Saturday, 26 January 2013

A Singular Thesis

The information revolution has brought our planet to an inflexion point. This is our generation's industrial revolution, and conventional wisdom of all sorts is suddenly in doubt. But is the universe really about to wake up? Are we about to look into the face of God? Ray Kurzweil thinks so.

Julian Jaynes rounds out his wonderful The Origins of Consciousness in the Breakdown of the Bicameral Mind with a sanguine remark that the idea of science is rooted in the same impulse that drives religion: the desire for “the Final Answer, the One Truth, the Single Cause”.

Nowhere is this impulse better illustrated, or the scientific mien so resemblant of a religious one, than in Ray Kurzweil’s hymn to forthcoming technology, The Singularity Is Near. For if ever a man were committed overtly - fervently, even - to such a unitary belief, it is Ray Kurzweil. And the sceptics among our number could hardly have asked for a better example of the pitfalls, or ironies, of such an intellectual fundamentalism: one one hand, this sort of essentialism features prominently in the currently voguish denouncements of the place of religion in contemporary affairs, often being claimed as a knock-out blow to the spiritual disposition. On the other, it is too strikingly similar in its own disposition to be anything of the sort. Ray Kurzweil is every inch the millenarian, only dressed in a lab-coat and not a habit.

Kurzweil believes that the “exponentially accelerating” “advance” of technology has us well on the way to a technological and intellectual utopia/dystopia (this sort of beauty being, though Kurzweil might deny it, decidedly in the eye of the beholder) where computer science will converge on and ultimately transcend biology and, in doing so, will transport human consciousness into something quite literally cosmic. This convergence he terms the “singularity”, a point at which he expects with startling certainty that the universe will “wake up”, and many immutable limitations of our current sorry existence (including, he seems to say, the very laws of physics) will simply fall away.

Some, your correspondent included, might wonder whether, this being the alternative, our present existence is all that sorry in the first place.

But not Raymond Kurzweil. This author seems to be genuinely excited about a prospect which sounds rather desolate, bordering on the apocalyptic, in those aspects where it manages to transcend sounding simply absurd. Which isn’t often. One thing you could not accuse Ray Kurzweil of is a lack of pluck; but there’s a fine line between bravado and foolhardiness which, in his enthusiasm, he may have crossed.

His approach to evolution is a good example. He talks frequently and modishly of the algorithmic nature of evolution, but then makes observations not quite out of the playbook, such as: “the key to an evolutionary algorithm ... is defining the problem. ... in biological evolution the overall problem has always been to survive” and “evolution increases order, which may or may not increase complexity”.

Kurzweil seems to be genuinely excited about a prospect which sounds rather desolate, bordering on the apocalyptic, wherever it manages to transcend sounding simply absurd. Which isn’t often.
But to suppose an evolutionary algorithm has “a problem it is trying to solve” - in other words, a design principle - is to emasculate its very power, namely the facility of explaining how a sophisticated phenomenon comes about *without* a design principle. Evolution works because organisms (or genes) have a capacity - not an intent - to replicate themselves. Nor, necessarily, does evolution increase order. It will tend to increase complexity, because the evolutionary algorithm, having no insight, is unable to “perceive” the structural improvements implied in a design simplification. Evolution has no way of rationalising design except by fiat. The adaptation required to replace an overly elaborate design with more effective but simpler one is, to use Richard Dawkins’ expression, an implausible step back down “Mount Improbable”. That’s generally not how evolutionary processes work: over-engineering is legion in nature; economy of design isn’t, really.

This sounds like a picky point, but it gets to the nub of Kurzweil’s outlook, which is to assume that technology evolves like biological organisms do - that a laser printer, for example, is a direct evolutionary descendent of the printing press. This, I think, is to superimpose a convenient narrative over a process that is not directly analogous: a laser printer is no more a descendent of a printing press than a mammal is a descendent of a dinosaur. Successor, perhaps; descendant, no. But the “exponential increase in progress” arguments that Kurzweil repeatedly espouses depend for their validity on this distinction.

The “evolutionary process” from woodblock printing to the Gutenberg press, to lithography, to hot metal typing, to photo-typesetting, to the ink jet printer (thanks, Wikipedia!) involves what Kurzweil would call “paradigm shifts” but which a biologist might call extinctions; each new technology arrives, supplements and (usually) obliterates the existing ones, not just by doing the same job more effectively, but - and this is critical - by opening up new vistas and possibilities altogether that weren’t even conceived of in the earlier technology - sometimes even at the cost of a certain flexibility inherent in the older technology. That is, development is constantly forking off in un-envisaged, unexpected directions. This plays havoc with Kurzweil’s loopy idea of a perfect, upwardly arcing parabola of utopian progress.

