2026-07-01 No AI regurgitating
==============================

Somebody asks a question. Here are the answer patterns rated from top
to bottom:

The person knows the answer. End of story. This is great.

The person boosts the question so that somebody else can answer. Nice!
Thank you.

The person connects emotionally. I like to know that I've been heard
and understood, but people with a lot of followers don't always
appreciate it.

The person connects anecdotally. I always like to read about other
people's lives but people with a lot of followers don't always
appreciate it.

The persons knows the answer but didn't realize that dozens of correct
answers have already been posted. Oh well. It's the fediverse and I
also don't always click through to the original (and that only works
for public posts, so sometimes there is no good solution).

The person disconnects emotionally, answering with misplaced humour,
irony, sarcasm, nihilism, etc. There is no answer, no emotional
support and no interesting diversion.

The person is condescending or shows a lack of benevolent reading
skills. It doesn't matter whether a correct answer is hidden somewhere
in the thorns.

The person asked a chat bot and pastes the generated text. It doesn't
matter whether a correct answer is hidden somewhere in the slop.

The person asked a chat bot and reports that no plausible text was
generated.

Related: Trying to manufacture permission by tante argues that people
who deliberately introduce their argument by admitting a chat bot has
generated the text are trying to absolve themselves of responsibility
and repercussions. "The chat bot did it!" is the universal excuse for
every botched job.

Maybe I should add this to Etiquette of online posting.

Social norms are tricky.

Recently I was talking with somebody about Longjing tea. I mentioned
that I had once bought Longjing 43 but that I didn't know what the 43
stood for. They say they asked a large language model (LLM) and then
they paste it's long, generated text into our chat. I tell them that
it hurts to see people use a LLM and say that when I search for the
term, the first hit is from the shop where I bought it. To which they
reply, that it was an interesting page but that it still didn't
explain where the number 43 came from. And therefore, they say, even
though they don't like LLMs, they still use them for searches like
this. Do I want to start a discussion about some AI companies
"optimizing the kill chain"? The water use and energy use? The abuse
of low-paid workers in Kenya? The deterioration of call-centres and
service-centres in general? The moment passed without me saying
anything.

We've had a long history of looking up things in lexicons and
encyclopedias. So I think more thought is required. We also need to
identify areas where LLMs are used but but where we didn't use books
and Wikipedia before, like when somebody tells me they had a LLM
"check" their new work contract. WTF does that even mean and how am I
supposed to react? I was so baffled that the moment passed without me
saying anything.

To me, this often feels like talking to people who believe in
alternative medicine like homeopathy, chiropractics, Bach flower
remedies, Chinese traditional medicine and so on. What do you even
say if the people involved are your friends and family. I usually say
that I am a firm believer in evidence-based medicine and that's where
the conversation ends.

Not all conversations are equal. What I can say for my own online
discourse, including conversations around the software projects I
maintain, is that I don't want to spend my free time talking to
machines. Therefore, I prefer to end such conversations as soon as
possible; I'm happy to add an explanatory note. If a person then
emerges to talk with, that's great. If not, I've saved myself the
time. (Just to be clear, I also don't want to spend my paid time
talking to machines.)

#AI #ButlerianJihad