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

8.3. Evolving Financial Models

For all of the aforementioned reasons, I saw the genetic and evolutionary algorithms as a great idea, one that could take the use of AI in finance to another plane, where trading and investment programs would successfully adapt to changing environments. My quant equity research group tried these algorithms in a variety of contexts, from short-term trading to longer-term forecasting.

At the start of this chapter, I recounted my adventures as I became an avid booster of genetic computation and a guest at Jumer's Bavarian house of bears and armor for the Genetic and Evolutionary Computation Conference.

Dave Goldberg's opening address to the conference included an insightful assessment of the state of evolutionary computing, in theory and in practice. Somehow, he managed to work in a story involving his Lithuanian grandmother's recipe for chicken soup, which began, "First, steal a chicken."

There was no Lithuanian chicken soup at that GECCO, but there were some amazing demonstrations of learning programs. Robot control strategies started out as random walks, and after a few hundred simulated generations, they were moving like R2D2 on a good day. There were novel circuit, network, and even protein designs produced by artificial genetic methods.^[]

The financial guys, many of whom I recognized from more Wall Street-oriented events, and I were trolling for ideas, people to hire, and software to take home. I found all three, and so, several years later, when ...