[ExI] Meaningless Symbols

Ben Zaiboc bbenzai at yahoo.com
Sun Jan 10 13:31:13 UTC 2010


> Damien Broderick <thespike at satx.rr.com> wrote:
> 
> On 1/9/2010 5:41 PM, Stathis Papaioannou wrote:
> > More to the point, what is the difference between real
> understanding
> > and pseudo-understanding? If I can use a word
> appropriately in every
> > context, then ipso facto I understand that word.
> 
> Not relevant. What you can do is exactly beside the point
> when 
> discussing what robot systems can do. A good Google
> translation now can 
> cough up a reliable translation from Hungarian (I know, I
> used it the 
> other day to turn part of one of my human-translated papers
> back into 
> English). It would be perverse to claim that the Google
> system 
> understood the words being translated, even though the
> complex program 
> was able to find appropriate English words and syntax. I
> understood it, 
> the machine didn't.

Google isn't a robot.  What a human can do *is* relevant to what a robot can do, because they both not only have a brain, but also a body. The word "move" is meaningless to Google because it has no experience of moving, so all it can do is relate it to another word in another language.  

To a system that does have the means of movement (whether that be via a real-world body or in a simulated environment), it has an experience of what moving is like.  That's the 'symbol grounding' it needs to make sense of the word.  It now has a meaning.  

"Using the word appropriately in every context" means that if you say "Could you move 2 metres to your left?" the system will be able to answer yes or no, and do it or not, depending on it's physical state and environment.  Moving 2 metres to the left is meaningless to Google, because Google doesn't have legs (or wheels, etc.).

If you hooked Google up to a robotic (or virtual) body, and gave it the means to sense the environment, and move the body, and hooked up words to actions, then it would be capable of understanding (assigning meaning to) the words, because they would now have a context.

Ben Zaiboc


      



More information about the extropy-chat mailing list