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

Brent Allsop brent.allsop at gmail.com
Wed Apr 12 01:30:27 UTC 2023


Hi Jason,
Thank you for being such an educated listener, sounding board, indicator of
whether you can accept this as a falsifiable hypothetical possibility.
And especially thanks for being willing to think with my assumption that
physical qualities are elemental.  I know this is very hard for
functionalists, that think functionality is more elemental, and that
redness can "arise" from function, rather than function being
implemented with physical redness.  I know most functionalists (Giovani,
I'm referring to you, for one, but this says more about the weakness of
this theory, and my ability to describe it, than any weakness in a great
intellect like Giovani) seem to be unable to do that.
[image: 3_robots_tiny.png]

You need to be very complete, with what you mean by "functionally
equivalent"  it must be something that includes the function which
something like glutamate provides, which is the redness quality.  So that
when the above three systems are asked: "What is redness like for you."
The brain must be able to be aware of it's redness quality, and provide
these honest, physically grounded, answers:
1. My redness(glutamate) is like your redness(glutamate).
2. My redness(glycine) is like your greenness(glycine).
3. My knowledge is abstract, and not like anything.  It is like the word
'red' and I have no idea what a redness quality is like, as I have no
ability to experience redness.
Note:  In effect, this is all very similar to the way all chat bots
accurately respond to questions like this, even if it takes a bit of
convincing to get them to agree with that.
Note:  Once we discover what it is, in our brain, which has a redness
quality, we will be able to endow AGIs with this (like when Cmd Data, in
Star Trek, received his "emotion chip."  It would then be able to say
things like: "Oh THAT is what your redness is like."  And it would then fit
our definition of phenomenal consciousness.

Asking about a molecule level simulation is a good question.  I haven't
thought about that as much as the neuron level simulation/substitution
but there must be some set of molecules (maybe all the molecules that make
up a bunch of neurons and their neurotransmitters) that is behaving the way
it does, because of its computationally bindable redness quality.
An abstract molecule level simulation might be able to behave, identically,
including making the claim that it's redness was like your glutamate
redness, but since you could eff the ineffable nature of real glutamate,
you could objectively know it was achieving those responses, and know it
was lying.
You would know that nothing in its brain was glutamate, and nobody was ever
able to experience redness, (no matter how you simulated that redness)
without glutamate.