It is what I call “perspective chauvinism” to judge former technologies by the standards and parameters set by the prevailing orthodoxy - being that of the new technology. Judged by such an arbitrary standard older technologies will, by degrees, necessarily seem more and more primitive and useless. The fallacious process of judging former technologies by subsequently imposed criteria is, in my view, the source of many of Ray Kurzweil’s inevitably impressive charts of exponential progress. It isn’t that we are progressing ever more quickly onward, but the place whence we have come falls exponentially further away as our technology meanders, like a perpetually deflating balloon, through design space. Our rate of progress doesn’t change; our discarded technologies simply seem more and more irrelevant through time.

Evolutionary development is constantly forking off in unexpected directions. This plays havoc with Kurzweil’s loopy idea of a perfect, upwardly arcing parabola of utopian progress.
Kurzweil may argue that the rate of change in technology has increased, and that may be true - but I dare say a similar thing happened at the time of the agricultural revolution and again in the industrial revolution - we got from Stephenson’s rocket to the diesel locomotive within 75 years; in the subsequent 97 years the train’s evolution been somewhat more sedate. Eventually, the “S” curves Kurzweil mentions flatten out. They clearly aren’t exponential, and pretending that an exponential parabola might emerge from a conveniently concatenated series of “S” curves seems credulous to the point of disingenuity. This extrapolation into a single “parabola of best fit” has heavy resonances of the planetary “epicycle”, a famously desperate attempt of Ptolemaic astronomers to fit “misbehaving” data into what Copernicans would ultimately convince the world was a fundamentally broken model.

If this is right, then Kurzweil’s corollary assumption - that there is a technological nirvana to which we’re ever more quickly headed - commits the inverse fallacy of supposing the questions we will ask in the future - when the universe “wakes up”, as he puts it - will be exactly the ones we anticipate now. History would say this is a na├»ve, parochial, chauvinistic and false assumption.

And that, I think, is the nub of it. One feels somewhat uneasy so disdainfully pooh-poohing a theory put together with such enthusiasm and such an energetic presentation of data (and to be sure, buried in Kurzweil’s breathless prose is plenty of learning about technology which, if even half-way right, is fascinating), but that seems to be it. I suppose I am fortified by the nearby predictions made just seven years ago, seeming not to have come anything like true just yet:

“By the end of this decade [i.e., by 2010] computers will disappear as distinct physical objects, with displays built in our eyeglasses and electronics woven into our clothing”

On the other hand I could find scant reference to “cloud computing” or equivalent phenomena like the Berkeley Open Infrastructure for Network Computing project which spawned schemes like SETI@home in Kurzweil’s book. Now here is a rapidly evolving technological phenotype, for sure: hooking up thousands of serially processing computers into a massive parallel network, giving processing power way beyond any technology currently envisioned. It may be that this adaptation means we simply don’t need to incur the mental challenge of molecular transistors and so on, since there must, at some point, be an absolute limit to miniaturisation, as we approach it the marginal utility of developing the necessary technology will swan dive just as the marginal cost ascends to the heavens; whereas the parallel network involves none of those limitations. You can always hook up yet another computer, and every one will increase performance.

I suppose it’s easy to be smug as I type on my decidedly physical computer, showing no signs of being superseded with VR Goggles just yet and we’re already three years into the new decade (he also missed the mobile computing revolution, come to think of it), but the point is that the evolutionary process is notoriously bad at making predictions (until, that is, the results are in), being path-dependent as it is. 

You can’t predict for developments that haven’t yet happened. Kurzweil glosses over this shortfall at his theory’s cost. 

A version of this article was first published on Amazon in 2010.

Monday, 21 January 2013

The future's so bright I've got to wear VR goggles which help me empathise.

If your kids already spend eight hours a day online, the future depicted in Pareg and Ayesha Khanna's Hybrid Reality might ring true for you. Others may be harder to persuade.
Hybrid Reality is the short monograph which I suppose serves as flagship publication Pareg and Ayesha’s Hybrid Reality Institute, an organisation whose raison d’etre seems to be the pursuit of unfettered wishful thinking about the potential of technology. Good luck to them: dreaming up whacky visions of the future does sound like fun, and while it’s hard to see any practical application for the Fortune 500 companies the authors claim as their clients, if they’ve managed to persuade these conglomerates otherwise, happy days. Especially if in the future, everything is going to be crowd-sourced and free.

Hybrid Reality is thus an attempt to sketch out a future based on extrapolating current trends of technological development: a (thankfully slimmer) companion-piece to Ray Kurzweill’s The Singularity Is Near.

