Have We Finally Become Slaves to Machines?

Late April, an obscure out-of-print paperback about the lifecycle of a fly became the single most expensive book on the Internet. In one day, its price skyrocketed from $32.95 to $23,698,655.93. It wasn’t because two billionaires were competing in a farcical bidding war; instead, a basic Amazon pricing algorithm went haywire. Programmed to raise the price incrementally if another seller raised his, the code went about its business in an endless feedback loop. At some point, a human stepped in to cool these boys out.

Gamer (and algorithm wiz-kid) Kevin Slavin says it’s evidence of programs in conflict without any adult supervision. In his riveting TEDGlobal talk delivered last month, Slavin shows how code like Amazon’s determine espionage tactics, stock prices, movie scripts, and ultimately, the physical landscape. And he warns that we no longer understand the code, nor its implications.

Exhibit A: 70 percent of the American stock market is now running on autopilot in the form of high-speed programs designed to hide, and then find, millions of shares in milliseconds. This code enables enormous trades to happen in the blink of an eye, but stealthily, so that they’re hard to track and detect. And while it works most of the time, sometimes these algorithms get into a feedback loop, a la the fly, and nine percent of the entire market can disappear. Which happened last year. Called the flash crash of 2:45, trillions of dollars vanished in five minutes. It was the biggest one-day point decline in history and automated high-speed algorithm deals were clearly involved. But which ones? Humans could only pull the plug. And to this day, we can’t pinpoint the exact root of the problem.

These maths are controlling us in compelling and disturbing ways, from Amazon pricing to determining which movies you watch on Netflix (60 percent are selected based on recommendations from algorithms), to which movies get made, and who will star in them (yes, there’s a program for that, too). And we are understanding less and less about what these programs are and how they work.

And they’re moving in next door. The Internet may be virtual, but speed-freak algorithms work best when they’re close to the source of data. So entire buildings in Manhattan are being hollowed out and steel-reinforced to accommodate stacks of servers, just to proximity to a certain 2.9 million square foot 1930’s hotel on 15th Street (which Google bought for $1.9 billion last year). The art deco brick building is one of the world’s premier “carrier hotels” — essentially a gigantic server hub. Distance from here means time and money. If you’re some poor computer all the way down on Wall Street, you’re too late.

Which is funny — because while high-speed rail can’t get traction in the U.S., one company is already blasting and hacking a straight line from New York to Chicago to lay down fiber-optic cable and get those algorithms the speed they need, to close the deal three microseconds faster. Moving people is so 20th-century, but data must get through. So, have we finally become slaves to machines?