My brother Mike's got a discussion
on symbol grounding over in his journal which he'd like feedback on.
For the uninitiated Symbol Grounding is the process by which we link the ideas in our heads to the outside world - linking "ham sandwich" to some actual object outside of ourselves. There's a growing movement in AI that says that the problem with many approaches to AI is that they don't even attempt to link their internal 'facts' with external 'experience' and just manipulate words in an attempt to sound conscious.
Anyway, my thoughts on ideas and their relation to the world go something like this:
Symbols connect to each other through their definitions. You end up with chains of definitions from unknown to known. For example you can work your way through the dictionary using those definitions - "A dog is a domesticated animal descended from wolves." "A wolf is a large predatory animal that hunts in packs." "A pack is a group of animals that runs together." And so on ad infinitum.
Now, the problem there is the ad infinitum bit. Dictionaries are limited in size, so eventually your definitions are going to lead back to themselves. Not only that, but unless you have an internal understanding of the meaning of a word then you have to go and look that up too.
This leaves us two options:
1) The definitions are eventually grounded, in which case they leave the realm of the defined and enter the realm of experience, which is undefinable in its own realm - defintions require symbols which are pointers to experience, they aren't part of the experience itself
2) The definitions eventually loop round and what you have is a circular reference, which is no good to anyone. A circular reference that doesn't refer to anything you know would look something like this - "Freebles are a groshnik kind of wifflebop." "Wifflebops are a frequent sight when weshlurfling your spikquoddle." "Spikquoddles are a common byproduct of less freebles." Without any idea what the words might point to, you have no way of getting into the loop and attaining any understanding.
So, either your definitions are looped and not based in reality, or they link to experience which is inherently subjective. In order to communicate you have to hope that your experiences are similar enough to the person you are communicating with that you have a degree of intersubjectivity (a shared approach to the experience) and thus can find ways to explain things to them.
One of the things that I find useful about the Myers-Briggs psychological test is that it indicates something about the way the people view the world and their approach to it. The idea that other people might not jus have different conclusions, but be working from distinctly different approaches to and experiences of the same situation is one that came to me very late in life. Previous to that I had assumed that any differences between people came from either incomplete information or from not thinking clearly enough about it. Realising that other people may in fact be experiencing a different world to me and that therefore I would never
be able to come to an agreement with them about the world caused a radical change in my thinking about communication. It meant that I lost a lot of my interest in arguing about the nature of reality and the way the world works unless I believe that the person I am talking to thinks about the world in a sufficiently similar way that we will eventually be able to reach an understanding.