Q. I. Review

One could argue that the ability to write a book is a good confirmation of intelligence. Not, you’ll note a test of intelligence (far too many of us would fail on that basis!) but you might say that any individual capable of writing a book, stringing together a bunch of original thoughts and revealing one or two new insights into the human condition is exhibiting intelligence.

The degree, however, to which any book demonstrates intelligence varies, of course, with the complexity of the concepts discussed and the qualities of argument within the book. And, on this scale, I’m afraid, Kevin Warwick’s “QI - The Quest For Intelligence” doesn’t rate very highly. About, I would say, on a par with a “Famous Five” novel. The Quest obviously failed.

How can I justify such a sweeping condemnation? The easiest way is to give a little bit of detail about one of his central themes – the “Intelligence Hypersphere”.

The raison d’etre for this novel construct is the entirely valid observation that you can’t use the widely used measure of intelligence – the IQ test – in any meaningful discussion or analysis of intelligence. The factors measured by such a test are too few and too parochial. The weightings given to the factors are arbitrary (should spatial abilities be considered less important than literary comprehension etc etc); and the resulting single number is too one-dimensional.

This, in turn leads to, at best, a misunderstanding of what constitutes intelligence (resulting in, for instance, poorly designed teaching methods) or, at worst, illegitimate rationalisation of prejudice (justifying slavery, for example, on the basis that the enslaved races are inferior to the slave-owners). Such evils arise, Warwick argues, from our attempt to treat intelligence as a one dimensional quantity. The prejudice against afro-americans is thus justified (in the minds of the bigot) by the observation that afro-americans score significantly below white Americans on the traditional IQ test. One of Warwick’s aims is to tackle this prejudice by weaning us off the single dimensional measure.

So far so good. Reducing prejudice can’t be a bad idea and we can easily accept the need for a more rounded measure of intelligence. We can also accept that factoring in all aspects of behaviour will allow us to measure intelligence in systems where previously we hadn’t recognised any. All living things can be easily seen to exhibit some features of intelligence, as, indeed, can many machines.

This is his second purpose in inventing the hypersphere; to get us talking about intelligence in this far broader perspective so that instead of asking simple questions like “do machines think?” we can be persuaded, if you like, that “thinking” itself is merely one of the parameters of intelligence and that machines (and less complex lifeforms than our own) can still be considered to be intelligent to a greater or lesser degree even if they can’t do one or two of the things we’ve (up till now at any rate) considered a vital component of intelligence. Such as thinking.

Even here I am in full agreement with him. I’d put it a lot more simply myself.

Intelligence is the ability to make a choice (or decision).

The beauty of that simple definition is its total scalability. The very first quantum leap was the first act of intelligence. All that has happened since is a massive evolution in the intelligence of the system. That intelligence can be measured (hypothetically at least) by the number of choices the system is capable of making.

“Comprehension” and “Thinking” then arise as emergent properties from systems capable of making sufficient choices. (Beyond a certain level of complexity, choice-making requires comprehension. I recognise that this begs the question – “what is comprehension?” but let me answer that elsewhere, I’m reviewing a book, not writing one!)

The first major disagreement arises from the pointless complexity of his model. He imagines each measure of intelligence as requiring a “new dimension” and that there may be millions of relevant parameters, leading to a multidimensional hypersphere where the zero point for any ability lies at the centre of the sphere and the maximum possible ability is represented as a point on the surface of the sphere.

My objection is that there is no need to invoke multiple dimensions. Three will do nicely. Imagine an ordinary sphere. A perfectly round plain white football or basketball will do nicely. Now at some arbitrary point write “Numerical Ability” on it. Then rotate the ball, again arbitrarily, but say about 45 degrees and write “Spatial Ability”. A further 45 degrees and write “Language ability”. You get the picture. Now you measure abilities along each radius of the sphere and mark the point along the radius which corresponds to the individual’s performance for that parameter.

The next step is to recognise that between any two surface points there can be an infinite gradation of other more or less related abilities. Sticking with the limited set we’ve so far discussed, how can we move, say, from Numerical ability to Language ability? I’ll just sketch in the main steps:

Numerical abilities consist, basically, of Addition, Subtraction, Multiplication and Division. Closely related to number crunching, however, we have symbolic logic (algebra) and calculus. From here we can introduce geometry followed by a short hop to spatial awareness, then shape recognition, pattern recognition, speech recognition and finally language.

I’m not arguing that those are necessarily the correct steps or that they’re evenly spaced. Just trying to make the point that there are intermediate abilities between what we think of as the main factors; and that these could form the basis of a sophisticated measurement and modelling system.

What you end up with is a three dimensional grid of thousands or millions of points within the sphere. If you now let the “skin” of the sphere collapse onto the grid, we’ve got a model of the individual’s intelligence which is, for want of a better analogy, potato shaped! And just for the fun of it, I’m going to refer to that, from here on in, as the IP – the intelligence potato.

