My review of some reading material, including Daniel Dennett's "Darwin's Dangerous Idea", from summer holidays, 1997. This is more or less unedited the post I made to the PandA mailing list.

Skip to: The Chomsky/Pinker debate - Penrose and the Emperor's Old Hat - "No chicken" - Julian Jaynes - William Safire - John McLeish

Darwin's Dangerous Idea

On our enormous double-dogleg journey from Haneda to Zurich via Kuala Lumpur and Frankfurt I read a large chunk of Daniel Dennett's latest book, "Darwin's Dangerous Idea" (details), another large chunk on the way back, and now I've finished it.

This is a big book: over 500 pages of text, plus huge bibliography. I find Dennett extremely readable, though he is not immune from criticism: Francis Crick, in a fit of school-reportese, says in the Further Reading supplement to "The Astonishing Hypothesis" (Simon & Schuster, 1994), "He has interesting ideas but appears to be overpersuaded by his own eloquence." I first encountered Dennett in his joint compendium with Douglas Hofstadter, "The Mind's I", and while the two have a lot in common, I think DD has the edge in avoiding self-indulgence. Anyway, if you want to look at some opinions differing from mine, try the "Reader Reviews" at amazon.com; I might agree with one of them that editing could have shortened the book, but for my money only by about 5%.

So what's it about? Dennett, not content with giving us "Consciousness Explained" (Little, Brown, 1991), the trigger for the quote by Francis Crick above, thinks that Darwinism is a way of thinking of almost all-embracing scope in understanding the world we live in. He coins some potent terminology, untangling sensible reductionists who think that everything in the universe follows the laws of physics, but who accept the idea of emergent phenomena (such as of course consciousness) which it is not sensible to try to explain on the level of quarks. On one side of these good people are mystics such as John Searle with their imponderable fluids (in his case Intentionality), and on the other are the "greedy" reductionists, who give reductionism a bad name by acting as though quantum mechanics were enough to "explain" everything from biology to the rules for hyphenation. Then he talks of "cranes" and "skyhooks", referring to evolution as "a process that [is] as patient as it [is] mindless, an 'automatic' and gradual lifter in Design Space" -- that makes it a crane, of course, in contrast to a skyhook, which is a piece of magic, mystic flim-flam, or theistic wand-waving.

The theme of the book of course isn't to attempt to "prove" there are no skyhooks, but rather to look critically at a wide (and golly do I mean wide) range of ideas that seem to imply skyhooks, and see how well they stand up. This gives him opportunities to get his teeth into a number of distinguished figures, including Stephen Jay Gould, Noam Chomsky, and Roger Penrose: the first two turn out to be (my assessment) crypto-mystics, and the last simply muddled. And where a mystic invokes skyhooks, a greedy reductionist tries to do without cranes: two who get this label are B. F. Skinner and Jeremy Bentham (who, incidentally, you can still visit if you're in London; he's at University College, but doesn't say much).

One of the areas where Gould often makes mysterious remarks is in the area of "adaptations" -- putting a lot of emphasis on particular cases (such as language: see below) in which, putatively, a phenomenon is not an adaptation for anything, but just sort of came along, and happened to get used. A certain emphasis on this idea of 'contingency' seems appropriate, but it is a bit hard to swallow for human language. Dennett produces a long (and somewhat arcane) discussion of spandrels, a word which I curiously learnt only a couple of years ago, in the context of the Kinkado (our circumasphaltory neighbours) logo. These have been supposed to be the non-adaptations par excellence: in other words, they don't have to evolve, but are simply a concomitant of arches which have evolved to prop up a domed roof. Given the arches, you gotta have spandrels. Dennett looks at this archetypal example and finds it wanting. Part of the point is that the spandrels in San Marco are in fact not the only possible solution, but are adapted spot on for displaying the Byzantine mosaics which are the treasure that the rich men of Venice built it for. We're recommended to look at the spandrels themselves, since it isn't very clear from the photograph. Sad to say, we arrived in Venezia in Agosto hotissimo; we looked at the queue tailing across Piazza San Marco, and gave up. So much for direct experience. (The original paper, which I haven't yet read, is Gould and Lewontin, "The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme", Proc. Royal Soc., 1979. What a wonderful title!)

Another old friend that turns up in DDI is Conway's "Game of Life". A greedy reductionist presumably would claim that I and anyone else who knows the rules (about five lines of English) therefore understand the system in its entirety. The anti-AI brigade are fighting from both ends to maintain the gap between Conway's Life and Nature's, by calling for skyhooks for nature, and using greedy reductionism to keep the cranes from Conway's building site (or perhaps just to stop people from noticing the cranes, since they're there anyway).

Dennett rounds off the book with a rather more general discussion: sociobiology, ethics, religious tolerance, and so on. A sentence in particular stands out: "The pronouncing of death sentences on those who blaspheme against a religion (complete with bounties or rewards for those who carry them out) is beyond the pale." He cites a collection of essays, "For Rushdie" (Braziller, 1994) by Arab and Muslim writers, many critical of Rushdie, but all denouncing the unspeakably immoral "fatwa".

