Thursday, August 1, 2013

Edward Feser on Kurzweil on Artificial Intelligence

Edward Feser’s review of a recent book by Ray Kurzweil can be found here:
Feser, Edward. 2013. “Kurzweil’s Phantasms,” First Things 232 (April): 51–53.
Ray Kurzweil’s book is How to Create a Mind: The Secret of Human Thought Revealed (New York, 2012).

Now Edward Feser is an Aristotelian Thomist and Classical theist, who blogs here.

Although there are other critical reviews of the book (here and here), Feser’s is considerably more sceptical, because of its apparent repudiation of a materialist/physicalist theory of mind.

A rough list of the various categories of theories of mind can be seen below:
(1) Idealism;

(2) Dualism:
(1) Platonic dualism
(2) Cartesian dualism
(3) Occasionalism
(4) Property dualism.
(3) Epiphenomenalism;

(4) Materialism/Physicalism:
(1) Reductive materialism
(2) Eliminative materialism
(3) Connectionism
(4) Functionalism
(5) “mind as chaos” theory
(6) “mind as physics” theory.
Kurzweil’s “Pattern Recognition Theory of Mind” seems related to connectionism, and is obviously a materialist/physicalist theory of mind.

First, Feser appeals to the theory of phantasms and concepts developed by Aristotle and Thomas Aquinas (Feser 2013: 51). Our minds work by abstracting from experience of reality. An individual sense experience can represent something in reality, and from “sense experiences” we abstract intellectual knowledge. The “phantasm” is a mental image of an object or phenomenon, and from these the intellect can generate “concepts” or what we conventionally think of as universals: concepts of types/classes, properties, relations (for more on this, see Kenny 1993).

Feser contends that intellectual knowledge in the sense of abstract thought involving universals is not to be identified with material processes in the brain (Feser 2013: 51). That is to say, Feser is pushing a dualist theory of mind. Feser argues that non-human animals are incapable of true abstract thinking about universals (Feser 2013: 51).

A very worrying thought for Feser must be the prospect that science will one day create artificial beings with real consciousness and intelligence, for their existence would be a grave blow to the dualist theory of mind. Whether and when science can do such a thing is open to question, but Feser’s arguments against it are not very convincing.

The dualist assertion that our highly abstract thought involving universals is not caused by, or causally dependent on, brain processes is highly dubious.

First, functional magnetic resonance imaging (fMRI) can identify the neural correlates involved in abstract thinking (Gilead et al. 2013), and, secondly, neuroscientists have already identified the crucial process of neural connection “pruning,” by which human ability in higher-level abstract thinking progresses from the poor ability of young children to the sophisticated form seen in later adolescence and adulthood (Garlick 2010: 86).

Secondly, another highly questionable assertion of Feser’s is that non-human animals are incapable of abstract thought and (at least) simple perception of universals.

Now nobody denies that animals are incapable of true language or doing science or social sciences, nor under current physicalist theories of mind should we expect them do to so.

We need only note here the crude behaviourism behind the Nim Chimpsky experiment and its failure, and the findings of other ape language studies that chimp language abilities, after intense teaching and even on the most generous estimates, are comparable to those of 2 year old children only, and never progress to that of 3 or 4 year olds.

It seems clear that only humans have the genetically-determined ability to learn and speak syntactically complex natural language spontaneously and easily. Our language ability is most likely the result of Darwinian evolutionary adaptive selection, this means that we have biological structures that cause language and that all other animals lack.

Nevertheless, the great apes, and other higher animals, do not necessarily require language in order to have at least simple abstract thinking of the type that refutes Feser’s dualist position.

Before we even get to the great apes, it looks like man’s best friend is capable of abstract thought:
Scientists have welcomed dogs into a select club of species capable of using abstract concepts. The research showed that dogs are able to mentally sort objects into categories, a talent for abstract thought that has only been shown in birds and primates before.

The researchers trained four dogs with a touch-screen test which involved choosing between two images which appeared simultaneously. One was an image of a dog, the other a landscape. The animals indicated their preference with a prod of their noses. When they chose the dog they were rewarded with food. If they chose the landscape they had to wait a few seconds before the same two images appeared again.

Next the team tested the dogs – two border collies, an Australian shepherd and a mongrel – with an unfamiliar set of dog pictures and landscapes.

To pass they would need to realise that ‘dog’ is a category of object that unfamiliar objects also fall into. Their success dropped slightly from about 80% to 72%. ‘It shows us that dogs are able to use more or less abstract concepts,’ said Dr Friederike Range, who carried out the work with colleagues at the University of Vienna.”
Randerson, James. “Dogs smarter than we think, Study Shows,” The Guardian, 6 December 2007
Now obviously nobody asserts that dogs or higher animals have the type of abstract thinking of human beings in all of its richness and sophistication, but it is a matter of degree not of kind, and this experiment appears to show dogs having some simple ability to intuit what Feser would call a “concept” as distinct from a “phantasm”.

