Already a Bloomberg.com user?
Sign in with the same account.
Sonia Schulenburg is trying to prove that artificial intelligence can outdo the human kind when it comes to picking stocks. Born and educated in Mexico City, she moved to Edinburgh, where she competed as a bodybuilder while earning a doctorate in artificial intelligence. Her research helped her launch a computer-driven investment fund that instantly adapts to changing market conditions. Schulenburg's firm, Level E Capital, which she runs from Edinburgh's Georgian New Town, has attracted $40 million from investors including Peter Burt, Bank of Scotland's former chief executive officer, and Baillie Gifford, which invests money for seven of the world's 15 largest pension funds. While many funds use computer formulas to trade, "this is definitely different because it's dynamic," says Michael MacPhee, a partner at Baillie Gifford, which invested $15 million last year with Schulenburg. "This thing actually adapts, it's intelligent."
Guided by a computer server that sits in a six-foot-high air-cooled cabinet, Schulenburg's Maya Fund makes as many as 1,000 trades a day, focusing on companies in the FTSE 100 and Standard & Poor's 500-stock indexes. It can both buy stocks and sell them short, betting they will fall. Unlike typical algorithm-based funds, it uses artificial intelligence to change its buy and sell criteria as it digests new market data "on the fly," says Schulenburg, 44, who has three sons. "My children say, 'Mummy makes robots, the robots live in the computer, and the computer trades stocks.' "
Because the fund can alter its trading decisions quickly, its investments tend to fall less than the overall market during a selloff. It also doesn't always capture the full gain of a market upswing. The result is a fund that delivers steady returns with low volatility, according to Schulenburg. In the 12 months since the Maya Fund began trading in April 2010, its worst month relative to the market was July, when it fell 2.4 percent and the FTSE 100 index rose 6.9 percent. The best month was May, when the fund rose 1.1 percent and the index dropped 6.6 percent. Since the start of this year, when Schulenburg fully invested the money from her clients, the fund has risen 5.3 percent, compared with the 2.9 percent advance of the FTSE 100.
Colin McLean, the CEO and founder of SVM Asset Management in Edinburgh, is weighing an investment in the Maya Fund. "It gives a reasonably consistent return on a liquid portfolio, and it will compare well with other safer investments such as bonds or cash," he says, adding that the fund hasn't been "around long enough to have been tested by different markets."
Schulenburg began thinking about applying artificial intelligence to the stock market during the 1994-95 financial crisis in Mexico when she was studying computer engineering at the Instituto Tecnológico Autónomo de México in Mexico City. After graduating, she took a job in the systems department at a brokerage that's now part of Banamex, Citigroup's (C) Mexican unit, and watched the trading floor during breaks to observe the behavior of traders. "Looking at all the panic, it needed a more systematic approach," she says. In 1995, Schulenburg moved to Scotland from Mexico to work on a doctorate at the University of Edinburgh's artificial intelligence department. Her thesis was titled "Can Learning Classifier Systems Represent Competent Traders?"
During that time, Schulenburg also competed in natural, or non-drug-enhanced, bodybuilding events. She placed third in the U.K. championships for her weight in 1998 and 2000. She also was a Scottish champion. While Schulenburg no longer competes, she says she trains sporadically.
Burt says he invested a "few thousand pounds" in the Maya Fund to allow it to start trading with real money in June 2008. "It was seed money to give her a chance, to see what she could do," Burt says. "Can you get a computer to make a better decision than a fund manager? I don't think it's yet proven, but there is an increasing amount of evidence that a computer can pick out trends that a human finds it harder to pick out."
The trading system the Maya Fund uses isn't patented because that would mean disclosing all the details, which would make it easy to copy, says Schulenburg, who aims to raise $1 billion from investors over the next five years. She believes the fund's low volatility will be a big selling point. "We don't offer that roller-coaster performance," Schulenburg says. "When people are celebrating, we don't, and when people are crying and committing suicide, we don't."
The bottom line: By programming a computer to adjust to market conditions in real time, Schulenburg aims to deliver steady returns with low volatility.