How to Create a Mind: The Secret of Human Thought Revealed
by Ray Kurzweil
Viking, 352 pages, $27.95
What goes into making a human mind? Two key elements, distinguished by Aristotle and Aquinas, are phantasms and concepts—a distinction entirely overlooked by pop-science writer Ray Kurzweil in his How to Create a Mind. This is no small error in a book purporting to show that science has not only explained human thought but will soon enable us to duplicate it in a machine.
A phantasm is a representation of a sensible object, the sort of thing you know introspectively as, say, the mental image of a triangle or a dog’s bark. It can have a degree of generality, as when you visualize a triangle without imagining any specific size or call to mind a generic barking sound. But it will also always have features that not every member of the class possesses. An image of a triangle will be equilateral, scalene, or isosceles but not all three, and the dog’s bark you bring to mind will be high-pitched or deep but not both.
A concept, by contrast, is truly universal. Your concept triangle applies to every triangle without exception, whether equilateral, scalene, or isosceles, acute, obtuse, or right, whether red, green, or blue, large or small. Your concept bark applies to the barks of toy poodles, German shepherds, and Dobermans alike. That they have a universality that phantasms can never have is one reason scholastic thinkers like Aquinas would argue that concepts cannot be identified with phantasms or images.
Another is that concepts can have an unambiguous, determinate, or exact content that phantasms cannot have. Nothing inherent in the phantasm of a black isosceles triangle determines that it represents triangles in general, or black triangles, or isosceles triangles, or black isosceles triangles—or even triangles at all, as opposed to a dunce cap or an arrowhead. The concepts isosceles triangle and dunce cap, by contrast, unambiguously represent isosceles triangles and dunce caps.
This distinction has a great many philosophical implications. For example, it provides the basis for the conclusion that while sensation and imagination are identifiable with material processes in the brain, genuinely intellectual activity or abstract thought is not. The first lack the strict universality and determinacy of content the second have.
The distinction also helps explain the difference between human beings and other animals. Nonhuman animals can form general images and thus exhibit behaviors (such as ape or dolphin mimicry) that superficially resemble our higher cognitive functions, but they cannot form true concepts or exhibit the strictly intellectual or abstractive capacities—for scientific and philosophical theorizing, for example—that presuppose them.
Any serious account of the human mind must distinguish phantasms and concepts. Empiricist philosophers like Berkeley and Hume notoriously conflated them, often treating concepts as if they were nothing more than mental images. (The bizarre conclusions for which these thinkers are famous ultimately rest on this crude philosophical error. Consider Berkeley’s view that no material substances underlie the colors and shapes we perceive and Hume’s claim that we have no concept of an abiding self underlying our thoughts and perceptions. All we can imagine, after all, are the colors, shapes, thoughts, and perceptions themselves and not any underlying substance.)
Given that even prominent philosophers are guilty of confusing concepts and phantasms, the conflation is, unsurprisingly, an occupational hazard of works of pop philosophy—which brings us back to Kurzweil’s book. Kurzweil, an influential inventor and author of bestsellers on technology, futurism, and artificial intelligence (AI), was recently named director of engineering at Google.
He is well known for making bold and sometimes eccentric claims—in The Singularity Is Near he predicted a new civilization in which the boundaries between human and machine, reality and virtual reality, will be blurred. The views expressed in his latest book, How to Create a Mind, are for the most part more mainstream, even if expressed with less caution and nuance than some other writers on AI might recommend.
Kurzweil puts forward a variation of the thesis that the brain is a computer and the mind the program run on the computer. The idea that thinking is really nothing more than mechanical computation has become commonplace, bolstered by the materialism that prevails in the modern academy and made familiar within the wider culture by the science-fiction themes of popular entertainment. (Think of HAL in 2001: A Space Odyssey, the Star Trek character Data, or Steven Spielberg’s movie AI.)
Assume materialism, factor in the extraordinary practical blessings computers have afforded us”what can’t they do?”and the notion that the mind is just a kind of computer might seem plausible. All it takes to seal the deal is some conceptual sleight of hand. Enter Kurzweil.
