[ExI] Against AI Understanding and Sentience: Large Language Models, Meaning, and the Patterns of Human Language Use

Giovanni Santostasi gsantostasi at gmail.com
Mon Apr 17 19:24:03 UTC 2023


No matter how may papers these philosophers write about this topic a single
empirical piece of evidence will destroy their positions. Well, if they
listened to logic and cared about the empirical fact. Anyway, just using
some arguments based on what these philosophers think LLM are about (and
for the vast majority they don't and also we don't given we don't know what
is happening the black box). One has to do the experiment. Experiments like
the one below. I already posted in another thread but it deserves to be
posted here too. If this is not real understanding I don't know what
understanding really is. And nobody does then. If we had an exchange with a
child as the below we would think this is a very attentive and clever
child.


Ok I told GPT-4 I like the idea of the cloud but I pointed out that the sky
is blue and the cloud are white usually. It went ahead and fix that. Easy
task but notice the error it made.
[image: image.png]
Gio:
Do you notice a possible mistake in this rendition (just the last
modification about the use of the color blue), do not yet write the new
code before we agreed on this possible mistake, go ahead and tell me what I
think it is wrong.
GPT-4:
Yes, I see the mistake in the last modification. The background sky color
is covering the entire SVG area, including the ground portion. The sky
color should only cover the area above the ground, while the ground should
have its own color. This can be fixed by adjusting the height of the sky
rectangle and adding another rectangle for the ground.

On Mon, Apr 17, 2023 at 8:46 AM Gordon Swobe via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> More academics pushing back against the hype about "conscious large
> language models."
>
> I read the paper and found some interesting discussion about
> "distributional semantics," and about some of the reasons these models seem
> so proficient in the use of language. The authors draw on Wittgenstein and
> emphasize the patterns in common language usage as distinct from only the
> raw statistics.
>
> Abstract
> Large language models such as ChatGPT are deep learning architectures
> trained on immense quantities of text. Their capabilities of producing
> human-like text are often attributed either to mental capacities or the
> modeling of such capacities. This paper argues, to the contrary, that
> because much of meaning is embedded in common patterns of language use,
> LLMs can model the statistical contours of these usage patterns. We agree
> with distributional semantics that the statistical relations of a text
> corpus reflect meaning, but only part of it. Written words are only one
> part of language use, although an important one as it scaffolds our
> interactions and mental life. In human language production, preconscious
> anticipatory processes interact with conscious experience. Human language
> use constitutes and makes use of given patterns and at the same time
> constantly rearranges them in a way we compare to the creation of a
> collage. LLMs do not model sentience or other mental capacities of humans
> but the common patterns in public language use, clichés and biases
> included. They thereby highlight the surprising extent to which human
> language use gives rise to and is guided by patterns.
>
> ....
>
> "Humans are prone to attribute agency even to geometric shapes that move
> in seemingly
> intentional ways (Heider and Simmel 1944). They are all the more inclined
> to anthropomorphic misinterpretation when interacting with a seemingly
> intelligent system of unprecedented power. Especially susceptible are those
> who are lonely, socially disconnected, or otherwise vulnerable (Epley,
> Waytz, and Cacioppo 2007), but given the natural propensity of immediately
> ascribing agency, anybody may be tempted to anthropomorphic
> misinterpretations."
>
>
> https://philsci-archive.pitt.edu/21983/
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230417/58ae3438/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 8487 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230417/58ae3438/attachment.png>


More information about the extropy-chat mailing list