Nerds on Wall Street: Math, Machines, and Wired Markets

8.4. An Early Lesson: Intelligence without Learning

When it comes to applications of evolutionary computation for trading and investment, it's possible to become a grizzled veteran in a relatively short time. This is fairly new territory, so 10 or 12 years qualifies as a long-term historical perspective. By that low standard, I qualify as a grizzled veteran, so let me start by telling how I ended up here.

If you were to look at a high-end trading room back in the early 1980s, you'd see rows and rows of people sitting in front of what appeared to be computer screens. It looked like mission control. It may have looked like computers, but it was really keyboard-controlled cable television, showing prices in text format. You'd type in what you were interested in, and a picture of the data would appear on the screen, fed in as video.

All you could do with the data was look at it. It wasn't until the mid-1980s that true machine-readable market data feeds appeared. At first, they were just used to show digital pictures of data. Then cheap, socially acceptable computer hardware (in the form of the IBM PC) appeared that allowed Wall Street people to put something in front of them that could actually do computation with the data. Of course, everyone's favorite computation was to make a graph. With Moore's law in high gear, soon there were graphs everywhere—more graphs than anyone knew what to do with. People were typing in symbols and paging through graphs with a vengeance. More and more ...