Evolution has produced organisms with varying degrees of mental complexity and abstract thinking. Human beings are no doubt at the top of the pyramid and no other animal can match them, but it seems we are hardly alone in the general ability to think in a basic abstract manner with general concepts such as kinds/types/classes.

To return to the great apes, it seems that chimps and baboons have cognitive structures in their brains that permit some kind of abstract thinking (Estep 2006: 121; Fagot et al. 2001). Other experiments show that some varying degrees of abstract thinking exist in chimpanzees, orangutans, baboons, monkeys, dolphins, sea lions, parrots, and pigeons (Vyshedskiy 2008: 17; Katz et al. 2002). To take an example, it is suggested that the different but consistent warning sounds that vervet monkeys use when they see leopards, eagles and snakes respectively suggest a simple ability to think abstractly about types of animals, and even a rudimentary grasp of the concept of kinds/sets (a type of universal) (Vyshedskiy 2008: 18).

Of course, the critic might complain that these tests have not really measured or detected abstract thinking in animals, but that would require counterarguments and evidence, not lazy appeals to Aristotelian metaphysics or dualism.

Secondly, the critic might respond that this sort of intelligence is only seen in certain individual animals only, and not generally. But that can hardly be a strong objection. Why? The reason is that Darwinian evolution, sexual reproduction, and environmental development obviously mean that there could be important individual differences between members of the same species in terms of intelligence and even in the ability to think abstractly. Some individual dogs or chimps can be much smarter than others. Perhaps some of them really do attain a degree of ability in abstract thinking that others do not.

It follows from all this that the hard and unbridgeable gulf between (1) human beings and (2) other animals in terms of abstract thinking and use of at least simple universals like sets or types is unconvincing.

While the human mind has true language and animals do not, nevertheless higher human intelligence most probably emerged by Darwinian evolution. If higher intelligence is the product of evolution, then it is also likely that it can be engineered, even if the eventual artificial technologies from which an artificial mind would depend on causally for its existence might very different from modern computers.

Estep, Myrna. 2006. Self-Organizing Natural Intelligence: Issues of Knowing, Meaning, and Complexity. Springer, Dordrecht.

Fagot, J., Wasserman, E. A., and Young, M. E. 2001. “Discriminating the Relation between Relations: The Role of Entropy in Abstract Conceptualization by Baboons (Papio papio) and
Humans (Homo sapiens),” Journal of Experimental Psychology: Animal Behavior Processes 27: 316–328.

Garlick, Dennis. 2010. Intelligence and the Brain: Solving the Mystery of Why People Differ in IQ and How a Child Can Be a Genius. Aesop Press, Burbank CA.

Gilead, Michael, Liberman, Nira and Anat Maril. 2013. “From Mind to Matter: Neural Correlates of Abstract and Concrete Mindsets,” Social Cognitive & Affective Neuroscience

Katz, J. S., Wright, A.A., and J. Bachevalier. 2002. “Mechanisms of Same/Different Abstract-Concept Learning by Rhesus Monkeys (Macaca mulatta),” Journal of Experimental Psychology: Animal Behavior Processes 28.4: 358–368.

Kenny, Anthony John Patrick. 1993. Aquinas on Mind. Routledge, London.

Kurzweil, Ray. 2012. How to Create a Mind: The Secret of Human Thought Revealed. Viking Books, New York.

Morell, Virginia. 2013. Animal Wise: The Thoughts and Emotions of our Fellow Creatures. Crown, New York.

Potì, Patrizia. 1997. “Logical Structures of Young Chimpanzees’ Spontaneous Object Grouping,” International Journal of Primatology 18.1: 33–59.

Premack, D. and A. J. Premack. 1983. The Mind of an Ape. Norton, New York.

Randerson, James. “Dogs Smarter than We Think, Study Shows,” The Guardian, 6 December 2007

Smith, Dinitia. 1999. “A Thinking Bird or Just Another Birdbrain?,” October 9

Vyshedskiy, Andrey. 2008. On the Origin of the Human Mind: Three Theories: Uniqueness of the Human Mind, Evolution of the Human Mind, and the Neurological Basis of Conscious Experience. MobileReference, Boston, MA.

1 comment:

  1. >>While the human mind has true language and animals do not,

    Evidence shows that the communication system certain species of dolphins and whales has a complexity comparable of higher than human languages as English. Whether these species have true language, is not known, unless we would be able to decipher these communication systems.