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

Ben Zaiboc ben at zaiboc.net
Sun Apr 16 09:00:06 UTC 2023


On 15/04/2023 21:48, Brent Allsop wrote:
> All of these differences you describe will be objectively observable 
> differences you will see, when you look in their brains, when they 
> hear the word Smaug.
> ...
>
> I am asking, what are the differences in their subjective knowledge?


What does the word 'subjective' mean?.

If something like a brain-state can be measured (which is necessary, in 
order to compare it with something else), then what you've measured is 
an objective fact.

Subjective states, by definition, can't be measured, and so can't be 
compared. You can't 'look into someone's brain' and know what they're 
experiencing, you can only see what patterns of information are active 
and map them to patterns that occur in other brains and infer what is 
being looking, at, heard, etc., based on what the other brains are 
looking at, hearing, etc.

What's going on inside someone's mind (as opposed to any measurable 
activity in their brain) is private to them, incommunicable to anyone 
else: SUBJECTIVE.

Even if we could measure every detail of someone's neurology when 
they're looking at a picture of Smaug, and can tell that what they're 
seeing is a cow with wings, there's no way we can know what that feels 
like to them.

Subjective states cannot be compared.

The fact that two people can agree on the name of the colour of an 
object has got nothing to do with their subjective experiences, it has 
to do with what sensory inputs they associate with a word for the 
colour. When a spaniard says "Azul", I have no idea, unless I already 
know some spanish, that it corresponds to the english word "Blue". But 
when he points to the sky and says "Azul", then points to a blue ball 
and says "Azul" and so-on, I'll soon get the idea. But if I cut open his 
brain and examine every single detail of his neurology, there's no way I 
can tell what he experiences when he looks at a blue thing (even if 
certain bits of his brain are added to mine). It is completely 
impossible for us to know what the other is experiencing. Which doesn't 
matter at all, as long as we have a common vocabulary for similar 
sensory inputs (similar because no two people are exactly the same, in 
both their sensory apparatus and in what they pay more, or less, 
attention to).

But I don't expect you'll pay any attention to this, or think about it. 
You'll just see the differences between what I'm saying and what you 
believe, and, yet again, concentrate on that, and 'correct' me. Again.


Can you see how these conversations are going?:

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: But that's not how the brain works. As far as 
we know, blah blah A, blah blah C, blah blah F, blah blah W

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: So you mean that blah blah F, blah blah M, 
blah blah D, blah blah B?

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: That's logically inconsistent. Look: Blah 
blah A, blah blah C, blah blah M, blah blah W

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: OK, let's put it another way: Blah blah Z, 
blah blah Y, blah blah K, blah blah E

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: No, I don't think so, Blah blah G, blah blah 
J, blah blah X, blah blah P

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

Just about everybody else: However, Blah blah D, blah blah A2, blah blah 
D4, blah blah Z

Brent: Blah blah Red, blah blah computationally bound, blah blah 
knowledge of, blah blah glutamate, blah blah [the same old diagram]

And so on, round and round and round and round, getting absolutely nowhere.

When you have an opinion that runs counter to all established science on 
the subject, just repeating yourself ad nauseam won't get you anywhere.

Ben




More information about the extropy-chat mailing list