[ExI] Emily M. Bender — Language Models and Linguistics (video interview)

Jason Resch jasonresch at gmail.com
Mon Mar 27 21:47:37 UTC 2023


On Mon, Mar 27, 2023, 5:20 PM Gordon Swobe <gordon.swobe at gmail.com> wrote:

>
> On Mon, Mar 27, 2023 at 3:02 PM Jason Resch via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
> It certainly gives us that impression, but on careful analysis of what is
>>> actually going on, we can see that is the human operator who attributes
>>> meaning to those symbols. GPT is merely very good at arranging them in
>>> patterns that have meaning to *us*.
>>>
>>
>>
>> I think that's why this particular example is so important to escape that
>> trap, because mathematical structures are objective. Which vertices are
>> connected by which edges isn't something that can be faked or
>> misinterpreted, it simply is.
>>
>
> I thought I had already mentioned that as Giovanni pointed out correctly,
> mathematics is a kind of language. The fact that GPT can translate English
> words about mathematical relationships into the language of mathematics is
> certainly impressive, but it doesn’t “escape that trap.”
>

Mathematics can be viewed as a language. But it's a language that describes
objects that exist in reality. That is to say, math contains "referents."
Godel's incompleteness confirms that mathematical structures exist beyond
any earthly or human description of them.


> When ChatGPT 3.5 first went online, I saw on twitter several examples of
> how it had failed to make those translations correctly, and I understand
> GPT-4 is much better at it, but it is still merely manipulating the symbols
> of English and Math.
>

But please explain how you think it acquired the capacity to interpret the
symbols in order to correctly draw *an image of the house", not mere
symbols about the house and not a list of mathematical language about the
house, but *an accurate picture of the house*.

It seems we're talking past each other at this point so I don't know if any
further progress can be made on this subject.

Jason

>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230327/0a2869ca/attachment.htm>


More information about the extropy-chat mailing list