100 IDEAS LATER

Dick Pountain/Fri 15 November 2002/1:11 pm/Idealog 100

So, I've made it through a hundred issues without drying up. Phew... Maybe I've repeated myself occasionally, but  sufficiently disguised to render it palatable I hope. Idealog 1 in 1994 was entitled 'OOPS Upside your Head' and was a moan (that will NEVER change) about the slothful way Microsoft was introducing full object-orientation into Windows. (It also showed a familiarity with funk somewhat at odds with that glowering photograph.) Well Microsoft has finally got all its OOP ducks in a row and called it .NET. It's only taken 8 years and only requires a whole generation of punters who'd pinned their career hopes on VB programming to start again from scratch. C'est la vie or la guerre or something. Of course Microsoft still doesn't get it that OOP was meant, before all else, to make life easier for programmers: it will finally add generic templates to C# with Yukon, next year or the year after. And it's too soon tell whether or not those injured punters will revolt - that is, will Visual Basic become the Cobol of the 21st century or will C# become its Algol 68?

My other reflection on these past 8 years is that computers became sufficiently fast that it's no longer very interesting to inquire how they work. They're now fast enough for everyone but the deepest of nerds, though regrettably they have at the same time corrupted many people into precisely that kind of deep nerd. It's a great irony that the whole computer industry should stem from a single mathematical observation by Alan Turing that says that given certain minimum capabilities, anything that can be computed by one machine can be computed by any other given sufficient time. Modern PCs are several orders of magnitude faster than is required for their most popular business task, word-processing, and few users ever test their limits on the second most popular, spreadsheet calculation. Yet if you venture anywhere that nerds hang out - newsgroups, conferences, and yes, magazines - you'll find the sole topic of conversation is what kind of RAM, which cache architecture and what over-clockable chip can shave a few milliseconds off the benchmark. Computing, as I've said here before, has become an extension of the hot-rod car business.

This is merely one symptom of our pathological relationship to time in general. My favourite moment from the Simpsons is when Moe boasts to Homer about his new MilSpec deep-fat frier which can 'flash-fry a buffalo in 40 milliseconds' and Homer responds '40 milliseconds! Awww, I want it NOW!' I could get all metaphysical and spout about morbid fear of passing time due to loss of faith in god and an after-life, but I won't. Because of course, in computing the real reason for all this haste is much simpler and more mundane: games. The big lie that computers exist for enhancing business productivity still clings on by the skin of its teeth, but even hardware manufacturers will admit if pushed that games performance now drives most aspects of the development of PC technology. Those deep nerds are talking about real-time video rendering performance, not word processing.

Not that I have anything against games per se. Indeed I continue to be fascinated by Games Theory as a way of explaining many aspects of the modern world, most particularly in the sphere of economics. We've just witnessed its spectacular coming-of-age, when a group of UK academics used it to devise the rules for auctioning the Third Generation mobile phone licences, and were so successful that they've almost bankrupted half of Europe's telcos. What larks!

My love affair with the Python language continues to bloom, and one of the things I've been using it for is to program games like Iterated Prisoner's Dilemma to prove to myself that this stuff really does work. Prisoner's Dilemma is a game where two captives must choose whether to support or denounce each other, and the penalties are rigged such that they're best off if both support, worst off if both rat, but either can gain an advantage by ratting first. If you iterate this scenario so the prisoners must choose repeatedly, not just once, then there is a best long-term strategy, which is to retaliate by ratting exactly once after every time your mate rats on you. I've proved this to my own satisfaction using a Python test suite that took a short Sunday afternoon to write and occupies less than 50 lines of code.

Such mathematical games are more than just fun nowadays. Black and Scholes won the 1997 Nobel Prize for their equation used to price share derivatives, whose application lead to a massive derivatives boom that almost destroyed the world financial system in 1999 when their Long-Term Capital Management hedge fund collapsed and had to be bailled out with billions of taxpayer's money. In an important recent book ('Machine Dreams: Economics Becomes a Cyborg Science': Cambridge) Philip Mirowski suggests that the basic agent imagined by all economists since Adam Smith and before - the 'rational egoist' or homo economicus who seeks only to maximise his personal profit - doesn't actually describe the way people behave in real markets very well at all, and that as a result our economic models are deeply flawed. Mirowski proposes that markets are 'machines' in Turing's sense, whose structure governs their behaviour in ways more or less independent of assumptions about the greed (rational or otherwise) of the agents. Place your bets now as to whether there will be a PC Pro 200 in which I spell out the political implications of this discovery...