March of the algorithms raises issues

By Nicole Bullock. This article originally appeared on the Financial Times website, on September 16th, 2014

Smart, self-teaching algorithms are taking over from people in homes, cars and hospitals. Are they going to do the same to the capital markets?

To an extent the machines have already won. High-frequency trading (HFT) has accounted for the majority of trading on US markets for several years. But there is a qualitative difference between statistical modelling software that is primed to pick up on market cues to obtain an edge over rivals, and genuine “machine learning” algorithms that teach themselves tactics.

So-called “black-box” artificial intelligence techniques have found favour in recent years in ecommerce and recommendation engines. While they seem to work, the precise mechanism by which they arrive at answers is often opaque to human observers. This could pose a problem for regulators, experts say, as they need to be able to pinpoint how a decision was made and who made it to apportion legal responsibility.

“The question is, are you happy to have your systems trade on something you don’t understand?” says Donald MacKenzie, a sociologist at the University of Edinburgh who studies financial markets. “To prove market manipulation requires demonstration of intent.”

Martin Wheatley, chief executive of the UK’s Financial Conduct Authority recently acknowledged that this world of self-improving artificial intelligence is “perhaps closer than we think”.

Indeed, profit-constrained HFT, hedge funds and asset managers are eyeing “smart” approaches to gain a competitive advantage. German financial technology group GFT, which works with big global investment banks, has partnered with Massive Analytic, a big data start-up, to develop trading software based on “artificial precognition”.

“We are looking to build algorithms that tell you when the next black swan event will take place,” says George Frangou, Massive Analytic’s chief executive. “It’s about understanding not just what’s likely to happen, but what could happen.”

That approach rests on huge amounts of data, and global regulators are preparing to oversee this brave new world.

The US Securities and Exchange Commission have launched MIDAS, a big data visualisation tool that collates a billion of time-stamped daily records. Europe is planning tighter regulation of the algorithms used in trading.

Global spending on such solutions is set to rise. The money dished out on parsing unstructured data for electronic trading is estimated to grow by 12 per cent over 2014 to reach $228m, according to a study by the Aite group.

Meanwhile, only half the 22 capital markets firms that responded to a survey by Aite Group said they had a so-called “big data” strategy in place.

Much of the reluctance to talk is simply competitive caginess, but an even more fundamental problem, especially for banks, is that they cannot use smart algorithms until they have visibility across all their trades and exposures.

About New York Institute of Finance

With a history dating back more than 90 years, the New York Institute of Finance is a global leader in training for the financial services and related industries with course topics covering investment banking, securities, retirement income planning, insurance, mutual funds, financial planning, finance and accounting, and lending.  The New York Institute of Finance has a faculty of industry leaders and offers a range of program delivery options including self-study, online and in classroom.

For more information on the New York Institute of Finance, visit the homepage or view in-person and online finance courses below: