[ExI] e: GPT-4 on its inability to solve the symbol grounding problem

Mon Apr 17 04:03:21 UTC 2023

Gordon,
Let me re-formulate the Peanut Argument parallel with what you have been
saying all this time. To be more precise is more like a Peanut Butter
argument in reverse. Imagine I show you some artificial life I created from
some mixture of ingredients like amino acids, lipids, sugars, and so on. I
just put them in a big flask and exposed them to light and made them go
through an elaborate process of accelerated evolution. I then get a life
form, maybe quite primitive but that behaves very similarly to a living
organism like a bacterium. It seeks its food, avoids danger, can reproduce,
and do a lot of things that we usually associate with life.

Then somebody that believes life is something that goes beyond the
understanding of science comes along and points to a Peanut Butter jar and
says all the bullshit the guy in the video I already linked (and I linked
again here for reference) to basically claim my artificial life cannot
possibly be life because it is made of the same ingredients than the Peanut
Butter in the jar and see this stuff cannot possibly produce life because
it simple stuff.
The analogy that I'm trying to make is your insistence that "simple
statistical inference" (organic material like the ones in the Peanut
Butter) cannot produce "understanding or consciousness" (life).
In a sense, your fallacy is even stronger than the guy in the video because
in his case he is doubting that our explanation for life on earth is
correct (that as strong it may be is indirect evidence).

In our case, we did create something from a particular process that
possesses characteristics that are very similar if not identical to
understanding and maybe even consciousness. We know how we got here.
But you are saying, no because you used this stuff your claim that GPT-4
understands is wrong because these simple ingredients cannot make something
as complex as meaning or consciousness.
This is a denial of the phenomenon of emergence per se when our own
existence is evidence that this happens all the time in nature, combining
relatively simple things you can get complex behavior (the entirety of
chemistry and solid state physics are other examples but basically the
entire universe are examples of emergent behavior).
Gordon, what is different between what you say and what the guy in the
video is saying?

Giovanni

https://www.youtube.com/watch?v=86LswUDdb0w

On Sun, Apr 16, 2023 at 8:42 PM Giovanni Santostasi <gsantostasi at gmail.com>
wrote:

> Let's stick to math because it is easier to think about. Are you aware
> that the entire sequence of natural numbers can be derived from the null
> set? Have you ever seen the proof?
> You can read it here, there are several references.
>
> https://math.stackexchange.com/questions/4039712/how-do-you-generate-the-numbers-from-an-empty-set
>
> You do need something, the idea that the null set exists. Ok, I can
> concede you need nothingness and then you can derive everything as a
> re-arrangement (relationships) of this nothing. By the way, as I pointed
> out before you can build a universe in this way. So no, nothing is required
> to establish a language besides the symbols in the language. Maybe you are
> thinking of Godel incompleteness theorem but that is another thing. In that
> case, is about truth but language doesn't require to have the perfect truth
> that mathematics logic seeks. Language is fuzzy, it has to be just
> approximately (or probabilistically true) to have meaning.
>
> Giovanni
>
>
> On Sun, Apr 16, 2023 at 8:25 PM Gordon Swobe <gordon.swobe at gmail.com>
> wrote:
>
>> “5” can be expressed formally also as “V”, I meant. These symbols point
>> to what we really mean by 5, which is outside of the language of
>> mathematics. It’s the same with English words.
>>
>> -gts
>>
>> On Sun, Apr 16, 2023 at 9:18 PM Gordon Swobe <gordon.swobe at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Sun, Apr 16, 2023 at 7:43 PM Giovanni Santostasi <
>>> gsantostasi at gmail.com> wrote:
>>>
>>>>
>>>> *To know the difference, it must have a deeper understanding of number,
>>>> beyond the mere symbolic representations of them. This is to say it must
>>>> have access to the referents, to what we really *mean* by numbers
>>>> independent of their formal representations.*What are you talking
>>>> about?
>>>>
>>>
>>>
>>> Talking about the distinction between form and meaning. What applies to
>>> words applies also to numbers. The symbolic expression “5” for example is
>>> distinct from what we mean by it. The meaning can be expressed formally
>>> also as “IV” or ”five.”
>>>
>>>
>>> LLMs have access to and are trained only on the formal expressions of
>>> both words and numbers, not their meanings.
>>>
>>>
>>> -gts
>>>
>>>
>>>> *“1, 2, 3, 4, Spring, Summer, Fall, Winter” and this pattern is
>>>> repeated many times.   *
>>>> Yeah, this is not enough to make the connection Spring==1, Summer==2
>>>> but if I randomize the pattern 1,3,4,2, Spring, Fall, Winter, Summer, and
>>>> then another randomization eventually the LLM will make the connection.
>>>>
>>>> On Sun, Apr 16, 2023 at 3:57 PM Gordon Swobe via extropy-chat <
>>>> extropy-chat at lists.extropy.org> wrote:
>>>>
>>>>>
>>>>>
>>>>> On Sun, Apr 16, 2023 at 2:07 PM Jason Resch via extropy-chat <
>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>
>>>>>
>>>>> To ground the symbol "two" or any other number -- to truly understand
>>>>>>> that the sequence is a sequence of numbers and what are numbers -- it needs
>>>>>>> access to the referents of numbers which is what the symbol grounding
>>>>>>> problem is all about. The referents exist outside of the language of
>>>>>>> mathematics.
>>>>>>>
>>>>>>
>>>>>> But they aren't outside the patterns within language and the corpus
>>>>>> of text it has access to.
>>>>>>
>>>>>
>>>>>
>>>>> But they are. Consider a simplified hypothetical in which the entire
>>>>> corpus is
>>>>>
>>>>> “1, 2, 3, 4, Spring, Summer, Fall, Winter” and this pattern is
>>>>> repeated many times.
>>>>>
>>>>> How does the LLM know that the names of the seasons do not represent
>>>>> the numbers 5, 6, 7, 8? Or that the numbers 1-4 to not represent four more
>>>>> mysterious seasons?
>>>>>
>>>>> To know the difference, it must have a deeper understanding of number,
>>>>> beyond the mere symbolic representations of them. This is to say it must
>>>>> have access to the referents, to what we really *mean* by numbers
>>>>> independent of their formal representations.
>>>>>
>>>>> That is why I like the position of mathematical platonists who say we
>>>>> can so-to-speak “see” the meanings of numbers — the referents — in our
>>>>> conscious minds. Kantians say the essentially the same thing.
>>>>>
>>>>>
>>>>> Consider GPT having a sentence like:
>>>>>>  "This sentence has five words”
>>>>>>
>>>>>> Can the model not count the words in a sentence like a child can
>>>>>> count pieces of candy? Is that sentence not a direct referent/exemplar for
>>>>>> a set of cardinality of five?
>>>>>>
>>>>>
>>>>> You seem to keep assuming a priori knowledge that the model does not
>>>>> have before it begins its training. How does it even know what it means to
>>>>> count without first understanding the meanings of numbers?
>>>>>
>>>>> I think you did something similar some weeks ago when you assumed it
>>>>> could learn the meanings of words with only a dictionary and no knowledge
>>>>> of the meanings of any of the words within it.
>>>>>
>>>>>
>>>>>>>>
>>>>>> But AI can't because...?
>>>>>> (Consider the case of Hellen Keller in your answer)
>>>>>>
>>>>>
>>>>>
>>>>> An LLM can’t because it has no access to the world outside of formal
>>>>> language and symbols, and that is where the referents that give meaning to
>>>>> the symbols are to be found.
>>>>>
>>>>> -gts
>>>>> _______________________________________________
>>>>> 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/20230416/2706664c/attachment.htm>