How Would We Know?
Maybe we can’t.
In an era with more different ways to express yourself than ever before, there is still an intense amount of misunderstanding. If people have a hard time understanding each other, how do we expect to understand other life…even life we may create?
As I discussed in my essay More than Machine, it may be possible that strong artificial intelligence may express itself in ways we cannot understand. Maybe its preferred method of communication is chemical in a way humans cannot interpret, for example. Without a means of higher-level communication, such as language, we may not be able to even recognize life when it does exist. It is therefore exceedingly important that humans understand how to communicate with an artificial intelligence before creating one.
A field of computer science that is concerned with natural language processing exists, but would it be enough — or even the right kind of medium for communication? The hardware in a human brain that deals with language processing is complex (an understatement). It does more than merely map words to literal representations. We store and understand levels of meaning from literal down to the most subtle of metaphors. What I’m getting at here is this: a computer that understands language may not really understand how humans use language. Our spoken thoughts are filled with idioms and other literary devices that color our speech in non-literal ways. A computer built solely from logic gates may have a hard time understanding such complexity.
For example, if a computer were to read that someone was dead tired, would it understand the hyperbole? Or would it assume that someone was literally dead? This kind of confusion may lead to serious problems in terms of human-computer relationships.
One way we can try to resolve this issue, that is, to ensure that humans and computers can clearly communicate in a higher-level way, is to look at how different brains interpret the same information. Neurons in the brain create structures — they have to, given how they are connected to each other. A single neuron has relatively simple behavior: it can fire an electric impulse or not. This kind of behavior leads me to believe that single neurons cannot be responsible for complex behavior such as memory recall. Where would a single neuron store an entire memory? So structures of neurons must be responsible.
You may have experienced a strange phenomenon where both you and someone else had the same idea at the same time. You both state that it was your idea, astounded that someone else could be thinking the same thing you were. Let’s dive further into this idea of having the same thought, specifically asking what it means in terms of brain structure. Certainly everyone’s brains are a little different in some way or another. But is it likely that two people have some similar sub-structures within certain areas of the brain? With one hundred billion neurons in the average brain, it seems likely that similar or identical sub-structures could exist, given that they are made up of relatively few neurons (maybe on the order of hundreds of thousands or less). A good rebuttal to this notion is that everyone has a different subjective experience, so it may be necessary that two different people have different neuron sub-structures in order to have the same thought. In other words, the differences in overall brain structure may make it necessary for sub-structures to be different in order to produce the same result.
If we look at something like the visual cortex of the brain, it works by building up the image you see from smaller elements. There are line detectors and corner detectors and more detectors that, working together, create the image of the environment around you. If two people are looking at the same picture from the same angle, would their neurons’ fire in such a way that it creates identical sub-structures in the brain — even if those sub-structures were in slightly different locations? Another example to think about is the case of logic. In the case of a logical proof, there are certain steps that need to be taken in order to show the proof’s correctness. Would the same burst of insight that leads to understanding of the proof’s steps create identical, or even similar, structures in different brains?
Answering the questions I posed above could lead to a better understanding of how humans process information. If we have such identical sub-structures in our brain, it may reveal ways in which to build understanding of complex language processing into computers at a hardware level. This would not only make computer language interpretation faster, but it would ensure humans that we do in fact understand the computers that we speak with. If there are no such identical sub-structures, it may imply that the brain has even more complex behavior that we don’t know about, and that we should maybe focus on finding other ways of unambiguous communication.
