[ExI] Bender's Octopus (re: LLMs like ChatGPT)

[William Flynn Wallace]

But a picture dictionary doesn't work for an abstract system, since all
they have for their knowledge of the picture is yet another abstract word
like redness.  brent

Huh?  Can't use an example of redness?  ???  Also, to increase stimulus
generalization, you would include several photo examples.  Many words which
have an abstract use also have concrete examples, such as 'floor'.  I can
see it will have trouble with honesty or love.   bill w

On Sun, Mar 26, 2023 at 3:02 PM Brent Allsop via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

>
> Your referent for the word redness is the subjective quality your brain
> uses to represent red knowledge.
> So, a picture of red in a dictionary works for you, as your brain produces
> a redness experience when you look at it.
>
> But a picture dictionary doesn't work for an abstract system, since all
> they have for their knowledge of the picture is yet another abstract word
> like redness.
>
> On Sat, Mar 25, 2023, 10:57 AM William Flynn Wallace via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
>> I won't argue what a referent means.
>>
>> I agree.  It is just what John would say but in different words:  he
>> would emphasize, as I do, that for definitions you need examples, and that
>> is why I, tongue not totally in cheek, wrote that you should give an AI a
>> picture dictionary.   bill w
>>
>> On Sat, Mar 25, 2023 at 3:41 AM Ben Zaiboc via extropy-chat <
>> extropy-chat at lists.extropy.org> wrote:
>>
>>> Reading these conversations over the last few days, it has struck me
>>> that some people keep referring to 'real' things, usually using the word
>>> 'referents' (e.g. an apple), as though our brains had direct access to
>>> them and could somehow just know what they are.
>>>
>>> But we don't.
>>>
>>> Think about it, what is "An Apple"?
>>>
>>> It's a term that we associate with a large set of sensory and memory
>>> data, including language data, but mostly things like visual, textural,
>>> taste, smell, emotional, etc., data stored as memories.
>>>
>>> Seeing as we all have different memories associated with the label "An
>>> Apple" (because some of us were sick the first time we ate one, some of
>>> us are allergic to something in apples, some of us have a greater
>>> impression of sweetness, or sourness, when we eat one, some of us once
>>> discovered a maggot in one, some people have only ever eaten Granny
>>> Smiths, others only Braeburns, or Crab Apples, and so on and so on...),
>>> then 'An Apple' is a different thing to each of us.
>>>
>>> There is no spoon! Er, Apple. There is no Apple!
>>> Not as a 'real-world thing'.
>>>
>>> "An Apple" is an abstract concept that, despite the individual
>>> differences, most of us can agree on, because there are a lot of common
>>> features for each of us, such as general shape, some common colours, a
>>> set of smells and tastes, how we can use them, where we get them from,
>>> and so on.. The concept is represented internally, and communicated
>>> externally (to other people) by a linguistic label, that refers, for
>>> each of us, to this large bunch of data extracted from our senses and
>>> memories: "Una Manzana".
>>>
>>> It's all 'nothing but' Data. Yet we all think that we 'understand' what
>>> an Apple is. Based purely on this data in our brains (because we have
>>> access to nothing else).
>>>
>>> So this idea of a label having 'a referent' seems false to me. Labels
>>> (data in our heads) refer to a big set of data (in our heads). Where the
>>> data comes from is secondary, diverse, and quite distant, when you trace
>>> the neural pathways back to a large and disparate set of incoming
>>> sensory signals, scattered over space and time. The meaning is created
>>> in our minds, not resident in a single object in the outside world.
>>>
>>> This is my understanding of things, anyway.
>>>
>>> Ben
>>> _______________________________________________
>>> extropy-chat mailing list
>>> extropy-chat at lists.extropy.org
>>> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>>>
>> _______________________________________________
>> extropy-chat mailing list
>> extropy-chat at lists.extropy.org
>> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>>
> _______________________________________________
> 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/20230326/dc74a51a/attachment.htm>