On Sat, Jan 3, 2026 at 8:51 PM Alejandro Colomar wrote:
>
> One may consider gambling with the fact that it might be difficult to
> prove it. In a project that I maintain, not being a democracy, I'll ban
> someone with high suspicion. One may also consider gambling on how much
> ISO would ban anyone only with high suspicion.


I find that a bit worrying, especially as some people have a very
unique way of writing -- and not just the neurodivergent. Even some
neurotypicals have some very unique ways of writing. It would be
horrible if somebody were to write something by themselves and then
later to be "found" to have used an LLM. If you run the 1776 US
Declaration of Independence through some LLM-detectors, you get a
false positive:

    https://www.forbes.com/sites/jodiecook/2024/07/04/ai-content-detectors-dont-work-the-biggest-mistakes-they-have-made/


> I find it interesting when people admit trusting the quality of AI (or
> other humans, FWIW). To me, it tells a lot where their threshold for
> quality is. "Tell me who you go with and I'll tell you who you are."


'Trust' is a strong word . . . I tend to second-guess everything that
comes out of an LLM and then go try verify it myself. But I do the
same thing with Wikipedia and a few other sources too. It's rare that
I would accept something immediately as fact, even if it comes from
National Geographic. Some of the stuff that comes out of ChatGPT
version 5 in "long thinking" mode though is very impressive to me. I
don't find it embarrassing to admit this even though I'm getting the
vibe here that I'm supposed to be embarrassed about it. I think a lot
of people just still have an aversion toward the idea of an artificial
or unnatural form of intelligence or intellectualism.