[ExI] The symbol grounding problem in strong AI

[Stathis Papaioannou]

2009/12/17 Gordon Swobe <gts_2000 at yahoo.com>:
> --- On Tue, 12/15/09, Stathis Papaioannou <stathisp at gmail.com> wrote:
>
>> http://users.ecs.soton.ac.uk/harnad/Papers/Py104/searle.comp.html
> ...
>
>> So, Searle allows that the behaviour of a neuron could be
>> copied by a computer program, but that this artificial neuron
>> would lack the essential ingredient for consciousness. This claim
>> can be refuted with a purely analytic argument, valid independently
>> of any empirical fact about the brain. The argument consists in
>> considering what you would experience if part of your brain were
>> replaced with artificial neurons that are functionally equivalent
>> but (for the purpose of the reductio) lacking in the the essential
>> ingredient of consciousness.
>
> Glad to see you read that article.
>
> I don't understand why you say you refuted anything with a purely analytic argument that does not depend on any empirical fact, when your argument consists of imagining an empirical fact! But that's besides the point...

The form of the argument is such that it is true if the premises are
true: that is, IF it is possible to simulate the behaviour of a neuron
with a computer program THEN it is also possible to simulate
consciousness.

Return to your simplified brain X-X-0-0-0-0, where X are the
artificial neurons in the visual cortex and 0 are the biological
neurons in the association, language and motor cortex. The X neurons'
job is to behave in such a way that the 0 neurons can't tell that they
aren't 0 neurons. According to Searle, this masquerade should be
possible. As a result, the subject with the cyborgised brain will tell
me correctly how many fingers I am holding up, declare that everything
looks normal and that he feels just the same as he did before the
operation. This is what *must* happen. It's true in all possible
words, true such that even an omnipotent God couldn't make it not
true. Please explain if you disagree!

Now, it is logically possible that although the subject will behave
exactly the same as if no change to his brain had been made, his
consciousness would be different. That is, he might be blind and not
notice that he is blind, or he might notice that he is blind smiles
and says everything is just fine while attempting in vain to
communicate his terror. The first possibility would make the notion of
consciousness meaningless, for if nothing else, we understand that
having a perception means that we realise that we have the perception.
The second possibility would mean that the subject is thinking without
his brain, since his brain is constrained to behave normally. Both
these scenarios seem quite implausible, if logically possible. Much
easier to simply say that the subject would be normally conscious.

> It looks like you want to refute Searle's claim that although a computer simulation of a brain is possible, such a simulation will not have intentionality/semantics. It won't on Searle's view have any more semantics than does a computer simulation of anything have anything. A simulation is, umm, a simulation.

While a simulation of a thunderstorm is not wet, a simulation of a
brain is conscious. That's the difference between brains and
thunderstorms.

> I once wrote a gaming application in C++ that contained an imaginary character. Because the character interacted in complex ways with the human player in spoken language (it used voice recognition) I found it handy to create an object called "brain" in my code to represent the character's thought processes. Had I had the knowledge and the time, I could have created a complete computer simulation of a real brain.
>
> Assume I had done so. Did my character have understanding of the words it manipulated? Did the program itself have such understanding? In other words, did either the character or the program overcome the symbol grounding problem?
>
> No and No and No. I merely created a computer simulation in which an imaginary character with an imaginary brain pretended to overcome the symbol grounding problem. I did nothing more interesting than does a cartoonist who writes cartoons for your local newspaper.

A complex enough game character probably would be conscious. There are
gradations of consciousness: bacterium, ant, lizard, mouse, dog,
human, superhuman AI.


-- 
Stathis Papaioannou