this post was submitted on 19 Aug 2026
571 points (98.8% liked)
Fuck AI
8002 readers
2636 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I invented a screw driver that screws things all the way in 80% of the time.
You still need to get a regular screwdriver to check every screw to make sure it's screwed in all the way, but 80% of the time, it is!
Pretty amazing!
It does costs $3,000 dollars to buy, but think of the time you'll save.
What's the probability to be in the 80% of the time? 10%?
π€
Whatβs Wikipedia? Do you all read whatever article in front of you or do you look for the source article? Surely you are all checking sources on everything and not just acting like ai is the only place you have to do it!
Yes. If the article doesn't contradict anything that I already know and the correctness of the information isn't required to avoid bodily harm or legal repercussions then I don't usually check the sources.
There's a world of difference between the reliability of a continuously, openly and cooperatively updated encyclopedia and that of a stochastic word prediction machine that ingested the entire internet - and only gods (and its owners) know what else.
Wikipedia's articles tend to improve in accuracy over time, as more people see the same article, can discuss it and correct wrong information / add missing facts and sources.
While LLM models tend to become more capable at certain tasks, they aren't increasingly reliable at providing accurate information. And there aren't the same stabilizing forces in play as with a human-edited encyclopedia. You can get correct answers today and incorrect answers tomorrow for the same question. Maybe it's because the model was updated. Maybe it's because you used sightly different words to ask the question. Maybe nothing changed and it was simply a random deviation because 0-temperature models aren't flexible enough.
The point is, you have no way to know. No way of knowing why the answer is wrong this time. No way of ensuring it isn't wrong again tomorrow. Not even any way of ensuring it isn't wrong in the same way tomorrow.
So it's not just that the likelihood of a wrong answer is higher with AI, it's also much harder to predict what kind of error it might make or when or why and adjust accordingly.
You're essentially playing roulette for information.
Is that trolling?
Are you trolling, bro?