Over the last several years, many of the traditional tools and economic assumptions used for understanding the financial system have been slowly replaced by more powerful tools originally developed by physicists, biologists, or even engineers.
Dr. Hamid Benbrahim, a data scientist and expert in finance, machine learning, and robotics, explains how the financial system is a complex adaptive system and that scientists are “making good progress” in applying the newest and most successful tool in finance—artificial intelligence.
Here are a few excerpts from his interview (click here for audio) that recently aired to our subscribers.
FSN: Dr. Hamid Benbrahim, you recently gave a presentation at the Sante Fe Institute where you discuss the financial system as a complex adaptive system. Can you explain what that means?
Benbrahim: These are systems where you have a large number of participants, like people for instance, that interact with each other. They are semi-autonomous, so they have some free will, but they are also guided by some kinds of guidelines, rules, or incentives to make them do what they do. And when you have these large number of interactions...all kinds of interesting things emerge. There are mechanisms of self-organization, like you can notice in herds for instance—herding behavior. All people herd together when they all cross the street at the same time or they all buy the same stock or dump the same stock and so forth. And we see these kinds of interactions in many other systems. In biology for instance, when you look at ant colonies or termite mounds. All of those are interactions of complex adaptive systems...you see birds flocking and making all sorts of interesting patterns—those also are characteristics of complex adaptive systems. And the reason why this is interesting in the stock market is exactly that: you have a large number of people trading and computers trading many times a second.
FSN: Given that the financial system exhibits many of these same properties we see with complex biological or physical systems in nature, what sort of tools must we use to understand the system better?
Benbrahim: When we start thinking about the stock market or financial system from a system point-of-view...then we are incorporating many pieces that affect the dynamics of the system. And those pieces include social behaviors or social events, particularly economic events, and so forth. So all of these things are parts of the system and they affect the behavior of the markets. And also the interactions within the system itself—the interactions between traders or investors also affect the dynamics of the system. To do that, then we need to apply the best tools that are available to us. And many of them...can be found in physics...social sciences...and psychology.
FSN: So a variety of tools from different disciplines are now necessary to understand the markets. Also, whenever you speak of the financial system as a complex adaptive system, there’s obviously an emphasis on this notion of adaptability, which means static models obviously don’t work very well. Given your expertise in finance and machine learning, how successful do you see A.I. in working to solve this problem?
Benbrahim: That's a great question. Machine learning has incredible promise in addressing many of these problems. The reason I got into machine learning to start with was because I was thinking about complex systems and how do we control [them]... Machine learning basically starts with a premise of saying let me have the system figure out on its own what it is the right thing to do. So you could say if your goal is to minimize volatility in your portfolio, then an intelligent system in theory should be able to observe the behavior of the market and should be able to learn from experience as it starts making adjustments and, with every new adjustment, it looks at the outcome and determines whether it was a positive outcome or negative outcome, and then learns from that to come up with the best configuration possible. So I strongly believe that machine learning is the right tool for understanding this level of complexity. And...today we have tremendous amounts of computing power and we have access to lots of data so I'm very hopeful that we're getting to a point where we can apply machine learning to address many of these problems. We're barely scratching the surface but we're making good progress...
To hear this full interview with Dr. Hamid Benbrahim and to gain unrestricted access to all our premium interviews and content, please login or CLICK HERE to subscribe. In addition to hearing from money managers, institutional investors, and other experts regularly interviewed for FS Insiders, you’ll also gain access to our entire archive of past interviews with noted heavy-weights like Kyle Bass, David Rosenberg, Gary Shilling, Marc Faber, and more.
Audio Link | Other Expert Interviews | iTunes | YouTube | Comments or Inquiries
