Desktop version

Home arrow Communication

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Cognitive and machine-oriented approaches to intelligence in artificial systems

Abstract

This chapter presents the different possible routes to building an Artificial Intelligence (AI) system. On one hand, it presents the design assumptions underlying cognitive approaches to AI and, on the other, it presents the tenets of machine-oriented approaches aimed at obtaining AI systems able to exhibit intelligent behaviour without making any assumption about the biological or cognitive plausibility of the implemented mechanisms. It additionally introduces the reader to the main instances regarding the debate on the levels of analysis of computational systems (cognitively inspired or not).

Nature- vs. machine-inspired approaches to artificial systems

It is possible to draw a broad distinction between two different computational approaches for the modelling of intelligent behaviour in artificial systems. We can distinguish, indeed, between “natural/cognitive/biological” inspired systems and “machine” oriented systems. The former explicitly take inspiration (at different levels of abstraction) from natural systems, and their “heuristics”, to design an equivalent artificial system able to exhibit the same intelligent behaviour by employing, to the greatest extent possible, the same mechanisms. Machine-oriented systems do not take inspiration from how nature solves problems, but rather develop algorithms and engineering solutions by focusing on the computational challenges posed by the problem itself. The latter strategy is a perfectly reasonable and useful approach to developing Artificial Intelligence (AI) algorithms by avoiding cognitive or neural inspiration as well as claims of cognitive or neural plausibility. Indeed, this is how many researchers have proceeded in the last few decades and this approach nowadays represents the mainstream research adopted in modern AI. In their influential textbook, Russell and Nor- vig (2002) frame this kind of approach by referring to the analogy with earlier engineering artefacts: they state that “the quest for ‘artificial flight’ succeeded when the Wright brothers and others stopped imitating birds and started using wind tunnels and learning about aerodynamics” (3). This excerpt suggests that taking inspiration from nature can be, somehow, a sort of drawback with respect to the strategy of pursuing a more machine-oriented approach for building useful technologies. Recent progresses in AI seem to confirm this overall hypothesis (at least for systems working on specialized and well-defined application domains). In particular, the example of the airplane and of its different flight mode with respect to birds, has become popular in AI because it points out how - from an engineering perspective - it is important to individuate the right element to investigate (i.e., the laws of the aerodynamics) in order to find different possible technical solutions. As Simon famously said, indeed, “the flight of airplanes does not much resemble the flight of birds, except in the fact that both can sometimes stay aloft” (see Simon, 1979: 203).

For the purposes of this book, this example is also relevant since it points out that the mere “functional resemblance” in terms of generated output (i.e., the ability to fly in this case) between organisms and machines is not sufficient for explaining that function through a model capable of reproducing the essential features of that organism. In other words: a similar behaviour/output can obviously be obtained through completely different processes/mechanisms. I would say that this is an important caveat to take into account when distinguishing the cognitive-oriented approach from other machine-oriented AI approaches. To better describe this difference we need to introduce the distinction, reflected in the design phase, between functional and structural artificial models.

 
<<   CONTENTS   >>

Related topics