[ExI] LLM hyperbole

[BillK]

On Fri, 26 Jun 2026 at 19:59, spike jones via extropy-chat <
extropy-chat at lists.extropy.org> wrote:
>
> Ja, and I really miss OK.  Everything is awesome and excellent and
perfect.
> Drives me nuts.  On the phone with the medics, they ask questions, I
answer,
> everything is perfect or awesome.  I want OK back.
>
> Ja.  We also lost bad.  Anything less than OK sucks now.  That one I don't
> understand, for that term can be a good thing.  The original meaning was
all
> good.  I don't understand how a term could turn thru pi radians.  I want
> good, bad and OK back.
>
> spike
> _______________________________________________


There are also the changes due to the widespread use of AI
People use AI to edit their writing and even start to talk like AIs.
BillK

iAsk AI -
The Linguistic Influence of Large Language Models

The integration of Large Language Models (LLMs) into daily life has
initiated a measurable shift in human communication, creating a "closed
cultural feedback loop" where the patterns generated by AI are increasingly
reflected in human speech and writing.  Researchers have observed that
specific vocabulary favored by models—such as "delve," "meticulous,"
"pivotal," and "intricate"—has surged in frequency across both academic
papers and casual, unscripted podcasts since the public release of
ChatGPT.  This phenomenon suggests that humans are not merely using AI as a
tool, but are subconsciously adopting its stylistic cadence, transition
phrases, and grandiose adverbs, effectively training their own
communication patterns on the output of the machines.
Mechanisms of Linguistic and Cognitive Drift

The transition toward "LLM-speak" is driven by several factors, including
the convenience of productivity tools and the rapid diffusion of technology
into professional and creative workflows.  In academia, studies have
quantified this shift, finding that nearly 17.5% of computer science papers
and a significant portion of peer review texts contain content drafted or
modified by AI.  This adoption is often motivated by the need for speed and
the pressure to manage high volumes of communication, which leads users to
rely on the model's structural templates.  Beyond simple vocabulary, users
report a change in their own cognitive processes, noting that the "rhythm"
of AI-generated text—often characterized by specific transition structures
and a preference for em dashes—begins to influence their original,
unassisted writing.  This creates a form of "cognition leakage," where the
logic and style of the model outcompete original human thought, potentially
flattening the diversity of human expression as both humans and AIs
continue to train on a narrowing subset of increasingly homogenized data.
---------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260626/27ff941f/attachment.htm>