Custom Search

The Role of Artificial InteUigence

There is one potentially significant exception to everything I've said above. Computer software and hardware technology are now advancing at such a rapid pace that the development of artificial-intelligence-based trading systems will, I feel, in the very near future, become a significant force in the marketplace. I refer here specifically to systems based on neural networks. Until recently, the computer hardware and software required to mimic the functioning of the human brain were extremely costly and, in many cases unavailable other than to government and large industry. However, with the advent of affordable computer memoВэry as well as faster computer chip speeds, the ability to analyze vast amounts of data on many levels simultaneously has opened the door to many new and powerful possibilities in the field of artificial intelliВэgence. Artificial intelligence is the terminology used to describe machine-based or computer-simulated intelligence which attempts to simulate the human problem-solving process. Applications of this new technology are already being used in science, business, and industry, particularly in the area of industrial robotics.

A sophisticated branch of artificial intelligence, termed neural netВэworks, actually models its software on theories of human thinking, attempting thereby to replicate the human learning process and moreВэover to create a computer program which learns from its errors and which performs better with more experience. Although both these goals are certainly lofty, they are not unrealistic in terms of current and anticiВэpated technology. Because neural networks seek to mimic the intellectuВэal processes of the human brain, this new technology promises to spawn a vast new area of trading methodology, one which is based on computВэer-learning models. If the promise of the neural networks in the area of stock and futures trading turns out to be as significant as I think it will, then this new technology could very well change system development dramatically.

Because each neural analysis program is distinct in terms of the marВэket forces it examines and the learning model it employs, very few neurВэal programs will reach the same conclusions at the same time. Hence, neural analysis also promises a multiplicity of approaches which, in the final analysis, would be considered an extremely positive factor for the markets, since the players would not all be playing from the same side at the same time or from similar timing frames. Because of how neural analysis reaches its conclusions, the focus is primarily on pattern recogВэnition and evaluation of current relationships which appear to be drivВэing the markets. Since forces underlying market trends change from day to day and week to week, neural analysis programs which are sensitive to such changes will also change. Neural systems are dynamic; they are not subject to the static limitations of traditional trading systems. This, of course, promises to eliminate one of the major objections to traditionВэal trading systems, which is that they are not responsive to underlying market conditions which can change dramatically over time.

Finally, neural analysis programs evaluate market inputs as do tradiВэtional trading systems. However, neural analysis programs can evaluate a vast number of inputs (by inputs I mean indicators), analyzing the combination of indicators which are driving market prices at any given time. Neural programs can combine the evaluation of technical as well as fundamental inputs concomitantly. In this way, neural analysis proВэgrams represent a quantum leap, possibly a several-quanta leap, from traditional systems. You'll be hearing much more about these new sysВэtems in the next few years.