In fairness, Hybrid Reality quickly moves beyond stock platitudes about crowdsourcing, but where it does it does so without much credibility. The text is plastered with buzzwords borrowed from other disciplines and deployed with carefree abandon: 

accelerated evolution creates what we might call a Heisenbergian or quantum society: we are particles whose position, momentum and impact on others, and the impact of others on us, are perpetually uncertain due to constant technological disruptions.

Okayyy. Amongst the rhubarb there is a point to be made about rapidly disrupting technologies, but that’s not it. To the contrary, the rate of change is so fast that genuinely novel technologies and businesses have little chance to establish themselves, and that those which get a foothold do so as by fiat as sober business development, and then proceed to hammer everyone else into the ground. In such a nasty, brutish and short environment conditions favour not elegance and sophistication in design but the lowest common denominator. 

Breath-taking technologies of the sort which overflow this book, on the other hand, assume a sophistication which needs a warm and safe environment in which to incubate. Increasingly, new technologies never get the chance to be smart. It isn’t accelerated evolution that’s going on, but accelerated extinction.

In such a nasty, brutish and short environment conditions favour not elegance and sophistication in design but the lowest common denominator. 

I suppose you might expect a degree of credulity from faculty members of the “Singularity University” but, still, their vision owes as much to science fiction as it does to academic analysis and nothing at all to the traditional discipline of economics. Perhaps the dismal science, too, will succumb to the information revolution: cavalierly, Samuel Huntingdon’s maxim is reformulated so that it is not economics but technology that is “the most important source of power and wellbeing”. Older hands will recall hearing that kind of talk before, and it didn’t work out so well in 2003 when hundreds of “new economy” business models folded when it turned out they did need to generate revenue after all.

It’s easy to be a naysayer, of course, but all the same my hunch is that the Khannas’ monologue has little value for anything but excitable kite flying. Many of their assertions strongly suggest this pair really, literally, need to get out more. “Of the eight hours a day children today spend online, 1.5 involve using avatars…” they say, as if that initial premise may be taken as a given. Eight hours a day online? Which children are these, exactly? “Robots are incontestably becoming more ubiquitous, intelligent and social” and “represent an entirely new type of ‘other’ that we interact with in our social lives”. Elsewhere, “Technik”, as they put it, seems to have the power to change the laws of nature, and in the short term: “The average British citizen will likely live to be 100 years old”, they predict. Technik is so clever it can even grant us powers which we already have: In the future there will be virtual reality goggles, we are told, which can “sense other people’s stress levels”. Just imagine being able to do that.

Many of their assertions strongly suggest this pair really, literally, need to get out more. 

You can, in any case, read your fill here of all the ways the internet of things will provide an untold wealth of cool free stuff, but note the lack of any financial analysis: All this cool stuff requires effort: not just to design and conceptualise, but to manufacture, distribute, house, power, maintain and (to extent it can’t be fully computerised) operate. And effort, generally, requires money. Previous generations of technological development have shifted the labour demand curve upwards: automation has taken out repetitive, low value tasks but created more complex ones designing, building and maintaining the machinery to carry out these tasks: as a result we have grown busier with each development, not more idle - though our occupations have been more complex, challenging and rewarding. The Khannas’ brave new world would, by implication, flip that on its head.

For argument’s sake, let’s say the robots can fully take over, perform our manual labour, wipe bottoms, cure diseases and revolutionise production across all industries and agricultures so that human intervention is not required at all. Hard to see, but let’s say. Is a permanent state of situation of blissful, but chronic, total global unemployment a feasible basis for an economy?

As far as I know, man cannot live by Facebook likes alone. Last time I checked, rent wasn’t free. Nor was power, food, nor raw materials. As we go on, they’re getting harder (and costlier) to extract. So who will finance these lives of leisure? With what? Why? Who would provide services, when there was no-one to pay for them? Is it perhaps the case that personal labour, rather than being an unfortunate by-product of the “old economy” way of doing things, is in fact an immutable in the calculus of value?

Dreaming about amazing technologies which might be coming down the pike is the job of a science fiction writer. The academic question is less glamorous and more fundamental: how, within the new parameters of digital commons and in a post-growth world, can anyone devise a business model able to deliver them? These, it seems to me, are the really challenging questions, and you won’t find them addressed in this book.

Saturday, 12 January 2013

Occam’s Razorburn

Stephen Hawking’s latest book raises far more questions than it answers. Such as, why hasn’t he been reading Thomas Kuhn, and what really is the benefit of unifying theories which don't seem to need unification?

In which we meet yet another first-class scientist who wishes to self-identify as a second-class philosopher and a comedian from the back end of steerage.

Since few will buy A Grand Design for its wit we can forgive Stephen Hawking's appalling attempts to be funny, but it's not so easy to forgive his philosophical ignorance. Certain physical scientists might be better off unacquainted with the modern philosophy of science (though those who know it possess a welcome sense of perspective and humility). But not world-renowned cosmologists. Their field continually bumps up against the boundary of what science even is (and it doesn't have a "no-boundary condition", whatever that might be).

