Robots may have been at the factory gate in the 1950s, but it’s only recently that they’ve marched, on our orders, into offices, shops, and homes. Today, as software of what Wiener termed “the judgment-replacing type” moves from our desks to our pockets, we’re at last beginning to experience automation’s true potential for changing what we do and how we do it. Everything is being automated. Or, as Netscape founder and Silicon Valley grandee Marc Andreessen puts it, “software is eating the world.” 48
That may be the most important lesson to be gleaned from Wiener’s work—and, for that matter, from the long, tumultuous history of labor-saving machinery. Technology changes, and it changes more quickly than human beings change. Where computers sprint forward at the pace of Moore’s law, our own innate abilities creep ahead with the tortoise-like tread of Darwin’s law. Where robots can be constructed in a myriad of forms, replicating everything from snakes that burrow in the ground to raptors that swoop across the sky to fish that swim through the sea, we’re basically stuck with our old, forked bodies. That doesn’t mean our machines are about to leave us in the evolutionary dust. Even the most powerful supercomputer evidences no more consciousness than a hammer. It does mean that our software and our robots will, with our guidance, continue to find new ways to outperform us—to work faster, cheaper, better. And, like those antiaircraft gunners during World War II, we’ll be compelled to adapt our own work, behavior, and skills to the capabilities and routines of the machines we depend on.
* The internet, it’s often noted, has opened opportunities for people to make money through their own personal initiative, with little investment of capital. They can sell used goods through eBay or crafts through Etsy. They can rent out a spare room through Airbnb or turn their car into a ghost cab with Lyft. They can find odd jobs through TaskRabbit. But while it’s easy to pick up spare change through such modest enterprise, few people are going to be able to earn a middle-class income from the work. The real money goes to the software companies running the online clearinghouses that connect buyer and seller or lessor and lessee—clearinghouses that, being highly automated themselves, need few employees.
CHAPTER THREE
ON AUTOPILOT
ON THE EVENING OF FEBRUARY 12, 2009, a Continental Connection commuter flight made its way through blustery weather between Newark, New Jersey, and Buffalo, New York. As is typical of commercial flights these days, the two pilots didn’t have all that much to do during the hour-long trip. The captain, an affable, forty-seven-year-old Floridian named Marvin Renslow, manned the controls briefly during takeoff, guiding the Bombardier Q400 turboprop into the air, then switched on the autopilot. He and his cabin mate, twenty-four-year-old first officer Rebecca Shaw, a newlywed from Seattle, kept an eye on the computer readouts that flickered across the cockpit’s five large LCD screens. They exchanged some messages over the radio with air traffic controllers. They went through a few routine checklists. Mostly, though, they passed the time chatting amiably about this and that—families, careers, colleagues, money—as the turboprop cruised along its northwesterly route at sixteen thousand feet.1
The Q400 was well into its approach to the Buffalo airport, its landing gear down, its wing flaps out, when the captain’s control yoke began to shudder noisily. The plane’s “stick shaker” had activated, a signal that the turboprop was losing lift and risked going into an aerodynamic stall.* The autopilot disconnected, as it’s programmed to do in the event of a stall warning, and the captain took over the controls. He reacted quickly, but he did precisely the wrong thing. He jerked back on the yoke, lifting the plane’s nose and reducing its air speed, instead of pushing the yoke forward to tip the craft down and gain velocity. The plane’s automatic stall-avoidance system kicked in and attempted to push the yoke forward, but the captain simply redoubled his effort to pull it back toward him. Rather than prevent a stall, Renslow caused one. The Q400 spun out of control, then plummeted. “We’re down,” the captain said, just before the plane slammed into a house in a Buffalo suburb.