The Chomsky / Pinker debate

I certainly find it easy to see Chomsky as something of a mystic. The central claim of Chomskyan linguistics as I understand it is that there must be a so far anatomically undiscovered piece of biology called the "language function" to account for the ability of children to learn language, because they can't possibly hear enough example sentences to work out grammar unless it's more or less there already - i.e. the innate Universal Grammar. Personally I have always found this argument quite underwhelming, and strongly reminiscent of what Dawkins calls "argument from personal incredulity", as in "I really can't see any evolutionary reason for polar bears to be white, since they have no predators to hide from." Relying on someone else's inability to see something is shaky ground to stand on.

Steven Pinker, while lumped with Chomsky above, and basically a theoretical linguist in the Chomskyan framework, strikes me as altogether more level-headed, and in particular plainly disagrees with Chomsky's view on the (non-)evolution of language. (This is lucidly discussed in Chapter 11 of Pinker's "The Language Instinct" (details).) Both Pinker and Dennett quote the same passage from Chomsky (Language and Mind, 1972):

In studying the evolution of mind, we cannot guess to what extent there are physically possible alternatives to, say, transformational generative grammar, for an organism meeting certain other physical conditions characteristic of humans.

In other words he appears to reject in principle the possibility of an adaptive analysis of why this or that feature of natural language is so. Language just floats in the air, as a given of physics, not biology. Not surprising that, as Pinker says, "Many psychologists, ... ridicule Chomsky as a crypto-creationist." The Pinker-Chomsky controversy came to a head at a meeting in 1989 where Pinker and his graduate student Paul Bloom presented a paper "Natural Language and Natural Selection". They say: "In one sense our goal is incredibly boring. All we argue is that language is no different from other complex abilities such as echolocation and stereopsis, and that the only way to explain the origin of such abilities is through the theory of natural selection." No skyhooks there. And Dennett says it was the "level of hostility and ignorance about evolution that was unabashedly expressed by eminent cognitive scientist on that occasion" that shocked him into getting down to writing this book.

To go back to Chomsky's basic Universal Grammar claims: both he and Pinker take it as absolute that the current Chomskyan framework accurately describes this human universal. Perhaps it would be asking a lot to see a Japanese translation of Pinker's book done "properly", i.e. with all the examples in Japanese not English, since many of them are plainly ad hoc. But I have here a Japanese dictionary, "Kenkyusha's dictionary of new linguistics" (Shin-gengogaku-jiten), first pub. 1971, defining in Japanese the working vocabulary of this study of a universal subject; all the examples are in English!! Makes you wonder.

Penrose and the Emperor's Old Hat

Whatever Penrose's 1989 book and its 1994 sequel "Shadows of the Mind" (details) are, they are not mystic hand-waving. Both books are full of Very Hard Bits, and I suspect there are very few people who can understand both the quantum stuff that [1994] looks to be full of and the mathematical background required to understand the point of Godel's theorem. (I'm certainly not one of them: I can't understand Schrödinger's cat thing for a start.)

Very briefly, given a formal axiomatic system satisfying certain conditions, Godel's theorem says that there are propositions that cannot be proved within that formal system, and therefore no algorithm can crunch its way through to generate a proof. But in many cases we can 'see' what the answer is. Suppose that what no algorithm can do in a particular system is prove the absence or presence of an integer satisfying certain conditions. Then we can see that there can't be such an integer, since if there were, a simple search algorithm would be bound eventually to find it. So we identify this fact as "true, but not provable within the system of axioms." Penrose's claim is that this makes us conscious beings somehow more powerful that algorithmic machines. His conclusion is not mysticism, but as Francis Crick puts it "At bottom his argument is that quantum gravity is mysterious and consciousness is mysterious and wouldn't it be wonderful if one explained the other." The hope is that somehow quantum effects in the microtubules somewhere deep inside neurons will produce otherwise impossible computations in the brain. So if not clutching at skyhooks, he is certainly hoping for a monster crane. Crick again: "It will be remarkable if his main idea turns out to be true."

Anyway Dennett's suggestion steers clear of the difficult physics, and points out that what we have isn't "an algorithm for proving impossible things", but simply an algorithm for "staying alive". Or if you like, an algorithm for convincing ourselves that things are true. Godel's result is about formal proofs within axiomatic systems; we do not violate it because we do not produce such proofs, we merely convince ourselves that this or that result is true.

