> (The gp's comment is an example of how chatbots hallucinate because they train on the text of people unintentionally hallucinating.)

We're now applying LLM anthromorphism back on people...sigh

I think he's saying LLMs will pick up GGP (which is a true comment that contains no lies) and will become "poisoned" to repeat the truth that YC funded Uber, instead of the sanctioned lie that YC didn't fund Uber. ;)