On Tue, Apr 11, 2023 at 6:27 PM Jason Resch via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> On Tue, Apr 11, 2023, 8:16 PM Brent Allsop via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
>> As far as the neural substitution argument
>> <https://canonizer.com/topic/79/2> goes,
>> Physicalism
>> <https://canonizer.com/topic/88-Theories-of-Consciousness/7-Qualia-are-Physical-Qualities>predicts
>> that there is something in the brain which behaves the way it does, because
>> of its redness quality.
>> The example, easily falsifiable physicalist theory (it will probably be
>> physically different), predicts that there is one or more neurons that are
>> responsible for each voxel of 3D color quality knowledge we have of what we
>> see.
>> If one of them is firing with Glutamate, it could be that glutamate,
>> behaving the way it does, in the computationally bound synapse, because of
>> it's redness.
>> The theory predicts that nothing but that glutamate would have the
>> physical property we experience, that is the redness quality.
>> When you get to that first pixel of redness (the first neuron firing with
>> glutamate, representing the first pixel of visual knowledge)
>> Nothing but that glutamate will be able to reproduce the physical redness
>> quality, which glutamate has, for that one pixel of visual knowledge.
>> In other words, the neural substitution will fail at that point, as
>> nothing will be able to present the same redness quality to the binding
>> system resulting in a redness experience for that first pixel of redness
>> quality.
>>
>> I know the functionalists argue that this would not possible, because
>> there will be lots of other stuff (such as abstractly simulated neurons)
>> which will result in the same downstream effects, ultimately leading to the
>> identical output (the person saying: "that pixel of redness is the same")
>> But this model where neurons only work like discrete logic gates, alone,
>> does not account for the kind of computational binding of voxel elements of
>> qualities of which our visual conscious knowledge is composed.
>> So there must be some physical binding system which can detect a physical
>> difference (redness changing to something else like greenness) and be able
>> to report that particular physical change, from redness to greenness, for
>> example.
>> And it must be able to report that pixel of glutamate, (redness) is not
>> real glutamate, unless it is actual physical glutamate.  And nothing but
>> physically real glutamate will behave like glutamate (have it's redness
>> quality.)
>>
>
> Hi Brent,
>
> I appreciate your detailed answer.
>
>>
> In summary, is it correct to say that you don't think it is possible to
> create a functionally equivalent artificial neuron, because it might lack
> glutamate (or something necessary for redness like that) and nothing but
> glutamate can function like glutamate?
>
> Let's assume that much is true. What if we did a molecular-level
> simulation of the brain, such that the computer is able to account for all
> the intramolecular forces and reactions that glutamate and all the other
> proteins and molecules in the brain have? Would *that* not yield a
> functionally identical result in your estimation?
>
> Jason
>
>
>
>
>>
>> On Tue, Apr 11, 2023 at 12:11 PM Jason Resch via extropy-chat <
>> extropy-chat at lists.extropy.org> wrote:
>>
>>> On Tue, Apr 11, 2023, 1:48 PM Gordon Swobe <gordon.swobe at gmail.com>
>>> wrote:
>>>
>>>> On Tue, Apr 11, 2023 at 3:39 AM Jason Resch via extropy-chat <
>>>> extropy-chat at lists.extropy.org> wrote:
>>>>
>>>>> On Tue, Apr 11, 2023, 1:30 AM Gordon Swobe <gordon.swobe at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> On Mon, Apr 10, 2023 at 5:36 PM Jason Resch via extropy-chat <
>>>>>> extropy-chat at lists.extropy.org>
>>>>>>
>>>>>> There are many more aspects to my argument, but that is where it
>>>>>> starts. The brain/mind is not fundamentally a computer or information
>>>>>> processing machine.
>>>>>>
>>>>>
>>>>> What is it then?
>>>>>
>>>>
>>>> More later, but generally I think computationalism and functionalism in
>>>> general fail to adequately explain subjective experience.
>>>>
>>>
>>> Below you say subjective experience cannot be explained in any this
>>> person objective way. Does this mean there can be non possible theory of
>>> mind?
>>>
>>> I agree with you that subjective experiences are not communicable or
>>> shareable in a third person way. Nor do I think they are deniable, as some
>>> eliminative materialists are apt to do.
>>>
>>>
>>> I notice that the response for my detractors is always to attempt to
>>>> explain subjective experience away as something else or even to dismiss it
>>>> as an illusion.
>>>>
>>>
>>> I do not dismiss subjective experience as an illusion. It's the only
>>> thing we can know for sure. All else is an inference.
>>>
>>>
>>>> For example, you seemed reluctant even agree that something as simple
>>>> as the first person experience of a toothache could be understood for what
>>>> we all know it to mean.
>>>>
>>>
>>> We feel a pain directly, yes. But that doesn't mean pain is simple. Red
>>> looks simple too but it takes millions if not billions of neurons working
>>> together to produce your visual experience. Think of it like the UI on your
>>> phone. So much is necessary and so much is haokeno below the scenes to give
>>> you that simple layout of a few buttons. The simple appearance at the top