The crash, which killed all forty-nine people onboard as well as one person on the ground, should not have happened. A National Transportation Safety Board investigation found no evidence of mechanical problems with the Q400. Some ice had accumulated on the plane, but nothing out of the ordinary for a winter flight. The deicing equipment had operated properly, as had the plane’s other systems. Renslow had had a fairly demanding flight schedule over the preceding two days, and Shaw had been battling a cold, but both pilots seemed lucid and wakeful while in the cockpit. They were well trained, and though the stick shaker took them by surprise, they had plenty of time and airspace to make the adjustments necessary to avoid a stall. The NTSB concluded that the cause of the accident was pilot error. Neither Renslow nor Shaw had detected “explicit cues” that a stall warning was imminent, an oversight that suggested “a significant breakdown in their monitoring responsibilities.” Once the warning sounded, the investigators reported, the captain’s response “should have been automatic, but his improper flight control inputs were inconsistent with his training” and instead revealed “startle and confusion.” An executive from the company that operated the flight for Continental, the regional carrier Colgan Air, admitted that the pilots seemed to lack “situational awareness” as the emergency unfolded.2 Had the crew acted appropriately, the plane would likely have landed safely.
The Buffalo crash was not an isolated incident. An eerily similar disaster, with far more casualties, occurred a few months later. On the night of May 31, an Air France Airbus A330 took off from Rio de Janeiro, bound for Paris.3 The jet ran into a storm over the Atlantic about three hours after takeoff. Its air-speed sensors, caked with ice, began giving faulty readings, which caused the autopilot to disengage. Bewildered, the copilot flying the plane, Pierre-Cédric Bonin, yanked back on the control stick. The A330 rose and a loud stall warning sounded, but Bonin continued to pull back heedlessly on the stick. As the plane climbed sharply, it lost velocity. The air-speed sensors began working again, providing the crew with accurate numbers. It should have been clear at this point that the jet was going too slow. Yet Bonin persisted in his mistake at the controls, causing a further deceleration. The jet stalled and began to fall. If Bonin had simply let go of the stick, the A330 might well have righted itself. But he didn’t. The flight crew was suffering what French investigators would later term a “total loss of cognitive control of the situation.” 4 After a few more harrowing seconds, another pilot, David Robert, took over the controls. It was too late. The plane dropped more than thirty thousand feet in three minutes.
“This can’t be happening,” said Robert.
“But what is happening?” replied the still-bewildered Bonin.
Three seconds later, the jet hit the ocean. All 228 crew and passengers died.
IF YOU want to understand the human consequences of automation, the first place to look is up. Airlines and plane manufacturers, as well as government and military aviation agencies, have been particularly aggressive and especially ingenious in finding ways to shift work from people to machines. What car designers are doing with computers today, aircraft designers did decades ago. And because a single mistake in a cockpit can cost scores of lives and many millions of dollars, a great deal of private and public money has gone into funding psychological and behavioral research on automation’s effects. For decades, scientists and engineers have been studying the ways automation influences the skills, perceptions, thoughts, and actions of pilots. Much of what we know about what happens when people work in concert with computers comes out of this research.
The story of flight automation begins a hundred years ago, on June 18, 1914, in Paris. The day was, by all accounts, a sunny and pleasant one, the blue sky a perfect backdrop for spectacle. A large crowd had gathered along the banks of the Seine, near the Argenteuil bridge in the city’s northwestern fringes, to witness the Concours de la Sécurité en Aéroplane, an aviation competition organized to show
off the latest advances in flight safety.5 Nearly sixty planes and pilots took part, demonstrating an impressive assortment of techniques and equipment. Last on the day’s program, flying a Curtiss C-2 biplane, was a handsome American pilot named Lawrence Sperry. Sitting beside him in the C-2’s open cockpit was his French mechanic, Emil Cachin. As Sperry flew past the ranks of spectators and approached the judges’ stand, he let go of the plane’s controls and raised his hands. The crowd roared. The plane was flying itself!
Sperry was just getting started. After swinging the plane around, he took another pass by the reviewing stand, again with his hands in the air. This time, though, he had Cachin climb out of the cockpit and walk along the lower right wing, holding the struts between the wings for support. The plane tilted starboard for a second under the Frenchman’s weight, then immediately righted itself, with no help from Sperry. The crowd roared even louder. Sperry circled around once again. By the time his plane approached the stands for its third pass, not only was Cachin out on the right wing, but Sperry himself had climbed out onto the left wing. The C-2 was flying, steady and true, with no one in the cockpit. The crowd and the judges were dumbfounded. Sperry won the grand prize—fifty thousand francs—and the next day his face beamed from the front pages of newspapers across Europe.
