{#8} Despite great dissension, there does seem to be some agreement as to what are prime criteria associated with more than trivial intelligence. These include the abilities to: generalise: the effects of training and knowledge, i.e. use available knowledge to extend an existing knowledge pool, allowing the following of general rather than specific instructions and the production and invention of strategies; learn: skills, particular facts and general facts, i.e. make use of existing information in the absence of complete instruction; remember: store and retrieve pertinent data; adapt: alter the probability of making the same response to the same or similar stimuli, i.e. adapt to changes in environmental conditions through changes in behaviour; perceptively discriminate: recognise stimuli, motives, representations and facts as being same, similar or different; model: build, describe, compare (possibly using pattern matching techniques), interpret, modify and use complex structures including symbolic structures and representations like sentences, pictures, maps and plans for action; exhibit a notion of purpose: deliberate, make decisions, plan and act on the basis of multiple purposes, motives, data (sought for and un-sought for) and external stimuli by using these symbolic structures to create a representation of environment, self, own actions, possible futures, reasons for making choices and methods of inference and relating this purpose to available resources; solve problems and be creative: particularly to have competence in a variety of domains in such a way that expertise in two or more domains can be combined, creatively and flexibly (via linking common characteristics of separate data), in order to deal with novel situations or problems; communicate: via use of a language; reason: abstractly, verbally and numerically; think: be creative and intuitively make guesses and leap to conclusions about things including those near and remote, things not previously met and new possibilities; perceive and have perceptional experience: particularly with respect to an environment, i.e. have an understanding of spatial relations, etc.; and perform experiments: both on the external environment and on an internal model of that environment. It is clearly evident that the number of activities which could indicate intelligence are numerous and almost impossible to measure. As McCorduck (1979) quite rightly suggests, a general purpose intelligence is nearly impossible to specify, for in determining intelligence we draw on an arbitrary set of abilities that are a product of the biological and cultural evolution of the human race. Other abilities for performing intelligent tasks exist, or can be conceived to exist, while new abilities might develop in the future. In short, our concept of intelligence is far too "species specific" and subjective, and too little is known about intelligence, to define it precisely. Perhaps Kugel's proposal to identify intelligence with "the ability to figure out a procedure [for doing a task, solving a problem, etc.] rather than with the ability to carry out a procedure after being told how to do it" (Kugel, 1985, p.45), may serve as a good working definition, with Pedelty's definition of machine intelligence as "the selection of an appropriate response in the absence of complete human guidance" (Pedelty, 1963, p.10), serving as a somewhat more pertinent definition with regards to our question. If we do manage to define our terms in some specific manner, the next problem to arise is that of making a judgement on the level of intelligence indicated by some act. The use of psychologists' standard Binet Intelligence Quotient (IQ) tests as a measure of intelligence is inadequate, on the grounds that by their nature the tests seem limited to the view of intelligence as what they themselves measure (Nixon, 1982). As pointed out by Sloman (1978), when asked how we can tell whether a particular program (or entity) really does exhibit some human ability (specifically the ability to think or behave intelligently), as opposed to merely mimicking it, we can once again only answer: "We can't". Because of the fact that intelligent behaviour is easy to illustrate but difficult to define, perhaps what is needed is a list of generally accepted criteria for describing what something would need to do in order to be "thinking" (similar to the one presented above), and perhaps the final judgement can only be a personal judgement based on your own ideas as to how many of, and how well, these criteria are satisfied. Furthermore Pendelty (1963) postulates that since all of our criteria are in essence humanoid functions, to be regarded as intelligent, machines should be able to perform them with no more human guidance than that required by humans. The question of what constitutes human guidance then arises as does the question of how these criteria are to be satisfied. We could say that a criterion has been met if we have reason to believe that a machine's abilities adequately model those of a human being. The problem here, stresses McCorduck (1979), is that all too often machines cannot be said to think (or satisfy a criterion) unless they show superhuman skills-a result of the modern sceptic's interpretation of the words "reason to believe" and "adequate". Even if we are to assume a machine is "doing things the same way" as a human might, Sloman (1978) further argues that this does not resolve whether the action is intelligent or not. He cites the (possibly) different ways a Chinaman and an Englishman play chess as evidence that there is not even one definitive way for humans to think in certain circumstances, and hence to expect that our human notion of intelligence is absolute and to use this notion as a measure for intelligence in other beings (be they humanoid, machine or extra-terrestrial) is quite absurd.