| Previews | Ernie Jonker - Artificial Minds - Review | Calendar |
| Up to MM Magazine | On to Review Index | UpUp to MM Home |
|
Ernie Jonker
S T A N F R A N K L I N
Artificial Minds |
||
|
MIT Press (A Bradford Book), Cambridge, Mass. 1995, ISBN 0 262 06178 3, English text, 460 p., £ 22.95 |
An adventurous journey through the cognitive sciences: coffee with cake, sandwiches, and a surprise gift included in the price. During the trip, our host will give a demonstration of some remarkable products. This is more or less how Stan Franklin presents his book Artificial Minds. As a skilful chauffeur, Franklin drives his touring car past the most important sights in the land of cognitive scientists: good models, exiting controversies, broad vistas. And as would be expected, this leisurely outing is regularly interrupted by the 'message' which justifies this trip. Cognitive sciences is the name given to the reservoir of scientists who busy themselves with issues of the mind. Psychologists, neurologists, philosophers and information scientists often experiment, measure, think and tinker together in order to formulate theories on, and design models of, the mind. Only a privileged few will ever master the complete field of cognitive sciences into its most inaccessible corners. Stan Franklin does not pretend to be one of these. His own speciality is mathematics and information sciences. He calls himself an amateur philosopher, and he really does not wish to be considered an expert in any of the other disciplines discussed in his book. Perhaps this is the reason why Franklin has restricted his sight-seeing tour mainly to the more well-known attractions. Scientists such as Penrose and Dawkins have published best sellers, and names such as Minsky, McClelland, Laird, Skarda, Freeman, and many others, enjoy great popularity in the field of the cognitive sciences. For those who do not wish, or do not have time, to read the original texts, but still want to keep abreast of any developments, Franklin's book is a godsend. Franklin however, wants to do more than just show the current state of science. The message in Artificial Minds is a 'new paradigm' for the study of the mind. 'New', in so far as it is a melting pot of developments which, over the last few decades, have taken place in the cognitive sciences, and 'paradigm' in so far as it is a rough survey of doctrines which prevail in some theories about the mind, and which, in Franklin's opinion, are in fact essential. Franklin calls this paradigm the 'action selection' paradigm. It formulates the following theoretical criteria with regard to the mind: - the goal of the mind is to produce a subsequent action; - the mind is not something you either have or do not have; it is a gradual phenomenon; - the mind creates information out of sensations; - memories are not retrieved from a mental cupboard or database, but are reconstructed again and again for the benefit of the following action; - the mind does not consist of a single chunk, but is split up into independent modules which hardly exchange any information; - various systems/mechanisms are active within these modules; - the mind can be implemented into a machine.
If you replace 'mind' with 'brain activity', or, for my part, with 'intelligence', these views do not seem as controversial (any more). The more practice-oriented cognitive scientists are, in fact, inclined to pass over the whole issue of the mind. According to them, the mind does a good job in the X-files, but has no particular use in science. Franklin's book is called Artificial Minds, not Artificial Intelligence, and by referring to the mind, Franklin stumbles over a number of age-old controversies. One of these is the debate on whether or not it is possible to put the 'mind into the machine'. Can a machine, in principle, have primary experiences, such as, perception of the (mental) colour of a combination of light frequencies, meaningful thoughts, and consciousness? The discussion makes it clear that various paradigms exist side-by-side, while no crucial, choice enforcing experimental data can be submitted. Franklin favours the view that the mind can only dwell in an 'autonomous agent', an object acting with internalized motives and objectives, senses by which it can identify the current situation, and means by which it can pursue those motives and objectives. In other words: an object which can be referred to in terms of rational behaviour. The degree to which a mind can be attributed to such an agent depends on the complexity of the relationship between objectives, reality and actions. Franklin presumes that this can be observed in the agent: either by monitoring its behaviour, by opening it up or by building a model. If an agent can take in an object or an event with more than one sense, you would attribute more 'mind' to it than if it had only one sense. In the same way: if an agent can learn, can draw up a plan of action and make mental models of its world, it is more eligible of being labelled 'mentally active'. It makes no difference whether such an agent is made of flesh and blood or of silicon. For someone such as Searle, for example, this machine-with-a-mind story is useless. He wants to know how a machine, as a machine, can understand something. It is nicely convenient when a machine can process sensory information, decide what to do and then does it, but it is a question of whether we could see something if we were that machine, and whether we would grasp the meaning of it. Now, does Franklin make a valuable contribution to this discussion? The answer, for the greater part, is no, with a little bit of yes. No, because the autonomous agent, with a few capabilities give or take, does not exactly represent a new vision on the issue of the mind. Twenty years ago, philosophers such as Daniel Dennett and Paul Churchland wrote more competently and clearly on this subject. Yet the answer is also: yes. The crazy thing is, that such theoretical debates are now scarcely relevant to the practice of the cognitive sciences. The success of a cognitive scientist depends more on the technical realizability of a model, and on whether or not it works, rather than on his intuition and transcendental philosophies on the human mind. Franklin was right to focus mainly on the achievements of model builders. Albeit sporadically and with difficulty, these models are casting some doubt upon the long-established ideas on the autonomous agent. In the same way that the importance of knowledge in the form of symbolic representation has already been seriously undermined, the conventional views on perception, motives, objectives and decisions are now implicitly being called into question. Unfortunately, precisely due to Franklin's 'new paradigm', this development cannot be seen in its true light.
translation OLIVIER / WYLIE |
| Previews | Ernie Jonker - Artificial Minds - Review | Calendar |
| Up to MM Magazine | On to Review Index | UpUp to MM Home |