[ExI] all we are is just llms

[Ben Zaiboc]

On 22/04/2023 11:01, Gordon Swobe wrote:
> On Sat, Apr 22, 2023 at 2:43 AM Ben Zaiboc via extropy-chat 
> <extropy-chat at lists.extropy.org> wrote:
>
>     (you think that pointing is not a language? I suspect many deaf
>     people would disagree)
>
>
> Fine with me. Sign language is also a form of language.
>
>     This is why referring to linguistics is not helping.
>
>
> ? Because we are going to include sign language in our definition of 
> language, linguistics is not helping? Linguists consider sign language 
> also to be a form of language.
>
> In our primitive caveman example, in which he points at let us say an 
> animal, his first "words" in sign language translate to something like 
> "Look over there! See what I see?" Based on how frantic or calm is his 
> gesturing, his interlocutor might also know if his friend perceives 
> the animal as a threat or as food. Now he has two words. Before long, 
> Fred and Barney are also grunting identifiable noises as their sign 
> language evolves into more complex verbal language.
>
>     As I said earlier, it's the wrong discipline here.
>
>
> Language models model language and linguistics is the science of 
> language.
>
> > Referents, being internal conceptual models, /are made of language/. 
> They must be, because there's nothing else to work with, in the brain.
>
> Really? My brain has visual perceptions and sounds and imaginations 
> and non-verbal thoughts and music and many things going on that can be 
> /described/ with language but are not language.
>
> I understand what you are trying to say about the "language of the 
> brain" but I would say you are conflating neurology and language.
>
> The statement "referents are made of language" is simply false on the 
> definition of referent. Only a tiny subset of words in the English 
> language have language as referents. Linguists call them meta-words. 
> They are parts of speech and similar. For example, the word "nouns" 
> refers to the set of all nouns.

No no, you've completely shimmied past what I'm saying, there.

Partly my own fault:

    This is why referring to linguistics is not helping.


? Because we are going to include sign language in our definition of 
language, linguistics is not helping? Linguists consider sign language 
also to be a form of language.


No, I was referring to your whole reply. I'm looking at the fairground 
ride, trying to see how it works, how it compares to the way we work, 
and what that means for the future. You're analysing the cash flow.
Yeah, ok, forget the strained analogy.

I'm not conflating neurology and linguistics (which is what I assume you 
mean when you say 'language' here), I'm saying that neurology is the 
relevant discipline for analysing this, and linguistics is not.


 > My brain has visual perceptions and sounds and imaginations and 
non-verbal thoughts and music and many things going on that can be 
/described/ with language

Precisely.

 > but are not language

They are constructed with a specific language. My whole point is, as the 
brain can experience visual perceptions and sounds and imaginations and 
non-verbal thoughts and music and many things, when all it has to use 
are neural spike trains, which are binary signals, then all those 
experiences must necessarily be made from the brain's language of binary 
signals.

 > Language models model language and linguistics is the science of 
language.

If that was all they did, they wouldn't be very interesting or useful. 
Except to linguists.

I don't know much about linguistics, but I understand it to be the study 
of human languages. Not the study of AI. Just because "Large Language 
Models" has the word 'language' in it, doesn't mean that studying human 
languages is relevant. As I said before, we could call them Large Means 
of Communication Models. Would that make Communication Studies the 
relevant discipline?

You might find this guy's posts interesting:
https://seantrott.substack.com/p/humans-llms-and-the-symbol-grounding-fc4

Ben

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230422/4162367d/attachment.htm>