[ExI] Do all AI models represent “cat” in the same way?

[Mike Dougherty]

On Sat, Jan 17, 2026, 9:57 AM Adrian Tymes via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

>
> The claim is that Chinese or Native American fundamental depictions of
> things like "cat" are massively different from Western depictions, so
> when AI keeps distilling to Western depictions, that's proof that said
> other viewpoints are being excluded or not considered.
>
> That's the claim, anyway.  From what I have seen of the "true"
> cultures, many times the underlying concepts aren't actually all that
> different, no matter how much they may feel different - or, the things
> that are claimed as different amount to folklore, akin to saying that
> we should give real credence to concepts such as the Fair Folk and
> cryptobiology.
>

https://www.pnas.org/doi/10.1073/pnas.1015975108

https://altalang.com/beyond-words/language-shape-thought/

Obviously the obscure/oddity of isolated tribal groups will be
underrepresented in LLM training data - but the fact that we have these
references in non-AI mental models is interesting.

While looking for those links (Guugu Yimithirr, absolute direction) and
(Mundurukú, inexact number) I saw an overly dense article on communication
- and it made me think about the size/ complexity of the mental model and
the robustness/ complexity of the grammar/ language protocol used to
describe or convey the mental model from one computing context to another.

Multi-billion parameter LLM seem pretty good at convincing humans they
'understand' - I guess we're close enough to human equivalent because
[let's be honest] humans are so resilient to ambiguity and inexactness.  I
am curious how AI will evolve language to communicate with each other.

https://pub.towardsai.net/toon-vs-json-a-comprehensive-performance-comparison-446a2fb82f20

TOON is a data format that is 'cheaper' to communicate structured data to
AI than the format programmers had been using (JSON) after it became
obvious that it was better than the previous (XML) and archaic (CSV)
formats.

Idk if there was much more point made in this email for as long as it took
me to write it,  but these ideas were with sharing.  This is the modern
infovore analog to "look at these pretty rocks" ��

>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260117/9dc89afb/attachment.htm>