Core insights:
Bots are bad at saying "I don't know". They rather offer incorrect info. Or speculate.
Bots fabricate links and cite syndicated and copied versions of articles.
Multiple bots seem to bypass bot exclusion protocols.
Content licensing deals provide no guarantee of accurate citation.
Premium bots are worse than their free counterparts.
Bottom line: You can't really trust AI search engines (yet).
(via Columbia Journalism review)
https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php
@dw_innovation
This one sounds convincing:
"Premium chatbots provided more confidently incorrect answers than their free counterparts."
@dw_innovation fuck ai we need clean energy
@dw_innovation
Own tests: Even locally installed bot instances trained exclusively on specific documentation will often provide completely fabricated responses, confidently asserting that they are from the documentation. They can only be trusted if you already know the correct answer.
Bots are not bad at saying "I don't know". They don't know anything.
@dw_innovation
bullshit machines. Kruger Dunning machines
AI search engines are not just imperfect—they actively undermine journalism by diverting traffic, misattributing sources, and fabricating citations. This @cjr article is essential reading for anyone who cares about information integrity.