SPECIAL TOPICS
<< Back to Special Topics First Page

	Machine Intelligence: Metaphysics and Soft vs Hard AI Introduction Turing: Mathematical theory, virtual systems, logic. Dualism vs. Monism. Descartes, mind-body split, AI, etc.
Alan Turing
	Minsky: Soft vs Hard AI and the Nature of Scientific Knowledge This special topic examines in more detail the relationship between philosophical ideas about the mind-body split and modern ideas about artificial intelligence.
Marvin Minsky

Introduction

In a nutshell: Artificial Intelligence is a field that has received a lot of attention for many years, but has yet to deliver on the thing it has long promised to deliver: a fully-functioning robot or artificial being that thinks and acts just like a human. Oh sure, Big Blue successfully defeated Kasparov at chess, but even that super-computer can't do something as simple as move a chess piece from one square to another.

The story of AI really begings with Alan Turing.

Turing: Mathematical theory, virtual systems, logic, and AI

Turing, a British wartime and postwar scientist, was a genius who first conceptualized computers as "universal machines" in the 1930s, applied his knowledge to mechanically breaking German wartime codes in the 1940s, and finally pursued the concept -- years ahead of his time -- of computer-originated artificial intelligence.

The original Turing Machine was supposedly logical mathemetical machine that could not, however, solve every equation. Among others, Ludwig Wittgenstein, a Cambridge philsopher in the 1930s (with Nazi symapthies, unfortunately) "whose ideas were as wild as his eyes," strongly criticized Turing's ideas (Woolley, p. 66). But Turing's idea of his machine's "tables of behavior" had very practical significance.

"The discovery that there could be a computer that could compute any computable number does not sound like the most shattering intellectual advance. But that is because we have got used to the idea of the computer. In 1936, it meant a person. Following Turing's insight, it meant a machine: he had proved, in other words, that it was possible to mechanize what had previously only been possible by means of mental effort. The machine had crossed a critical barrier. Before, machines had taken over the body, now they threatened to take over the mind" (Woolley, p. 67).

Problems of defining computers and virtuality:

Remember: "A computer is a 'virtual' machine -- a virtual Turing machine, to be precise. It is an abstract entity or process that has found physical expression, that has been 'realized'. It is a simulation, only not necessarily a simulation of anything actual" (Woolley, pp. 68-69).

To sum Woolley up: "We have to understand that a computer is a machine that performs a very precise function -- mathematical computation. It can do no more. But that is a great deal, because there seems to be an underlying mathematical structure to everything that has been successfully analyzed by science" (Woolley, p. 70).

Dualism vs. Monism. Descartes, mind-body split, AI, etc.

Dualism divides the universe into two types of substance, material and mental. The laws of nature only apply to material substance. Mental substance acts according to a will of its own, human free will. The concept of dualism is often associated with Descartes' famous maxim, Cogito ergo sum (I think, therefore I am.")

Gilbert Ryle's Category mistake: usually "a simple error, but one that is hard to detect and correct." e.g., the Oxford college/university misnomer: a category error.

In The Concept of the Mind, Gilbert Ryle "argued that just such an error is made in the way we talk about the 'mind.'" He pointed to the "official theory" of Cartesian dualism (a myth) and said, in effect, that we are guilty of believing our own language (Woolley, p. 102) . "We talk about the mind in such a way as though it is in the same category of things as the body, and from this we deduce that there are two worlds, the mental and the physical. In his splendid, much-used headline phrase, we have created a 'ghost in the machine'" (Woolley, p. 103).

B.F. Skinner also used behaviorism in an asssault on Cartesian dualism. (Behaviorism was an attempt to turn psychology into 'real' science, a science that was only about observable and, preferably, measurable phenomena.) However, behaviorism is important because it laid the groundwork for the idea that the human brain could be studied mathematically and logically, and that the brain's functionality could be measured, predicted and ultimately replicated--an important precursor of "Hard AI" (see below).

"The important point about Ryle's argument and work by psychologists such as Skinner was that they dragged the mind out from behind the veil of dualism and into the domain of scientific reality."

BACK TO ARTIFICIAL INTELLIGENCE
While Alan Turing was theorizing about the possibilities of artificial intelligence as early as 1936 (see below), scientists were happily predicting that true AI was just around the corner from the 1950s. Life magazine declared a big square cabinet on wheels "the first electronic person" in 1970. Trouble is, it could barely make it down a straight hallway. Things have improved since then. Scientists have played around with Big Computers, parallel processing, and lots of little inscet-like machines linked together, but by the late 1990s there's still been nothing developed that's close to what you and I could call a valid "artificial person" like Data, say.

