|
Posthumans and Other Hedge Fund Managers
Hedge fund management naturally attracts highly competitive men and women. Not surprisingly, hedge fund marketing is just as competitive. Perhaps, that is why so many hedge fund managers have tried for so long to present their approach as posthuman. As the marketing manager of one such firm once said to me: “Our competition has no hope of keeping up with us. We use artificial intelligence. We don’t employ people, at least not to do anything important.” As one-upmanship goes, this is hard to top.
Actually, I don’t remember whether he said artificial intelligence or genetic algorithms or neural nets or data mining—maybe he just said he used a computer. I have listened to a lot of similar brags over the decades, and watched most of the braggarts crash and burn. I expect the same for most of the current batch. This is not because I’m an old fogey. Well, I am an old fogey—when I started in this business, punch cards were still a novelty—but that’s not why I see problems here. Nor do I lack understanding or sympathy. I don’t believe in souls or spirits or angels or gods. As far as I can tell, we are computers made of neurons. And, in turn, I see no reason why a computer could not be made to think better than we do.
But that doesn’t mean that anyone owns a posthuman right now. I find it hard to believe that George Bush would have led us into the Iraqi debacle if there were a weakly godlike machine sitting somewhere in a subbasement of the White House or the Pentagon. Considering that there are no noncontroversial estimates of human IQ, it is hardly surprising that there are few estimates of machine intelligence. My own highly unscientific estimate is that the best we can do right now is to make an AI as smart as a cricket or, perhaps, a cat. In terms of patience, tenacity and sheer ruthlessness, I would put my own cat, Snowball, up against any three hedge fund managers you might choose. Still, I find it hard to believe that anyone in the hedge fund industry would be bothered by being on the other side of Snowball’s trades, if Snowball did trade, that is.
This has at least two implications for the fund-of-funds manager. First, and most obvious, a hedge fund manager who claims he just follows the orders of some kind of posthuman might be a genuine visionary, but much more likely he is a fool or thinks you are a fool. Second, a hedge fund manager who makes more modest claims, who uses the tools or engines that are used to build posthumans (i.e., genetic algorithms, data mining and the like) is certainly worth talking to. Given the current state of the art in the investment industry, he may have an important competitive advantage. Unfortunately, much more likely than not, he will piss that advantage away.
Illusions of the mind
There are illusions of the mind that are every bit as real as optical illusions are and much, much more important. In the same way that knowing you are looking at an optical illusion does not make the illusion go away, knowing you are a victim of an illusion of the mind does not make that illusion go away either. For example, in a game involving luck and skill, most people overestimate the amount of skill involved. Speculators need to be familiar with these illusions. It takes an act of will to recognize the illusions and do the right thing.
Lists of these illusions can be found in books on cognitive psychology, in (ahem) my own books and in David Aronson’s, Evidence-Based Technical Analysis. (By the way, Aronson’s analysis applies to all investment methods, not just technical methods.) Aronson criticizes popular forms of technical analysis on the basis of their vulnerability to these illusions and on the lack of evidence that these tools work. Further, he proposes solving these problems by the application of statistics and data mining.
Data mining allows us to test massive numbers of investment methods and find the “best” ones. The reader will notice a similarity with optimization. Data mining is a more powerful technique. Aronson then presents a number of techniques for deterring if the methods found are statistically valid. (Disclosure: David Aronson is a friend of mine and a former client.)
Data Mining.
Aronson’s criticisms here are exactly on target. Moreover, data mining has a proven (noninvestment) track record for finding new facts. For example, retailers use data mining to find products that their customers would like to buy. Amazon knows that people who have bought my book have also bought the books of Ralph Vince so they make those books available to potential buyers. Even better, because techniques such as data mining do not have our biases, they could well discover investment facts and relationships that we could not find on our own—no matter how much time we had.
Unfortunately, Aronson’s stress on investment methods, which is entirely typical of the investment industry, is exactly wrong. An investment method is a way of exploiting an investment fact or relationship—the discovery and validation of facts and relationships must come first. More important, this is radically more difficult than most people in the industry believe. It takes observation and thought and repeated experimentation. The technical term for this process is science. I don’t know another science that claims to be able to find anything important based on a single sophisticated look at the data.
As one sage put it, there is no limit to how far most people will go to avoid thinking. Apparently, this includes investors, and this includes building machines to do the thinking for them.
(Aronson’s book has a number of virtues I haven’t mentioned. It makes several difficult topics accessible; it presents a number of novel techniques for managing data mining. And most important, despite a large number of statements that should be hedged, it clearly presents an honest man trying to think his way through a difficult issue.)
Evidence-Based Technical Analysis, David Aronson, Wiley Finance, 2007, $ 95.00
|