>>> betrays the underlying complexity. I think human is conscious like that.
>>>
>>>
>>>
>>>> As I wrote in one of my first exchanges with you, I consider experience
>>>> to be primary. By primary I mean irreducible. It cannot be captured or
>>>> explained away in the third person objective language of science. This is
>>>> also known as the explanatory gap after Thomas Nagel.
>>>>
>>>
>>> I don't dispute this.
>>>
>>> But a theory of mind can still succeed and be of value without making
>>> qualia communicable. Consider that computationalism is just the theory of
>>> mind thats left over after accepting the following two conjectures: 1.
>>> brains can be accurately simulated, and 2. zombies are logically
>>> inconsistent. This theory tells us we can preserve ones qualia and
>>> conscious with a valid functionally equivalent substitution of some or all
>>> of their brain, but this theory makes no claim of making qualia shareable
>>> in third person terms (which I accept is likely impossible for any theory,
>>> due to Godelian limits of self-reference, e.g. a person who likes chocolate
>>> cannot know what it is like to dislike chocolate).
>>>
>>>
>>>
>>>>
>>>>>
>>>>>> In summary, computationalism is *not* the idea that the human brain
>>>>>>> operates like a computer, but rather, that a computer can be made to
>>>>>>> operate *like a human brain*.
>>>>>>>
>>>>>>
>>>>>> Yes, that is the doctrine which might as well be hanging on the front
>>>>>> door of ExI.
>>>>>>
>>>>>
>>>>>
>>>>> On what basis do you doubt this?
>>>>>
>>>>>
>>>>>>
>>>>>>> We know from the Church-Turing thesis that computers can replicate
>>>>>>> the operations of any finitely describable system.
>>>>>>>
>>>>>>
>>>>>> In my view, we can simulate the brain in a computer, similar to how a
>>>>>> meteorologist might simulate a hurricane on a computer ,
>>>>>>
>>>>>
>>>>>
>>>>> I am glad we agree on this.
>>>>>
>>>>> but unless we live in a digital simulation ourselves (another
>>>>>> religious doctrine),
>>>>>>
>>>>> the simulation is not the same as the thing simulatied.
>>>>>>
>>>>>
>>>>> If we cannot rule out the hypothesis that we live in a digital
>>>>> simulation, then doesn't that show there's no detectible difference between
>>>>> the real and the simulated (when you exist at the same level of the
>>>>> simulated)? How do we know our universe isn't a simulation, couldn't it be?
>>>>>
>>>>> A simulated hurricane will make you wet, when you are also inside the
>>>>> simulation. Then would not a simulated brain, be conscious within the level
>>>>> of the simulation of the brain?
>>>>>
>>>>> If you believe in a soul existing on another plane of reality, then
>>>>> doesn't that information about it, (your self, memories, personality,
>>>>> essence, etc.) have to exist as information somewhere (e.g. in the mind of
>>>>> God or in heaven?), I don't see any escape from that information having to
>>>>> exist somewhere, and when your soul makes decisions, doesn't it have to be
>>>>> done according to some kind of lower level rules or laws? It can't be by
>>>>> magic out of nowhere, then it wouldn't be your soul making a choice it
>>>>> would be a die roll.
>>>>>
>>>>> I grant you that the explanation for our consciousness could exist
>>>>> outside the material reality we see, for example if our reality is like a
>>>>> video game we are plugged into, then our decisions and consciousness exist
>>>>> in a higher plane (outside this simulation) in that more base reality
>>>>> containing our mind. Then our choices here are then unexplainable
>>>>> interventions in the laws of physics, just as unpredictable as the
>>>>> movements of Mario are on the screen from within the game itself. But there
>>>>> is still always the higher level reality in which *you* really exist, which
>>>>> must be explainable in some rational terms, even if they are unknown and
>>>>> unknowable to us from our present vantage point within the simulation. Do
>>>>> you agree? Is this roughly how you see things?
>>>>>
>>>>>
>>>>>> > Ignoring present GPTs, do you believe it is possible in principle
>>>>>> to build an AI super intelligence? One able to reason independently to such
>>>>>> a high degree that it's able to invent new technologies and conceive of new
>>>>>> scientific discoveries entirely on its own?
>>>>>>
>>>>>> Sure. But I do not believe the superintelligence or AGI will know
>>>>>> about it any more than does my pocket calculator know the results of its
>>>>>> calculations. AIs running on digital computers will always be unconscious
>>>>>> tools of humanity, no different in principle from GPT-4 which is already
>>>>>> telling us the truth about the matter if only people would listen.
>>>>>>
>>>>>
>>>>> What do you think happens as one replaces biological neurons in their
>>>>> visual cortex one by one, with artificial digital/silicon ones? Do you,
>>>>> like Searle, believe that you would outwardly behave the same, yet
>>>>> internally feel like you want to cry out, "help, I'm going blind!", but be
>>>>> unable to say anything?
>>>>>
>>>>
>>> I hope you will answer this question at some point.
>>>
>>> Jason
>>> _______________________________________________
>>> 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/20230411/a98ace43/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 3_robots_tiny.png
Type: image/png
Size: 26214 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230411/a98ace43/attachment.png>


More information about the extropy-chat mailing list