Modelling Human Intelligence

The goal of modelling, or replicating human intelligence is a two-pronged goal. One half of the project is the modelling of the cognitive structures of the mind. Cognitive science uses AI in part to construct models based on theories of how the mind works and then test the plausibility of these models. This can be, however, a problematic practice, mostly because of the confusing relationship between cognitive science and neurophysiology. Although cognitive science is still dedicated in many ways to a dualism of mind and body, very few if any would still claim that the mind's function is unconnected to the structure of the brain. As a result, attempts to model human intelligence seem to take two forms: formal symbolic logical systems and subconceptual neuron- based systems. These two approaches are known respectively as Good Old Fashioned AI (or GOFAI) and connectionism and will be explored in detail in the next section.

The most obvious problem of modeling human intelligence is clearly that we do not have a rigorous set of criteria for what counts as human-like intelligence. Alan Turing anticipated this (as he seems to have anticipated everything else) with the comment:

The extent to which we regard something as behaving in an intelligent manner is determined as much by our own state of mind and training as by the properties of the object under consideration. If we are able to explain and predict its behavior we have little temptation to imagine intelligence. With the same object, therefore, it is possible that one man would consider it as intelligent and another would not, the second man would have found out the rules of its behavior.

It is therefore difficult to find a rigorous definition for this goal of AI; the Turing Test is arguably the best test we have for human-like intelligence, although one ethical question that be raised in a later section is whether the Turing Test is the best test for intelligence we can use. Nevertheless, most AI research today takes as its goal the replication of human intelligence, either explicitly by trying to uncover the rules of human cognition through a hypothetical artificial model, or implicitly by using human cognitive models as the models for an expert system. There is therefore quite a bit of overlap between the two goals of AI presented here. Although they arise out of two different uses of computers, as tools for research or for industry, it has become clear that one goal cannot be achieved without also achieving the other.