In 2011, the reigning Jeopardy! champion Ken Jennings went head-to-hard drive with Watson, an IBM supercomputer, for two rounds on the quiz show. The computer scientist lost; Watson left with $1 million in prize money.
Like all contestants, Watson had to rely on its own memory—no Googling permitted. But that memory included 200 million pages of content, including encyclopedias, newswire articles, and the full text of Wikipedia. Watson received questions in text form as Alex Trebek read them aloud, analyzed the syntax for key phrases and words, searched its servers for related content, and algorithms produced a list of answers ranked by probability percentage. Under “Final Frontiers” for $1,000, Trebek asked, “Tickets aren’t needed for this ‘event’, a black hole’s boundary from which matter can’t escape.” Watson: “Event horizon – 97%; Mass – 11%; Radiation – 10%.” And with that, IBM was $1,000 richer.
One day, in the not-so-distant future, computers may be able to do with financial markets what Watson does with trivia: scan and analyze massive amounts of data, sort it, and come up with the most probable moneymaking solution. Dashboard systems used by cutting-edge institutions like the Alberta Investment Management Corporation tell investors where their money is; big data systems will one day tell them where to put it. Just as in Jeopardy!, the key to profitable investing is not just having the right answer, it’s getting there first. Big data systems for institutional investors will likely function similar to a souped-up Google, searching billions of news articles and online texts for pieces of information relevant to trading in any asset class. Watson picks up on grammar and syntax, in addition to keywords, to comprehend Trebek’s answers-as-questions. Likewise, developers are working on applications to pick up on rumors and sentiments in online chatter and tease out patterns between these factors and changes in the markets. As artificial intelligence becomes more, well, intelligent, it could potentially detect a circulating rumor, find a couple of news articles that back it up, and inform the institutional user of a likely opportunity.
“If you’re a money manager, you want to buy a stock when its price doesn’t reflect the available information about it,” says Churchill Franklin, executive vice president and incoming chief executive officer of Acadian Asset Management. The quant-savvy firm specializes in institutional asset management, and is based in Boston. “You want to be the first one in the world to know that something’s happening, and buy that stock,” Franklin continues. “Then everyone else finds out and buys, pushing up the price. After that, you want to be the first one to sell. In the world of asset management, any small edge you can get is valuable.”
This technology—along with myriad other applications for institutional investors—is still in the prototypical stage, according to Franklin. But they are coming, and soon. “Basically, here’s the bottom line: the world of data is exploding,” he says. “Every day a vast amount is being created, via everything from the GPS on John Deere tractors to people with cell phones walking around in Africa. The pace of change is just escalating faster than anyone can grasp.” He wouldn’t venture a timeline, although a diverse group of programmers, physicists, and MBAs are hard at work on financial applications at Acadian. “Three years ago, the computer industry looked completely different from what it does today. Take iPads, for instance, or how surprising it really is that Google maps exists at all. Thinking about those leaps forward, you just begin to realize, ‘Holy smokes, there is a lot of information here.’ We’ll be surprised on the quickness of advancement.”
Already, big data applications are migrating into finance from online retail and social media, where they are already integral to the likes of Amazon and Facebook. Jim Smith, a Wells Fargo executive vice president and head of the digital channels/data and analytics group, outlined his firm’s plunge into the data deeps in a recent Wall Street Journal column. For years, he says, banks have kept track of customer phone calls and used the information to project service needs. At the moment, quantity of content is not the limiting factor—the amount of available data is plenty big. It is the organizing that presents an issue, although developers have made significant strides recently. “Big data techniques allow us to link data streams together to unveil new insights into customer experiences,” Smith says, “including understanding customer tone.” First customer tone, then hints of surprise in brokers’ Bloomberg messages mentioning a certain stock? It’s possible, says Franklin. If an asset manager’s data system also detected a mild uptick in searches for the stock, and a few large buys, institutional clients might be in for a deal.
Wells Fargo isn’t the only finance giant that’s long on big data. In March, Citigroup announced a high-profile new hire: Watson.