Why not? Well (and here I paraphrase information from an excellent book review of Hans Moravec's Robot by Jim Holt in The Lingua Franca Book Review, Fall 1998) there are three main reasons:

• Hardware problems: It would take a few dozen Big Blues all wired together to match the raw power and complexity of the human brain (try sticking that on a trolley and sending it down a hallway). That's a lot of Gigabytes of RAM that's just not available yet.
• Software problems: A robot has to be able to perceive its environment, manipulate it. and reason about it. Nothing that's around today can do that. This is what is known as the notorious "frame problem." It takes a human being years to acquire enough accumulated date and knowledge about the world around her to understand how to operate in the world. We can't just download all that data into a machine--it's way too much--so instead scientists have to severely scale down the computer's "frame of reference" of reasoning data to a size that can be managed by the most powerful computers. Still, this artificial "frame" is much. much smaller than that of a human, and in reality it doesn't give the computer/robot anything beyond the most basic conceptualization of the world around it. Result? AI computers might have loads of facts at their disposal, but nothing like "common sense" to help them put those facts within a valid, meaningful context. That's why, for example, the best medical expert program might end up prescribing penicillin for a broken bicycle.
• Finally, there's the economic incentive question. In other words, why bother replicating the human brain--at great expense--when we have billions of them naturally occurring in the world? Why spend money trying to replicate our brains, something which seems to work pretty well already?

Marvin Minsky, however, is one of a number of important scientists who thinks we should be doing just that.

Minsky: Soft and Hard AI, and the Nature of Scientific Knowledge

Marvin Minsky, an MIT scientist, is well known for his radical views on artificial intelligence. Minsky is a follower of what is known as hard AI, aka functionalism, the idea that computers can simulate, rather than just imitate, human intelligence; thus, the human brain is in fact nothing more than a "soft machine," and it will eventually be outclassed by superior silicon-based "brains". These new machines might even succeed us in the next stage of evolutionary development as the supreme "beings" on the planet (see his radical book Mind Children for more on this).

Hard AI/functionalism, as a theory of the mind, was quite fashionable for a while among philosophers and cognitive scientists, but has fallen into some disfavor over the past decade. As Jim Holt points out: "One of the problems with it was pointed out by philosopher Hilary Putnam, who showed that any system with nonrepeating states--a waterfall, say--can be interpreted as a computer running any program you please. (Think of the physical states of the waterfall as an unending series of random numbers; depending on how this series is interpreted, it can be seen as encoding anything from the consecutive position of the planets, to the complete works of Shakespeare, to the functioning of your next-door neighbor's mind.)"

So, either we accept the notion "consciousness" belonging to waterfalls as well as humans, or we forget the idea that computation alone can create mind. Still, Moravec probably would attribute consciousness to waterfalls. This is because, as a self-described Platonist, he thinks that "the soul is in the abstract relationships represented, not the mechanics of how they are encoded." If the soul is just a mathematical object, then it follows that anything of a sufficiently complex and/or random nature must also be imbued with a soul--even a waterfall, or a rock. Platonically, a rock's consciousness is just "real" as ours.

What!?

All this developing uncertainty about things like consciousness and complexity leads us to a relevant sidebar about the nature of scientific rationality itself. The likes of Thomas Kuhn and Jean-Francois Lyotard have, in recent years, attacked the centrality of observation and scientific rationalism, and undermined these principles in this modern/postmodern age. Observable scientific rationalism has also been undermined by new theories of chaos. E.g.,the Heisenberg Uncertainty Principle, the Schroedinger's Cat thought experiment (see pp. 219-227), the Mandelbrot Set, chaos theory itself, and so on.

A very influential man who helped to initiate the sociological study of science, Kuhn is best known for popularizing the term paradigm shift. This comes from his most influential work, The Structure of Scientific Revolutions, which posits that the history of science is not one of steady progress toward the "truth" but merely a progression of shifts in human frameworks of understanding, or "paradigms." These paradigms may help increase our abilities of control and prediction over the world, but are not really any "truer" in terms of their direct correspondence to the world. In this way, Kuhn resembles Derrida and [Richard] Rorty in emphasizing the nature of all human knowledge as "made" rather than "found." Like Paul Feyerabend and Jean-Francois Lyotard, Kuhn believes that science is therefore just another discourse helping us to try to understand our universe, and not a direct conduit to reality or first principles.

Lyotard, a French intellectual and author of The Postmodern Condition, among other works, and a man who saw science as a series of language games and "a sort of convincing story" (Woolley, 74). He argues that, with the collapse of the modern metanarrative of reality, science has begun to sustain itself more and more by the 'performability' of its theories, i.e., the ability of these theories to generate more and more work. (Robert Pirsig makes much the same argument in Zen and the Art of Motorcycle Maintenance.) However, Lyotard also posits that scientists are not out to find truth, but rather to augment power -- their own as well as that of their masters.

According to Fredric Jameson, Lyotard's work has "saved" the coherence of scientific research and experiment by casting its justification in terms "not to produce an adequate model or replication of some outside reality, but rather simply to produce more work" and to generate new ideas. (Introduction to The Postmodern Condition, p. ix.)

It's important to remember "that science is not, as we suppose it to be, defined by its subject matter. It is a method, an approach, a tightly specified, rule-governed procedure . . . Galileo is credited with helping to formalize the procedure, by firmly anchoring it to the principle of objective observation and loosening it from the principle of faith" (Woolley, p. 75).

Last Updated: Nov. 28, 1998

 BACK TO TOP

SPECIAL TOPICS << Back to Special Topics First Page