousands—I was going to say 'a thousand,' that’s an exaggeration—there are a bunch of cases in Utah called 'State versus Carter. 'Quite by accident, we cited the wrong State v. Carter. That’s our error. It doesn’t really support the proposition.'" https://midwestfrontierai.substack.com/i/174305891/utah-show-cause-ai-ok-misuse-not
In LLMs, the only meaning a term has is the (condensed) set of words with which it co-occurs. Words operate as if they have similar meaning when they occur in similar contexts. LLMs have no way to represent anything other than the context, so, as you point out, fabricated words (in your analysis, citations) are going to be more closely associated with the context of the rest of the document. Thanks for this great start on understanding hallucinations.
Human systems are overwhelmed by machine scale errors.
Excellent article. It reminded me of when I found a hallucinated case that change the first letters to get Casserole Tech v. Caserage Labs apparently from a real Laserage Tech v. Laserage Labs. https://midwestfrontierai.substack.com/p/plumberslocal-no-75-v-morris-plumbing
Also, a quote from Utah agreeing with your point about common surname v. U.S. "There are thousands—I was going to say 'a thousand,' that’s an exaggeration—there are a bunch of cases in Utah called 'State versus Carter. 'Quite by accident, we cited the wrong State v. Carter. That’s our error. It doesn’t really support the proposition.'" https://midwestfrontierai.substack.com/i/174305891/utah-show-cause-ai-ok-misuse-not
In LLMs, the only meaning a term has is the (condensed) set of words with which it co-occurs. Words operate as if they have similar meaning when they occur in similar contexts. LLMs have no way to represent anything other than the context, so, as you point out, fabricated words (in your analysis, citations) are going to be more closely associated with the context of the rest of the document. Thanks for this great start on understanding hallucinations.