Chapter 2 - What does it take to possess a minimal mind?

Part A - Methodology for Investigating a Minimal Mind

Appendix: Wolfram's neo-animism

Back to Chapter 2 Back to Chapter 2 part A SUMMARY of Conclusions reached References

Steve Wolfram (2002) espouses what I would call a "neo-animist" position with regard to the occurrence of mind (or "intelligence", to use his preferred terminology). He argues that although the idea of animism - which he defines as the view "that systems with complex behavior in nature must be driven by the same kind of essential spirit as humans ... has been seen as naive and counter to progress in science", this idea is actually "crucial" to science (2002, p. 845).

Wolfram's argument can be expressed in six steps. First, if anything can be said to be the distinguishing hallmark of intelligence, it has to be complex behaviour. Wolfram explicitly equates intelligence with complexity when he writes:

Yet in Western thought there is still a strong belief that there must be something fundamentally special about us [human beings]. And nowadays the most common assumption is that it must have to do with the level of intelligence or complexity that we exhibit (2002, p. 844, italics mine).

Second, complex behaviour can be defined as the ability to perform sophisticated calculations. Hence, "intelligence is associated with the ability to do sophisticated calculations" (2002, p. 822). Here, "calculation" is meant to be a general term. It does not matter whether the calculation is performed with numbers, the black and white cells in a cellular automaton, text, images or anything else. For example, the particles in a fluid could be used to perform a calculation. In fact, "it is possible to think of any process that follows definite rules as being a computation - regardless of the elements it involves" (2002, p. 716). This implies that we can think of natural processes as computations, where the rules are defined by the laws of nature, instead of programs written by human beings (2002, p. 716). The rules can be described as mappings or functions that take a system from one state to another. In other words, "all processes, whether they are produced by human effort or occur spontaneously in nature, can be viewed as computations" (2002, p. 715).

This invites the question: how sophisticated do the calculations performed by a system have to be in order for it to be called intelligent? The third step in Wolfram's argument is contained in his Principle of Computational Equivalence (or P.C.I.), which states that there is in fact an upper limit to complex behaviour in our universe, and that anything that achieves this upper limit can be considered intelligent. This upper limit of complexity is found in universal systems. A universal system is one that can be used to perform any calculation - that is, one that can be programmed to follow any rule - so long as the function described by the rule only applies to a finite number of states (2002, pp 642 - 644, 721). (There could conceivably be systems that can exist in any one of an infinite number of different states, but Wolfram argues that there is no reason to suppose such systems actually occur in nature.) So, anything that can be considered as a universal system qualifies as intelligent.

Fourth, Wolfram's Principle of Computational Equivalence also implies that any entity possessing intelligence - i.e. any universal system - is as smart as any other: "once one has a universal system such a system can emulate any of the kinds of systems that we considered - even ones whose construction is more complicated than its own" (2002, p. 720). Some universal systems may require more time and resources to complete their calculations than others, but any of them can eventually solve any problem.

Fifth, the same principle implies that universal systems are surprisingly commonplace: "a vast range of systems - even ones with very simple underlying rules - should be equivalent [to universal systems] in sophistication of the calculations they perform" (2002, p. 822). More precisely, it says that "unless it is obviously simple essentially any behavior that one sees should correspond to a computation of equivalent sophistication" (2002, p. 726, italics mine). In other words, "any piece of complex behavior that we see ... is at some level equivalent" (2002, p. 726). Not only do some artificial non-biological systems (e.g. computers) exhibit this kind of complexity, but many kinds of natural phenomena, such as the weather, or the flow of sand in a sand pile, or the motion of a turbulent fluid (2002, p. 822), also do so.

Sixth, it follows that intelligence can be found wherever there are systems with the ability to perform complex calculations. Since such systems are commonplace in nature, it follows that intelligence is ubiquitous in the cosmos. Wolfram approves of the primitive, animist notion that the weather has a mind of its own: when we say this, "we are in effect attributing intelligence to the motion of a fluid" (2002, p. 822).

Evaluation of Wolfram's arguments

Some critics might take issue with the second and third steps in Wolfram's argument - the equation of intelligence with the ability to calculate, and his denial that there are any systems exist that are capable of occupying any one of an infinite number of different states. For example, mathematicians sometimes make intuitive generalisations that defy reduction to concrete calculations. (Wolfram's own conjecture that almost all the systems in our world are universal systems is a case in point. Is this utterance a computation?) And yet, surely these generalisations qualify as intelligent utterances when made by their originators. Wolfram's response is that intelligence has to manifest itself in a concrete, physical process in order to generate results (2002, p. 721). In other words, a purely "general" intelligence would be utterly unrecognisable. Do we have any grounds for believing that every system in existence is finite, as Wolfram believes? No, but since scientists have hitherto been able to describe phenomena in the cosmos without having to posit systems capable of occupying an infinite number of states, we can set them aside on methodological grounds (Occam's razor).

Gray (2003) also questions Wolfram's assertion that a universal Turing machine represents the upper limit of complexity: it turns out that there is something called a universal CA, that can do more, because it has infinitely many parallel processors.

My own comment on Wolfram's remarkable tour de force is that the first step in the argument - the equation of intelligence with complexity - is the most questionable, because it excludes any notion of purpose. With regard to any intelligent behaviour, it is always legitimate to ask what it is for. What is the intelligent agent trying to do? And in fact, the reason why we tend not to regard phenomena such as the wind as intelligent is that there is no discernible purpose behind them.

To his credit, Wolfram is quite explicit about excluding purpose from his definition of intelligence, on the grounds that it is too hard to discern, even when we are dealing with the behaviour of other human beings. For instance, do we have any reliable means of distinguishing the utterances of someone speaking a foreign tongue from those of someone babbling in gobbledegook (2002, p. 825)? And what about bird song? It is very complex, but no-one can be sure if it really means anything (2002, pp. 826 - 827). Again, if an alien civilisation wished to send us a message, why should they not use the wind or any other medium to encode it? We can take Wolfram's scepticism a step further and ask whether aliens themselves could be embodied in the wind.

Let us start with human language first. As Wittgenstein argued (Philosophical Investigations I. 19, 23), the meaning of linguistic utterances can only be understood with reference to their users' form of life. A tape in Sanskrit may sound like gobbledegook, but in practice, the way we learn Sanskrit, or any other language, is to see what its users do with it: greet, command, offer, request, challenge, describe, narrate and so on. In order to discern whether one of these "language games" is being played, we have to thoroughly familiarise ourselves with the way of life of the people speaking the language. "Gobbledegook" cannot be a language for the very simple reason that nobody does anything with it.

Wittgenstein's notion of a "form of life" also suggests we cannot decide whether bird song means anything until we have familiarised ourselves with how birds live in their natural environment, as ethologists have attempted to do. We shall return to this question in chapter 4.

As for alien messages, it is conceivable that they might literally be blowing in the wind, but if Wittgenstein's proposal were correct, we would have to meet the aliens first and immerse ourselves in their way of life before we could recognise their messages, let alone understand them.

But, it may be asked, what if the aliens are right under our nose: what if they, too, are blowing in the wind? The correct response to this proposal is to ask what kind of systems could embody intelligence - as opposed to merely serving as a medium for conveying a message by an intelligence? The distinction between entities that generate messages and systems that transmit them seems to be a fundamental one. The fact that such a distinction cannot be drawn within the framework of Wolfram's computational theory of mind suggests that it is an incomplete account of what minds actually do. Entities with minds, like everything else in nature, certainly undergo rule-governed transformations, but that is not all they do.