[ExI] Why AI agents invent their own language if you let them chat.

[Adrian Tymes]

On Wed, Oct 7, 2026 at 6:22 AM John Clark <johnkclark at gmail.com> wrote:
> No. Appeals to authority are NOT always fallacies.

Yes they are.

Authorities may be a useful heuristic, such as if you can't repeat the
experiment.  But when they make simple claims of logic, you can follow
them.

> when they make a mistake, which is rare, they immediately print a retraction.

A mistake of facts, yes.  A mistake of sensationalism, not so much.

> But I'm supposed to ignore them but pay attention to you?

No.  You are supposed to evaluate what you can, regardless of source.

This is not about me.  As with any matter of pure logic, the speaker
is irrelevant.

If your emotions insist that I am trying to make a claim about me
here, then your emotions - not me, and not the journal Science - are
what are misleading you.  I recommend disengaging them and looking at
this from a neutral logical point of view - without starting off
trying to prove a position.

* The journal Science has presented some quotes from AI removed from
their original context.
* The journal Science presented them to suggest that they are hard to
understand.
* I pointed out that, if a quote is removed from its original context,
it can be harder to understand regardless of whether it would be easy
to understand in context.
* I presented my own recollection of seeing those claims in context
and being able to understand them, as a counterpoint to the claim that
these quotes are (without qualification) hard to understand.

I further suspect that you saw the original instances of these quotes,
in context, too - though I do not know if you remember this.