Post #2474574

2026-05-12 03:02 UTC

The repeated refrain that #AI is inaccurate &amp; hallucinates &amp; is therefore useless totally ignores how 1) it&#39;s getting more and more accurate and 2) even if it weren&#39;t, most questions we have while navigating life really aren&#39;t seeking the purest most scholarly-researched facts. Maybe I&#39;d be more intense about it if I were an academic librarian, but even the databases UIUC offers are implementing AI...

Replies (1)

@ErinPtah@universeodon.com 2026-05-12 06:01

These bots keep getting shoved into areas where you would think "basic accuracy" should be the minimum standard. My web host put an LLM interface in their customer service, and every time I've had to use it, it wastes my time with responses like "try changing this setting that doesn't exist." It sucks, and I hate that managers keep getting suckered into buying it. Users regularly get dangerously-wrong answers from LLMs. "Cleaning tip: Mix bleach with vinegar" isn't just "not the purest scholarly research," it's deadly! (You might think something like "a philosophical chat about questions that don't have objectively right answers" would be safer...until you look at the logs from pretty much any case of chatbot psychosis, and it turns out, nope, those can end in serious-to-fatal harm too.)

Open ##2962415