Financial Markets Are Guinea Pigs for Science of Complex Systems

Know how to make a small fortune through innovative investments? Start with a large one. . . .

No doubt, the folks in Orange County don’t find that joke quite so funny these days. But all these exotic derivatives and leveraged synthetics that got Orange County in trouble are really the products of the last major wave of financial innovations. They’re the software spinoffs of financial theories that began seeping from academe into the investment world back in the early 1970s. In pure monetary terms, it’s been the most significant technology transfer of the postwar era. Trillions of dollars, yen and Deutschmarks are traded on these equations.

But now technology is creating new playgrounds for theory. Twenty years ago, chess was the key experimental medium for artificial intelligence research. Today, as academic researchers look at “neural nets” that can be trained to detect data patterns and “genetic algorithms” that literally breed software solutions to complex problems, they’re going to the most dynamic, versatile and data-rich research medium in the world: the financial markets.

Neural network software doesn’t merely execute commands, but actually constructs models of the world by establishing connections between seemingly disparate bits of data, in ways somewhat analogous to the living brain. Conversely, genetic algorithms literally evolve software according to Darwinian principles of fitness. As computer processors become more powerful and less expensive, “growing” solutions becomes practical.

And the stock markets, the bond markets, the options & futures markets and the foreign exchange markets are now becoming the E. coli and Drosophila of these areas of advanced computer research; what fruits flies and lab rats are to biology, financial markets are becoming for complex software. The markets offer a superb research domain to rigorously test pattern-recognition and solution-finding in complex systems. And, if the software works, you can make a few extra bucks along the way.

“It’s not the only medium for attack but it’s certainly a key one,” acknowledges USC Computer Science Professor Bart Kosko, who specializes in fuzzy logic research. “This will be a fast-growing area. . . . Look at the billions of dollars that have been spent on model-based techniques to analyze the markets--like charts and the Black-Scholes Options Pricing equations. By contrast, these technologies let us play with ‘model-free’ techniques--to let the data tell their own story if there is a story to be told. So I think it’s obvious that there’s going to be a lot of research in this area and that financial markets will be a major part of that research.

“In fact,” Kosko notes, “one of my students (has developed software) for prediction of gold market prices.”

“I think there is another revolution going on but, never having done one of these things, I’m a little suspicious of it,” asserts Peter Bernstein, founding editor of the Journal of Portfolio Management and author of “Capital Ideas,” a history of the rise of the financial theorists on Wall Street. “The dream of having all this stuff measurable and predictable is an irrepressible dream.” But Bernstein observes that the people developing genetic algorithms and neural nets “don’t come from economics or finance, so I’m not really on their frequency.”

Much as the Wall Street investment houses eventually lured big-name academic financial theorists to their firms to help design a new generation of financial instruments, they are now sending their “quants” to academic conferences on neural nets and genetic algorithms to cherry-pick ideas and recruit consultants. “It’s what people talk about on the floor of these conferences and between the sessions,” says Larry O’Brien, editor of AI in Finance, a magazine that was launched earlier this year. “Every trading house has their rocket scientists looking at this. It is the primary thing; it is the current Holy Grail.”

“It used to be weather; now they’re using financial markets,” observes Susan Webber of Jaquish Advisers, who has worked extensively with options traders. “Certainly, data capture has gotten a lot better. A few years ago, all that people had was end-of-day price data. Now we have prices all through the day so, if there are patterns to detect or parameters to define, there is a lot better data to work with. It’s certainly a happy coincidence that financial markets make such a good research tool,” she adds, “but, frankly, I don’t think it’s just a coincidence.”

But the point that academics stress is that financial markets offer a tremendous resource for exploring new concepts in pattern recognition and software design. “It certainly has all the right characteristics of a challenging, worthy problem,” says Stanford Consulting Computer Science Professor John Koza, author of “Genetic Programming,” a best-selling text in the field. “It certainly has the potential to be self-funding.”

But “no one knows if it is even possible to model these things or if the markets are so efficient or so chaotic that they don’t matter. . . . I’m personally not doing anything in financial services but I know several people who are.” However, Koza adds, “This may be one of those rare research areas where people prefer to publish their failed experiments rather than their successful ones.”

Will a misguided neural net investment or a mutating genetic algorithm hedge help plunge a municipality into bankruptcy 20 years hence? It’s far too early to say what the impact of these emerging software genres will be on financial investment. But recently, a computer chess program beat world chess champion Gary Kasparov in several games of speed chess. Could beating Warren Buffet be that much more difficult?

*

Michael Schrage is a writer, consultant and research associate at the Massachusetts Institute of Technology. He writes this column independently for The Times. He can be reached at schrage@latimes.com by electronic mail via the Internet.

More Innovation

* For a collection of recent Innovation columns by Michael Schrage, sign on to the TimesLink on-line service and “jump” to keyword “Innovation.”

Details on Times electronic services, B4