The nature of intelligence - Artificial Intelligence

It seems that everyone has their own opinion on what intelligence is or isn’t. Intelligence is a concept that everyone knows about, but understands differently. As we have seen in the previous chapter, the way each person understands a particular concept will have its own unique ‘flavour’. Perhaps one of the most interesting quotes above is by Susan Sontag that uses an analogy between taste and intelligence. Taste is a complex sensation in four dimensions – sweetness, sourness, bitterness and saltiness. Similarly, intelligence is a complex concept, with multiple dimensions.

Intelligence is multi-faceted – its nature cannot be defined using one of these quotes alone; it requires all of them. As an analogy, try describing the Mona Lisa. One person’s description of the painting may be anathema to another person. To imagine that we can distil the Mona Lisa down to a few written words, and then naïvely believe other people will agree with us that it is the one and only definitive description, is like believing that people should only ever eat one type of food, or enjoy looking at one type of painting, or read one type of book. The Mona Lisa painting continues to inspire people to write more and more words about it. Similarly, intelligence is not something we can elucidate definitively. But that will not stop people from continuing to do so, since in so doing further insights can be gained into its nature.

Although definitions of intelligence are fraught with problems, we can look for desirable properties of intelligence that we can help us to describe the nature of intelligence. In other words, we can help define the nature of intelligence by describing what it ‘looks’ like or what it ‘tastes’ like. Using the taste analogy, we can think of these properties as being ‘ingredients’ in a recipe for intelligence–we need to mix them together in order to make a particular taste, which some people will like,while others may not, preferring alternative tastes. For example, we can use the analogy of African and Australian explorers trying to describe what a giraffe or platypus looks like to someone who has never seen it.

These explorers will use words(concepts) that they are familiar with, such as ‘long neck’ and ‘fish-like tail’, but their description will be ‘flavoured’ by their own unique perspective. Whatever words they come up with, they will have over-emphasized certain features and ignored other important ingredients.

Similarly, AI researchers with a background in knowledge engineering and the symbolic approach to AI will describe intelligence using ingredients such as the following:

  • the capacity to acquire and apply knowledge;
  • the ability to perform reasoning; and
  • the ability to make decisions and plan in order to achieve a specific goal.

AI researchers who prefer a behavioural-based approach will describe the intelligent behaviour of embodied, situated agents using ingredients such as:

  • the ability to perform an action that an external intelligent agent would deem to be intelligent;
  • the ability to demonstrate knowledge of the consequences of its actions; and
  • the ability to demonstrate knowledge of how to influence or change its environment in order to affect outcomes and achieve its goals.

If we think of intelligence using an analogy of mapping, as discussed in the previous chapter, then we might use the following ingredients to describe intelligence:

  • the ability of an embodied, situated agent to map environments, both real and abstract (i.e. recognize patterns to provide useful simplifications and / or characterizations of its environments);
  • the ability to use maps to navigate around its environments;
  • the ability to update its maps when it finds they do not fit reality; and
  • the ability to communicate details of its maps to other agents.

It is important to realise, however, that these are not definitive descriptions,just ingredients in alternative recipes for intelligence.

In the previous chapters, we have seen various examples(implemented as models in NetLogo) that have demonstrated some of these ingredients. In some respects, these models have exhibited a small degree of intelligence in the sense that if we observed a human agent with the same behaviour, we would deem that to be a sign of intelligence. In the next volume of this book series, we will also see other models that will demonstrate more advanced technologies. It can be argued, however, that these examples show no true intelligence–but of course that depends on your own perspective, and the ingredients with which you choose for your own recipe for intelligence.


All rights reserved © 2018 Wisdom IT Services India Pvt. Ltd DMCA.com Protection Status

Artificial Intelligence Topics