HN (and other forums) are great sources of myths. It goes like this:<p>- Person A asks a question and proposes a suggestion<p>- Person B responds to the suggestion and adds more complaints<p>- Person C either agrees or disagrees with A and B's claims, but adds an argument which adds more detail/fills in more gaps and adds more suggestions<p>At no point does anyone say "hold on, do any of you have any empirical evidence of <i>anything</i> you've said?" Everybody just kind of accepts every premise, and then argues it as if it's a valid premise. That acceptance of the false premise is the bedrock on which invented ideas become commonly-held beliefs.<p>I think this is part of what makes AIs hallucinate. These conversations get ingested by the AIs and make connections between these claims, which it will see again and again as the myths are repeated from forum to forum. Later it will make correlations between the conversations and there isn't anything solid enough for it to determine which is real or not. I've seen it recently when analyzing adhesives, where it will not only quote the wrong information about an adhesive (despite it already having the TDS in its database), but also it'll make up fake adhesives that are clearly a mash-up of two real adhesives.<p>(See, I've just done it... completely talking out of my ass about a subject I don't know anything about, but sounds plausible)