Now, clearly, individual IPs will be even more widely varied than traditional IQs. Warwick tries to argue that this makes it impossible to compare individuals in the simplistic way that IQs permit. It is a vacuous dream. Each IP will have one fundamental measurement that instantly reduces it to the simplicity of the IQ. It will have “volume”. And typically, we would quickly resort to referring to IPs by their volume, probably expressed as a percentage of the potential sphere we started with. The average human being, for instance might have an IP of 10 – meaning that the volume of their IP is 10% of the volume of the maximum possible IP. Human genius might have an IP of, say 15. Human Morons an IP of 7 or less. And straight away, the dream of using this more complex form of measurement to eliminate prejudice fades away. As soon as the relevant tests reveal that a target group has a lower average IP than the bigot group, the bigotry is reinforced in spades.

(I’ve always thought, frankly, that we should make use of these differences to attack the bigots with their own weapons. I’m quite sure, having met a fair few, that if we performed such wide ranging intelligence tests on the bigots themselves we could reveal to the world that, as a group, their average IP is well below the average for non bigots and this justifies our treating their assertions as the product of inferior analysis to say the least!)

So Warwick’s woolly liberalism is futile, but there are, nevertheless, some useful spinoffs from this approach. Dolphins, for instance, might turn out to have an IP very similar in size to humans, but their “potato” might occupy a radically different part of the sphere. Similarly, insects, primates and machines could probably also have a measurable IP, again in different parts of the sphere. Such a classification system would enable us to do exactly what Warwick tries to argue we can never do – compare apples with oranges. The IPs might have drastically different shapes and volumes, but it is still pretty obvious that an individual with a high IP is more intelligent than an individual with a low IP.

In fact it’s the most useful thing it could do. At a stroke it gives us a much better way, for instance to compare the intellectual abilities of chimpanzees with children, or ants with bees. Or bees with robots etc etc. As soon as we break away from limiting the discussion of intelligence to “that which makes us human” we can clearly see how all life is a continuum, that there are no real arbitrary break points or barriers beyond which the individual is intelligent and below which it is just a dumb animal or dumb terminal.

So much, then, for his foray into what constitutes intelligence and how best we might measure it. Why does he want to bother with this in the first place? Because, and this, ostensibly, is the real purpose of the book, if we don’t understand what intelligence is, then we won’t necessarily recognise it when we create it. We won’t see what’s coming and we won’t know what has hit us when our creations destroy us.

If you imagine that I’m overstating his position, here are a few specific quotes from the early pages:

"if we allow the robots of the future to be more intelligent than us it is they, not us, who will dominate the earth"

"How can we possibly stop this happening if we cannot agree what intelligence is in the first place? We must, I feel, fully investigate this subject in order to ensure our own survival."

"When this machine intelligence is working for humans there is no problem. But surely machines which can think for themselves and are intelligent in their own right, will eventually cause us problems. Certainly, if they are far more intelligent than us, will it still be possible for us to call the shots? If not, perhaps we are facing the prospect of a world run by machines rather than humans."

That bold line in the last example illustrates what you might think is a petty point. “Certainly” it begins. We are now expecting an assertion. Instead we get a question. Its not just bad, lazy or absent editing, its an example of what I call “PPD” – Post Prandial Dictation. The professor (for such he is, Professor of Cybernetics, no less, at Reading University) had clearly been offered an advance by the publishers to follow up his “best selling” “March of the Machines”. So he scheduled half an hour a day, after lunch, for his secretary to take notes, virtually verbatim, which became the book. It would seem that he had one too many good lunches.

In any case, you have to wonder why a professor of cybernetics, of all people, should be promoting this paranoid nonsense. And not just any old professor of cybernetics, but one who is rightly renowned for his direct participation in some leading edge experimentation. Warwick is the one who had the arm implant which allowed the University computer system to recognise him which, in turn, enabled it to open doors for him, prepare his computer screen for him and perform a number of other useful little functions.

Why should someone who is ostensibly leading the revolution suddenly be crying “er… hey guys I don’t like where we’re going” whilst, at the same time, actively seeking more funding for continuing in precisely the same direction. You’d think, if he really were that worried, the rational course would be to seek to shut down his own department and then campaign to shut down all similar research across the planet.

Call me a cynic, but I smell self-publicist here. Dr Susan Greenfield (who, in my view, has a much larger IP than friend Warwick) has, through her accessible writings on the workings of the human brain, become the (UK) media’s default spokesperson for all matters related to the brain. Methinks perhaps Warwick is angling to become the default for matters to do with the advancing intelligence of human technology. Let’s hope the media IP is large enough to see through that.