Inside the Curtiss C-2 was the world’s first automatic pilot. Known as a “gyroscopic stabilizer apparatus,” the device had been invented two years earlier by Sperry and his father, the famed American engineer and industrialist Elmer A. Sperry. It consisted of a pair of gyroscopes, one mounted horizontally, the other vertically, installed beneath the pilot’s seat and powered by a wind-driven generator behind the propeller. Spinning at thousands of revolutions a minute, the gyroscopes were able to sense, with remarkable precision, a plane’s orientation along its three axes of rotation—its lateral pitch, longitudinal roll, and vertical yaw. Whenever the plane diverged from its intended attitude, charged metal brushes attached to the gyroscopes would touch contact points on the craft’s frame, completing a circuit. An electric current would flow to the motors operating the plane’s main control panels—the ailerons on the wings and the elevators and rudder on the tail—and the panels would automatically adjust their positions to correct the problem. The horizontal gyroscope kept the plane’s wings steady and its keel even, while the vertical one handled the steering.
It took nearly twenty years of further testing and refinement, much of it carried out under the auspices of the U.S. military, before the gyroscopic autopilot was ready to make its debut in commercial flight. But when it did, the technology still seemed as miraculous as ever. In 1930, a writer from Popular Science gave a breathless account of how an autopilot-equipped plane—“a big tri-motored Ford”—flew “without human aid” during a three-hour trip from Dayton, Ohio, to Washington, D.C. “Four men leaned back at ease in the passenger cabin,” the reporter wrote. “Yet the pilot’s compartment was empty. A metal airman, scarcely larger than an automobile battery, was holding the stick.” 6 When, three years later, the daring American pilot Wiley Post completed the first solo flight around the world, assisted by a Sperry autopilot that he had nicknamed “Mechanical Mike,” the press heralded a new era in aviation. “The days when human skill alone and an almost bird-like sense of direction enabled a flier to hold his course for long hours through a starless night or a fog are over,” reported the New York Times. “Commercial flying in the future will be automatic.”7
The introduction of the gyroscopic autopilot set the stage for a momentous expansion of aviation’s role in warfare and transport. By taking over much of the manual labor required to keep a plane stable and on course, the device relieved pilots of their constant, exhausting struggle with sticks and pedals, cables and pulleys. That not only alleviated the fatigue aviators endured on long flights; it also freed their hands, their eyes, and, most important, their minds for other, more subtle tasks. They could consult more instruments, make more calculations, solve more problems, and in general think more analytically and creatively about their work. They could fly higher and farther, and with less risk of crashing. They could go out in weather that once would have kept them grounded. And they could undertake intricate maneuvers that would have seemed rash or just plain impossible before. Whether ferrying passengers or dropping bombs, pilots became considerably more versatile and valuable once they had autopilots to help them fly. Their planes changed too: they got bigger, faster, and a whole lot more complicated.
Automatic steering and stabilization tools progressed rapidly during the 1930s, as physicists learned more about aerodynamics and engineers incorporated air-pressure gauges, pneumatic controls, shock absorbers, and other refinements into autopilot mechanisms. The biggest breakthrough came in 1940, when the Sperry Corporation introduced its first electronic model, the A-5. Using vacuum tubes to amplify signals from the gyroscopes, the A-5 was able to make speedier, more precise adjustments and corrections. It could also sense and account for changes in a plane’s velocity and acceleration. Used in conjunction with the latest bombsight technology, the electronic autopilot proved a particular boon to the Allied air campaign in World War II.
