Re: AI slop "contributions" to IETF working groups

[Phillip Hallam-Baker <[e-mail]>]

LLMs aren't really Artificial Intelligence, they are SI, Superficial
Intelligence.

The reason SI is so popular in the management class is that most got where
they were through superficial expertise all along. People who got where
they are through bluffing are usually unable to believe anyone else knows
what they are talking about.

Back in the days of the Soviet Union, it was much safer to investigate
obvious nonsense like parapsychology because everyone in the field knew
they were a fraud and there was very little risk of being called out for
making a 'mistake'. It was the hard scientists who went to the gulag for
discovering that the remote viewing and psychic interrogation techniques
didn't work.

There are real applications but just like in the early years after the
discovery of radium, the number of people applying it to things that were
bad or actively dangerous was vastly higher than the number of valid uses.
Hence that time I got the fit of my shoes checked with an X-Ray machine
when I was a child.

Summarizing lists of data is a task it seems to do pretty well. But it
needs a skilled operator to craft the prompt. A use I was considering was
for a Quora like forum in which a user asking a question like 'where do I
find the paint for R2D2s blue parts' would be asked if an AI generated
response answered their question. The main goal being to reduce the number
of tedious posts attacking people for asking questions that have been asked
before which are usually vastly more of an annoyance than the original
question.



On Sat, Feb 7, 2026 at 6:50 PM Rob Sayre <[e-mail]> wrote:

> Hi,
>
> Nick Hilliard <[e-mail]> wrote:
> > Are there any general IETF guidelines for how to use LLMs
> > constructively? (Or have I finally attained pure dinosaur status by not
> > hopping on the llm bandwagon?)
>
> They explore things very broadly and very quickly, but they are unreliable
> narrators.
>
> https://en.wikipedia.org/wiki/Unreliable_narrator
>
> So, they are a fine tool to use, especially when picking through mounds of
> data. But sending it to the list unedited is another problem:
>
>
> https://www.amjcounselling.com/amj-blog/2015/03/24/contempt-for-reader-you-are-showing
>
> Not a bad treatment. Contempt for the reader. So, you take the Blaise
> Pascal quote typically translated as
>
> "I would have written a shorter letter, but I did not have the time."
>
> But it says:
>
> "I wouldn't have made this one so long if I'd had the leisure to make it
> shorter." ("Je n'ai fait celle-ci plus longue que parce que je n'ai pas eu
> le loisir de la faire plus courte.")
>
> It is an apology, not a confident statement.
>
> Basically, don't send things you haven't read yourself. :)
>
> thanks,
> Rob
>
>