3.2 Nouvelle Al
We needn’t argue over what proportion of human knowledge is non-propositional. It’s enough to say that an immense mass of human knowledge and faculties, much of it central to our lives, seems to be tacit in this way. It cannot be reached by introspection. Most of the workings of the human mind are hidden from us. The majority of psychologists would agree that what you see when you look into your own mind is only a small fraction of it. The explicit, rational, symbolic part of our minds is only the thin, visible surface layer of an ocean of cognition.
The argument in this course has been this: look at nature and everywhere you will find cases of non-propositional forms of intelligence. Animals solve problems constantly, without any need for symbolic reasoning. We do too, although we are symbolic reasoners as well. An understanding of this natural intelligence might help us with some of the problems that conventional Al has failed to solve. If we can work out how goal-directed, systematic, ordered problem-solving behaviour arises in the absence of rationality and planning; if we can make sense of the mechanisms that allow both animals and ourselves to perform complex actions without conscious thought; and if we consider how these can be replicated on a computer, then this could have the following effects on Al research:
- New computational tools, techniques and approaches may become available, and these may yield new insights into the classic problems of Al.
- New fields of enquiry may be opened up. Many problems that seemed irrelevant to earlier researchers, or simply did not look open to representation on a computer - probably because they involve non-propositional knowledge – might become solvable.
Some commentators have suggested that conventional Al has concentrated only on problems that are amenable to computer solutions – games like chess are a good example of this – and then labelled the processes required to solve them as ‘intelligent’ afterwards. A classic circular argument.
The project to bring insights about the mechanisms underlying natural intelligence to difficult computational problems is called biologically inspired computing, or often nouvelle Al. Our aim is to consider what these mechanisms are – how systematic, ordered, purposeful behaviour can arise without explicit representation – and to offer detailed examples of how they can be applied in real computer systems. The discussion becomes less and less biological as it proceeds, although I will still refer to concepts taken from biology and to examples taken from the natural world.