Shortly after the war, on a September evening in 1947, the U.S. Army Air Forces conducted an experimental flight that made clear how far autopilots had come. Captain Thomas J. Wells, a military test pilot, taxied a C-54 Skymaster transport plane with a seven-man crew onto a remote runway in Newfoundland. He then let go of the yoke, pushed a button to activate the autopilot, and, as one of his colleagues in the cockpit later recalled, “sat back and put his hands in his lap.”8 The plane took off by itself, automatically adjusting its flaps and throttles and, once airborne, retracting its landing gear. It then flew itself across the Atlantic, following a series of “sequences” that had earlier been programmed into what the crew called its “mechanical brain.” Each sequence was keyed to a particular altitude or mileage reading. The men on the plane hadn’t been told of the flight’s route or destination; the plane maintained its own course by monitoring signals from radio beacons on the ground and on boats at sea. At dawn the following day, the C-54 reached the English coast. Still under the control of the autopilot, it began its descent, lowered its landing gear, lined itself up with an airstrip at a Royal Air Force base in Oxfordshire, and executed a perfect landing. Captain Wells then lifted his hands from his lap and parked the plane.
A few weeks after the Skymaster’s landmark trip, a writer with the British aviation magazine Flight contemplated the implications. It seemed inevitable, he wrote, that the new generation of autopilots would “dispose of the necessity for carrying navigators, radio operators, and flight engineers” on planes. The machines would render those jobs redundant. Pilots, he allowed, did not seem quite so dispensable. They would, at least for the foreseeable future, continue to be a necessary presence in cockpits, if only “to watch the various clocks and indicators to see that everything is going satisfactorily.”9
IN 1988, forty years after the C-54’s Atlantic crossing, the European aerospace consortium Airbus Industrie introduced its A320 passenger jet. The 150-seat plane was a smaller version of the company’s original A300 model, but unlike its conventional and rather drab predecessor, the A320 was a marvel. The first commercial aircraft that could truly be called computerized, it was a harbinger of everything to come in aircraft design. The flight deck would have been unrecognizable to Wiley Post or Lawrence Sperry. It dispensed with the battery of analogue dials and gauges that had long been the visual signature of airplane cockpits. In their place were six glowing glass screens, of the cathode-ray-tube variety, arranged neatly beneath the windscreen. The displays presented the pilots with the latest data and readings from the plane’s network of onboard computers.
The A320’s monitor-wrapped flight deck—its “glass cockpit,” as pilots called it—was not its most distinctive feature. Engineers at NASA’s Langley Research Center had pioneered, more than ten years earlier, the use of CRT screens for transmit
ting flight information, and jet makers had begun installing the screens in passenger planes in the late 1970s.10 What really set the A320 apart—and made it, in the words of the American writer and pilot William Langewiesche, “the most audacious civil airplane since the Wright brothers’ Flyer”11—was its digital fly-by-wire system. Before the A320 arrived, commercial planes still operated mechanically. Their fuselages and wing cavities were rigged with cables, pulleys, and gears, along with a miniature waterworks of hydraulic pipes, pumps, and valves. The controls manipulated by a pilot—the yoke, the throttle levers, the rudder pedals—were linked, by means of the mechanical systems, directly to the moving parts that governed the plane’s orientation, direction, and speed. When the pilot acted, the plane reacted.
To stop a bicycle, you squeeze a lever, which pulls a brake cable, which contracts the arms of a caliper, which presses pads against the tire’s rim. You are, in essence, sending a command—a signal to stop—with your hand, and the brake mechanism carries the manual force of that command all the way to the wheel. Your hand then receives confirmation that your command has been received: you feel, back through the brake lever, the resistance of the caliper, the pressure of the pads against the rim, the skidding of the wheel on the road. That, on a small scale, is what it was like when pilots flew mechanically controlled planes. They became part of the machine, their bodies sensing its workings and feeling its responses, and the machine became a conduit for their will. Such a deep entanglement between human and mechanism was an elemental source of flying’s thrill. It’s what the famous poet-pilot Antoine de Saint-Exupéry must have had in mind when, in recalling his days flying mail planes in the 1920s, he wrote of how “the machine which at first blush seems a means of isolating man from the great problems of nature, actually plunges him more deeply into them.”12
The Glass Cage: Automation and Us Page 5