In 1973, when mathematicians invented an algorithm that could model the financial market, investment traders refused to use it – punching a calculator’s buttons wasted time. Now algorithms control 70 percent of the US equity market, running on supercomputers that buy and sell millions of shares at nearly the speed of light. Automized trading has become a race measured in milliseconds: a new fiberoptic cable is being laid on the Atlantic seabed that will shave five milliseconds off the communication time between financial markets in London and New York. It takes a human 300 milliseconds just to blink.
At such speeds, programming mistakes are explosive. During the 2010 “Flash Crash,” an errant algorithm flooded the US stock market with US$4.1 billion worth of contracts. Within seconds, other algorithms magnified the error, distorting the entire market. Expensive stocks sold for pennies, and the Dow Jones Industrial Average plummeted 9 percent: a loss of $1 trillion. Hours later, the market inexplicably rebounded. But when the system doesn’t autocorrect, humans can only try to contain the damage. In August 2012, an investment program at the US Knight Capital Group went rogue and was caught after 45 minutes. By then the firm had lost $440 million: $10 million a minute.