We need to figure out a way to train Artificial Intelligence (AI) to detect and comprehend symbols if we want to advance it from being a computer that excels in a limited domain to being a true intelligence. Whether we should instruct the AI on the nature of symbols via an objective, universal method or a subjective, interpretive-based one has been a major issue of discussion in this series.

Symbolic Behavior in Artificial Intelligence

At first look, it can be interpreted to indicate that anything can be represented by any symbol, or any "series of interrelated physical patterns." That might be how Simon meant it to be. We're not sure how to assist him if he did. Of course, some patterns can only be used to represent particular concepts. It is obvious that an elephant statue depicts an elephant, not a mouse. The gadfly behind the elephant asks, "But what if you call an elephant a mouse?" Afterwards, it stands for something that is referred to as a mouse. The point is, regardless of the name you give the animal, it's pretty evident what that statue stands for. 

What more could Simon's term mean, then? An even more sensible interpretation would be that anything can be represented by any type of "interrelated physical pattern." Put more simply, there is a symbol for literally anything. A heart is used to represent the intensely emotional and incredibly complex idea of love. No matter where you go, a circle is a circle. Anybody in almost any culture in the world knows they are looking at money if they see a little disc with a person's head on it. 

What, therefore, ought to be the definition of a symbol? To put it simply, a symbol is anything that stands in for another item, be it an object, sound, action, or idea. 

Having cleared it up, let's return to the concept of an unbiased interpretation of symbols.

The authors of the paper advocate for an interpretation-based method of AI training, as we have already discussed. They do accurately point out that the culture from which symbols come contributes significantly to their meaning. Is it reasonable to argue that an objective approach is impractical or, at most, inadequate in light of this? Unthinkable? No. Apathetic? Maybe. 

But think of an opera. They are still frequently completed in German or Latin. The singing will frequently be so stylized that you might not be able to distinguish anything, even if it is in your native tongue. 

You nevertheless learn something even when you are unable to comprehend every symbol that is being offered to you. You can hear the message that the singer's voice, the melodies she presents, and the overall tone of the music all transmit to the listener. In other words, the symbols that are being conveyed have universal qualities that go beyond specific cultural contexts. 

There are several symbols that can be used for this. We are aware that a statue stands for something specific when we see one. An illustration's expression can reveal something about the character's attitude. 

This universal ingredient is recognized, therefore it would appear that we can and ought to investigate a universal foundation for instructing an AI in symbol interpretation.