2026-04-27 03:25 UTC
It's 2026, and LLMs are still making up bullshit like it's 2022. Remember all the people who said in 2023 stuff like "hallucinations are a solved problem and all the popular models will roll out the non-hallucinating updates soon"? Um, they were wrong, consider not taking them too seriously in the future.
Replies (1)
-
@aharoni@wikis.world 2026-04-27 03:30
I am, however, concerned that lots of people will use LLMs to create articles anyway. Some will be caught, some won't be. My prediction is that "AI" bans won't really work, and what will actually happen is that the requirements for reliable sources are going to get even stricter. For example, birth and death dates are often not checked strictly, but they are probably going to be. This kind of sucks for reasons of practicality and equity, but it's eventually good for reliability.