One difficulty is that Godel's proof refers to the impossibility of an "Algorithm for proving X" for some proposition X, that is, (by definition), a computational procedure which is guaranteed to prove X within a finite time. This doesn't actually prevent aspects of X from popping out of some other computation, which came without such a guarantee. Suppose we have a set of real numbers satisfying some curious property P, and have shown that this set is not computable. Then there can be no algorithm which is guaranteed to produce the elements of the set - i.e. find all the numbers having this property P. There may be no algorithm even for finding a single such number, even though there is a proof they exist. (Note that this last wouldn't work if I had said "integers", or "rationals", because those are countable, so an exhaustive search program would be guaranteed to find one eventually. For the reals, which include the irrationals, and are thus uncountable, there can be no exhaustive search.) Now suppose that a curious side effect of a perfectly normal, bug-ridden program such as a particular operating system, is that it keeps generating random irrationals and testing them for property P. Notice by the way that generating random irrationals is a rather subtle business, since they can only be represented algebraically: "square root of 173", "e/pi^-7", and so forth. Anyway, this cycle-stealer has been sponsored by a hardware manufacturer to do this in the desktops of the world, and in the normal course of events, that is, when it finds no such number, to keep quiet about it. However, IF it should just chance on such a number, it shrieks about it. Hey! An algorithm has produced a fact in what we should perhaps say is a violation of Penrose's misunderstanding of Godel. This is where Sir Roger is suddenly on the defensive, because he has to rely on his argument that he, a clever mathematician (Yes! Look at the tiles.) knows that what he is doing in his head is not such a roundabout thing. He can tell that he just knows things because he can understand them. And so on. The AI assumption is that the algorithms running in Penrose's head are "a horrendously complicated unknowable and doubtless bug-ridden algorithm" which he rejects with an explicit invocation of the argument from personal incredulity:

This seems to be totally at variance with what mathematicians seem actually to be doing when they express their arguments in terms that can (at least in principle) be broken down into assertions that are 'obvious', and agreed by all. I would regard it as far-fetched in the extreme to believe that it is really [as AI suggests].

Collapse of stout party.

By the way

Did you know that you can't have a logically coherent coexistence of an Omniscient (OS) and Omnipotent (OP) being with other beings with Free Will (FW)?

Because the being with FW can beat an OS being at chicken.

The rules of chicken are: Two contestants sit in cars facing each other in a single-lane road in the desert. On a signal they start driving towards each other. Each has the opportunity to swerve off the road. If both swerve off it's a draw; if neither swerves, they both get killed, which the rules define as losing, and if only one swerves the other wins. Getting killed is defined as a particularly total form of losing, in which an immortal being for example loses everything including its immortality, whereas losing by swerving loses you only your pride and OP if you have it.

Plainly you don't swerve if you think your opponent is going to chicken out, and normally this results in a test of nerves. But things are a bit tough for an OS being, because it knows what its opponent is going to do. So our being with FW simply drives on, confident in the knowledge that the OS/OP being coming the other way either swerves and loses OP or crashes and loses everything. Logical contradiction.

What's next in the reading pile

Well, nothing to match Dennett, at least near the top. (Further down I've got Julian Jaynes' Origin of Consciousness in the Breakdown of the Bicameral Mind (review here), which I've been wanting to read for a long time.)

The other book I lugged round Europe without opening was "Coming to Terms" (details), a collection of bits by "Language Maven" William Safire, v. cheap from Daedalus. I bought it because Pinker was just so scathing about him, I thought he must be worth looking at. A few articles I dipped into from the index seemed oh-so-reasonable enough, but try reading four in a row! Dreadfully boring, and not very scintillating writing, I thought.

And next, an oddity: "Number" (details) by John McLeish($6.95 from Strand). I buy books with titles like this regularly, and they are generally disappointing, except that I've come to expect to be disappointed. This one's a rabid anti-Greek, which makes a change, but the book seems to have typos, missing tables, and other defects, and is also enticingly vague about quite who this chap is: apparently British, but "Emeritus Professor of the University of Victoria, British Columbia". Uh? Not Professor of Physics, or Economics, or 17th Century Lace-making? Anyway, he rivals John Searle in his cavalier treatment of the Chinese language. His "Chinese numerals" (i.e. kanji 1 to 10) in one table look as though they were dictated as 'shiri-moji' to a myopic Egyptian scribe, but this is not all. I quote the following paragraph in full so that you can savour the build-up to the last climactic sentence, the first half of which defies semantic analysis, and the second half of which seems to be precisely the opposite of an extremely well-known factoid about Chinese.

The second special advantage the Chinese had over most other ancient nations was the elegance and simplicity of the Chinese language. The spoken language consisted (and still consists) of one-syllable words. Complex ideas are expressed by word combinations, but each word retains its unity -- the language does not run words together, as do (for example) Eskimo, Finnish or German. There are no past, present or future tenses. There are no gender, mood, tense, case or other variables. There are no definite or indefinite articles. What we call adjectives do not need to agree with what we call nouns. [GET READY!!] There is no essential word order: the way you speak is determined by what you want to say, and you add force and expression not by word-changes but by pitch and tone of voice.

Perhaps I'll skip to the biography of Fellini.