So when Stephen Hawking claims that "philosophy has not kept up with modern science, especially physics" it suggests not only a lack of perspective and humility, but that Hawking has been skipping on some required reading.  Especially since, having written off the discipline, Hawking seems barely acquainted with it. He mentions few philosophers more recent than Rene Descartes (d. 1650). So it is hard to know who he thinks hasn't kept up.

Particularly when Hawking's first grand pronouncement is "model-dependent reality": the idea that there may be alternative ways to model the same physical situation with fundamentally different elements and concepts. "If [such different models] accurately predict the same events, one cannot be said to be more real than the other." Physics has, apparently, been forced into this gambit following recent failures to get unifying calculations to work themselves out. In any case it isn't quite the neat trick Hawking thinks it is.

Firstly, while model-dependent reality might be news to Stephen Hawking (he seems to think it the fruit of modern physics' womb) the philosophers he hasn't been reading have been talking about it for years, if not centuries, to the constant sound of scientists' excoriations. It is even part of Descartes' philosophical fabric (and, more tellingly, Darwin's, but picking a fight with modern evolutionists, while fun, is a story for another day). That is to say, it sounds like it is the physicists who are finally catching up with the philosophers and not the other way around.

Secondly, in the grand game of philosophers' football that Cosmology has become, the model-dependent reality play is something of a surrender before kick-off.  For if it is true that the same phenomenon can be plausibly accounted for in multiple, "incommensurate" (© Thomas Kuhn) ways, then the hard question is not about the truth in itself of any model, but the criteria for determining which of the (potentially infinite) models available we should choose in the first place. 

This question is not one for physics, but metaphysics. It necessarily exists outside any given model (© Paul Feyerabend). Here we meet our old friend, Occam's Razor. This isn't a scientific principle at all, but a pragmatic rule of thumb with no intellectual pedigree: all else being equal, take the simplest explanation. Occam's Razor is a favourite instrument for the torture of hapless Christians by grumpy biologists: all your tricksy afterlife wagers and so on fail because evolution is so much less complicated and has so much more explanatory value than the idea that an omniscient, intangible, invisible, omnipotent entity pulling strings we can't see to make the whole thing go.

But, alas, in seeking a grand unification of things that really aren't asking to be unified, cosmology reveals some almighty snags. Unification under Hawking's programme, if it is even possible, involves slaughtering some big old sacred cows. To name a few: causality, the conventional conception of space-time; the idea that scientific theories should be based on observable data and their outcomes testable. It bows to some truly heinous false idols too. For example: seven invisible space-time dimensions, a huge mass of invisible dark matter, an arbitrary cosmological constant, a potentially infinite array of unobservable universes which wink in and out of existence courtesy of a mathematically inferred "vacuum energy"). Hawking doesn't propose solutions to these problems, but seems to think they're a fair price for achieving grand unification.

I'm not so sure: other than intellectual bragging rights, the resulting unified theory has no obvious marginal utility. And it has political drawbacks: believing one's model to be the truth carries potentially unpleasant implications for the suppression of those who don't.

There are practical drawbacks, too. We are asked to reject existing theories, which still have quite a lot of utility, in favour of something that it infinitely harder to understand and work with. The accelerating expansion of the universe without any apparent acting force seems to violate Newton's second law of motion. Without an outrageous end-run, the first nanosecond of the Big Bang (wherein the universe is obliged to expand in size by ten squillion kilometres - i.e. far faster than the speed of light) seems to violate the fundamentals of general relativity. String theory requires seven necessarily unobservable space-time dimensions and/or entirely different universes, and even then doesn't yield a single theory but millions of the blighters, all slightly inconsistent with each other (hence the appeal to "model dependent reality).

From the camp which wielded Occam's Razor so heartily against the Christians, this seems a bit rich. If these are the options, then the razor might slice in favour of the big guy with the beard.

But these aren't the options. We could save a lot of angst, and perhaps could have avoided digging trillion dollar circular tunnel under Geneva, had we employed model dependent reality the way the philosophers saw it and not the scientists (and shouldn't we call a spade a spade and label it cognitive relativism, by the way?). Since it crossed the event horizon of observability modern cosmology has become arcane, stunt-mathematics. If there were a chance that it might deliver time-travel, hyperspace or a tool for locating wormholes to other galaxies or universes then one could see the point in this intellectual onanism. But none of that seems to be allowed. So we should therefore ask the question "but why? What's the point? What progress do you promise that we can't achieve some other way?" No one seems to be able to answer that question.

But if we park it, what's left of Stephen Hawking's latest book is some pretty ropey jokes.