Back to the paranoia. What evidence does Warwick offer in support of this potential doomsday scenario? Zilch. What reasoned argument does he present to convince us that intelligent machines would, without question, subjugate us or even destroy us? None at all. From page one to the bitter end, the charge is merely an assertion and painfully similar to the kind of assertions bigots tend to make about the targets of their bigotry. “They’re out to get us. They’ll take our jobs. They’ll eat our babies”

A more thoughtful analysis would accept that there are indeed risks arising from the increased intelligence of machines. Intelligent cruise missiles, capable of identifying (most of the time) a precise target in the middle of a city were an early example of how such intelligence could eventually become a threat to us all. If we choose to create “terminators” then we’ll all be in deep doodoo!

The crucial phrase in that last sentence is “if we choose”. There is little doubt that the technology could be made to go that way. What Warwick and other technophobes are postulating, however, is that technology will inevitably go that way even if we do not choose it. In other words, Warwick sees as inevitable that machines with sufficient intellect would, of their own volition, decide that it would be a good idea to do away with the human race. Thus, he implies (he never actually states it), that we’d better do something soon to prevent machines ever acquiring such intelligence.

Historical precedent, it must be said, is on his side. As probably the most intelligent species (to date) on this planet, we have, so far, swept all before us and wiped out many species in the process. But why? Simply because they either threatened our survival directly (eg carnivores) or competed for resources to which we wanted exclusive access (eg herbivores eating our grain). On occasion we’ve wiped them out (or nearly) as a result of pure greed rather than necessity (whale and elephant hunting come to mind). And sometimes superstition is responsible (Tiger’s and Rhinos, ferinstance, have been driven to the edge of extinction just so humans can use tiny parts of the animals to boost their flagging sex lives. Which wouldn’t be quite so bad if it had ever worked, dammit, but there is no evidence whatsoever that it does work any better than any other placebo.)

But how would any of this behaviour carry across to machine intelligence? We are not going to eat them. We don’t compete for resources in any direct sense. Indeed they will be responsible, in large part, for reducing the overall consumption of almost all finite resources. They will never need to be greedy and they will never be blessed with the crass ignorance which fuels superstition.

If ever machines evolve an entirely artificial intelligence it would be markedly different from our own in at least one fundamental respect. It would be detached and objective to a degree that it will never be possible for an organic lifeform to emulate. Indeed one of the really important stumbling blocks on the path to artificial intelligence lies in trying to conceive what on earth would motivate such an intelligence. Yes it could, perhaps, answer any question we throw at it, and, in the process, solve any problem we have. But would it bother (metaphorically speaking) getting up in the morning? Would it do anything at all, other than merely respond to questions or carry out any instructions we passed to it?

Why did you get up this morning? Because you had to go to work to earn an honest crust in order to feed the family, pay for the roof over your head etc etc. Why did you eat lunch? Because your brain told you that you needed food. Why did you have sex last night? Because you had a wonderfully primitive urge to pleasure yourself – a biological bribe designed to make your dna replicate itself.

None of these “drives” will exist for a purely machine based intelligence. We may choose to simulate them, but that is all they would be, simulations, not the real thing. And without such drives, it is not only difficult to imagine what would motivate them, it is impossible to imagine that they would be driven to perform the kind of horrors that Warwick vaguely hints at. In a nutshell, they’ll be entirely disinterested. They won’t give a damn.

In practice, though, I don’t see us going down that road in any case. The visible trend at the moment (and one which Warwick is a prominent example of) is for the melding of humans with machines. What we’ll see increasingly over the coming years is how machines are used to enhance human beings, not to subjugate them (although I admit the possibility that amongst such enhanced humans there may well be a proportion still sufficiently primitive to exploit their new powers to subjugate “ordinary” humans). The emerging new species is not a machine based intelligence at all, it will be a hybrid of human and machine. And if that species chooses to blast us into oblivion, it will, I’m afraid, be a human responsibility and no fault of the machines.

This is not to deny the option of genuinely artificial intelligence emerging alongside this new species and even constituting the first artificial species in its own right. But the intellectual - or IP - gap between the new human/machine hybrid species and this hypothetical machine-only species will be a lot less than the gap between present day humans and such a machine species. It may even be that the gap is in our (descendants’) favour. There may well be constraints on the intelligence of either organic or machine based lifeforms which only a hybrid can overcome – which would leave our descendants still as the top dogs in Earth’s evolutionary tree. (The real fun will be when we encounter our first off-world hybrid intelligence, particularly if it’s a few million years ahead of us.)

Until, or unless, however, someone can come up with a rational hypothesis to explain why a species with superior intelligence to our own would behave as badly as we’ve been known to, then Kevin Warwick’s vicious cyborgs will only frighten small children and other individuals with an under-developed IP.

In conclusion, if you bought the book in pursuit of the “Quest for Intelligence” you’ll be disappointed and no wiser after the event. If you’re a technophobe, it will reinforce your fears and prejudices and you’ll no doubt start quoting it as a biblical revelation of the dangers awaiting us. If you’re a technophile you’ll yawn and go looking for a more serious discussion.

QI – The Quest For Intelligence