The core of Kurzweil’s position is what he calls the “pattern recognition theory of mind” (PRTM). The basic idea is not new—it has antecedents in the “connectionist” or “neural network” approach to AI and ultimately in the associationist psychology of the empiricists—but Kurzweil provides an agreeably clear and concise exposition. The key notion of the PRTM is that of a “pattern recognition module,” a cluster of neurons that fire in response to a stimulus.
To take a highly simplified example, one module might be triggered by a vertical line and another by an oblong loop, which in turn together trigger a further, higher-level module that thereby recognizes the pattern P . Yet further modules recognize the patterns E , A , and R ; and a third-level module recognizes the pattern PEAR . In Kurzweil’s view, this sort of mechanism explains perception, language, and mental capacities more generally; and to implement such a mechanism in a computer would be to “create a mind.”
His critics have pointed out that existing AI systems that implement such pattern-recognition mechanisms in fact succeed only within narrow boundaries. A deeper problem, though, is that nothing in these mechanisms goes beyond the formation of phantasms or images. And while a phantasm can have a certain degree of generality, as Kurzweil’s pattern-recognizers do, they lack the true universality and unambiguous content characteristic of concepts and definitive of genuine thought.
Recall the image of a black isosceles triangle, which does not of itself represent either all triangles or indeed any triangles at all. Adding the word triangle to the image wouldn’t help. Is the image with the word meant to refer to triangles in general or some particular triangle, or to the word triangle rather than to triangles themselves, or to something else entirely? Nothing in any image itself will tell you, no matter how many elements you add to the image or how many other images it is associated with. Any system of images is inherently ambiguous because the images can be interpreted in so many different ways.
Kurzweil’s pattern-recognition modules face the same problem. Nothing in the modules or system of modules themselves, no matter how complex, can provide any universal reference or determinate, unambiguous content”that is to say, a content of the sort that true concepts can have. Does the module corresponding to PEAR represent pears themselves, or the word pear, or a set of shapes that merely looks like the word “pear,” or something else entirely?
You might think that the question can be answered by appealing to the module’s cause-and-effect relations to other modules and to external stimuli, but that only kicks the problem up to ever higher levels. Any system of material processes is always susceptible of alternative interpretations.
It is not only old-fashioned followers of Aristotle and Aquinas who would raise this sort of objection. The impossibility of entirely capturing the conceptual content of human thought in terms of images, material processes, and the like has been emphasized by philosophers as diverse as Ludwig Wittgenstein, W. V. O. Quine, Saul Kripke, and Daniel Dennett. (Where they differ is that the more materialistically inclined among them would bite the bullet and conclude that our thoughts must not have any determinate conceptual content after all—a position of dubious coherence at best.)
Even if Kurzweil’s model did not have this philosophical defect or face the practical problems identified by other critics, might it still succeed in capturing the sub-conceptual (and thus sub-intellectual) activities of nonhuman animals, which can have general phantasms or images (as opposed to truly universal concepts)? Even this concedes too much to Kurzweil.
Suppose a botanist were to model the various aspects of plant life—root-growth patterns, sap circulation, photosynthesis, and the like—in a computer program. He could hardly be said to have created a plant, any more than if he had made a painting or sculpture of a plant. What was created would merely be a simulation having a greater or lesser degree of resemblance to an actual plant.
There is no reason to think that the computer model of the mind Kurzweil describes would be any different—a mere simulation of a mind, but not the real thing. (Of course, this sort of objection has been raised by critics of AI before, most famously by philosopher John Searle. In response, Kurzweil offers nothing more than some brief restatements of criticisms that Searle answered at length more than twenty years ago.)
A more accurate title for Kurzweil’s book would have been something like How to (Partially) Simulate a (Subhuman) Mind. But that rather lacks the punch of the original, and whatever the substantive deficiencies in his ideas, Kurzweil—an effective entrepreneur when he isn’t writing pop philosophy—definitely knows how to market them.
Edward Feser is associate professor of philosophy at Pasadena City College.