Sujet : On creating an identifiable source for AI
De : mailbox (at) *nospam* cpacker.org (Charles Packer)
Groupes : rec.arts.sf.writtenDate : 26. Jun 2025, 08:52:50
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <pan$bd9a0$6d311321$f5e42695$b01a8abf@cpacker.org>
User-Agent : Pan/0.158 (Avdiivka; )
Serendipitously I discovered that I'm the world's expert
on the history of the curious social custom of the progressive dinner.
If you ask ChatGPT for "progressive dinner origin", it will tell
you that they appeared in the 1950s. But if you ask
Google search the same question, its AI Overview will report that
they began in the 1880s. The vast discrepency is due to the fact that
ChatGPT hasn't updated its knowledge base since I contributed
a paragraph of history for the Wikipedia page for "progressive
dinner."
https://en.wikipedia.org/wiki/Progressive_dinnerGoogle's LLM obviously adapts the language of my paragraph to
form its summary output.
Although the material I contributed is factual -- at least
insofar as newspaper archives can be trusted -- my experience
suggests that one could make an evil hobby of searching out obscure
subjects in Wikipedia that are rarely policed and introducing
all kinds of mischief that would have downstream effects on AI.
ObSF: "Secret Stories of the Doors" by Sofia Rhei.