by P. W. Singer
“They don’t fight,” said Keegan, playing with the last shrimp on her plate before carefully spearing it. “They beat up some machinery. In a fight, someone fights back.”
“Some global elites, who never set foot in the communities they want dominated by machines, they think they can buy them off with promises of GBI and free money. No, they have the right to work and will fight for it.”
“This guy makes no sense. Now the factory is broken, and with no GBI, they’ll get nothing,” Keegan said.
“At least they did something,” Jared replied. “Damn machines are taking everything. If it keeps up, there’ll be nothing but bots and their few gajillionaire owners. How do they think people are going to afford whatever they make?”34
On cue, the screen narrative shifted to stimulate what it had determined was a brewing debate between the couple over GBI, the proposed Guaranteed Basic Income program.35 In this case, the screen filled with a high-resolution image of Willow Shaw—close-cropped silver hair, tan skin, a form-fitting long-sleeved black T-shirt that stopped at the forearms, where e-ink tattoo panels swirled like darting tropical fish whenever he gesticulated with his hands. He had the look of somebody who spent their days outdoors, a high-mountain guide’s placidity and openness in his smile.
“I don’t think we should blame these workers, but help them.”
“Ugh. Not that guy again,” said Jared. “As if being everywhere in the net wasn’t enough, now he’s all over the feed too.”
Shaw was the most visible face in the new industry that had taken off in the wake of the last market crash. If the previous generation of tech behemoths had been about unleashing information, Shaw’s company KloudSky was about its intelligentization.36 The combination of mass data, cloud computing, and AI not only had created new market efficiencies but entire new marketplaces, such as letting companies bid in real time on mid-modal drone delivery opportunities for shopping rush-hour commuters. His firm took a slice of each and harvested even more data.
“What happened in Indianapolis shows that we’re not doing enough. The question of our modern world isn’t whether automation is coming or even whether it will bring benefits.37 We know that it will. The real question is whether those benefits will be distributed beyond the lucky few like me. The work of SANE found that there is so much we can do to change the way we do business in America, but we have to be sure to never lose sight of the values that make all of that change possible.”
“SANE. I still can’t believe they called it that,” Jared continued. “It’s such obvious marketing.”
Little known outside Silicon Valley a few years ago, in the last year Shaw had taken on a more public role as a member of the White House task force known as the SANE Commission, short for the Study of Alternative and Novel Economics.
“Yeah, but it works,” replied Keegan. “Like ‘pro-life’ or ‘metal-collar job,’ even if you know what they’re doing, the term still conveys a certain point of view every time you hear it.”
“That means understanding at the most human level that our basic desires are shared, they are universal—to be safe, supported, and productive members of society. Losing your job should not mean losing your future.”
“Oh, I know that. I can respect that. But it still grates on me. Kinda like that forehead of his.”
They both knew he was being petty, but the very image of Shaw offered Jared a convenient target, a veritable human embodiment of the change that he wrestled with every day from the couch. You could be mad at a company, but it was easier to hate a person.
“Technology is not destiny.38 AI’s impact on the American worker will be decided not by AI, but by us . . . by the policies that we vote on and the organizations we choose to build . . .or not. We can take advantage of the new opportunities provided by this new economy. Or we can miss this closing window to create shared prosperity for all.”39
“Yeah, it’s a bit too smooth,” Keegan said, trying to make peace. “Might be from those nanomites that the viz stars and billionaires are injecting to combat aging.”
The second she said it, she regretted it, realizing they’d soon be peppered with skin care and aging ads.
“People may think that the SANE Commission is coming up with something new here, but there is little that is more baked into our very history as a nation. Indeed, guaranteeing someone a basic income in a time of tech change was first written about by Thomas Paine, the very man whose ideas helped inspire the American Revolution.40 Today, in a new Industrial Revolution, we need that same kind of bold thinking—”
The buzzing of Jared’s Watchlet marked a pop-up notice breaking into the feed. Thousands of miles away, an elderly woman was stirring in her bed in a first-floor bedroom in a Craftsman bungalow and would soon need a reminder about her medicine. That would be another rewards opportunity for Jared. If he was back online before the actual client request for assistance came in, then it would be another prime.
“Back to the coal mines,” Jared said with a smile, and stood up. He couldn’t help himself, she knew. It wasn’t the drugs, but how he was wired. For all of the talk of giving people some kind of guaranteed income regardless of whether they worked, some people were defined by it. They drew who they were from it.
Shaw continued:
“There’s no fighting change. We have to embrace it or lose everything.”
Keegan nodded at Jared and switched the screen off. She leaned back in her chair and took a sip of wine, staring at how the liquid took the shape of the glass, adjusting to whatever environment it found itself placed in.
Jared headed over to the corner, to where Haley had already fallen sleep, curled up on a pillow with her toys laid out in front of her. He gave his daughter a kiss on the forehead and smiled at Keegan. For all their present distance, here was their joint creation. He headed back to the couch and put on the VR helmet, off to chase more primes. That drive was part of what had so attracted her—type A all the way, he’d powered his way through law school and now he’d do the same through whatever came next.
Keegan stood up from the table and knelt over her daughter, who hadn’t stirred at the kiss.
“Butterfly, time for bed,” she whispered, and carefully picked her up, the girl’s warm body slightly moving to rest her head over her mom’s shoulder. Carrying her to her darkened bedroom, Keegan tucked her daughter under the covers. Just out of reach, Baz hovered in the corner as it wirelessly charged from its wall-mounted power station—a nighttime sentry that Haley claimed kept away any nightmares. Keegan gave her a kiss on the forehead, just where her husband had, and then placed the Lego spider bot on a shelf across from the bed. She reset its camera to face the girl, her own nighttime sentry for warding off the bad.
Keegan quietly closed the door behind her and checked on Jared, whose head was now bopping back and forth, animated again in conversation with his ward. The hallway, lined with black-and-white photos of her and Jared on vacations pre-Haley, led to the office that had previously doubled as the guest bedroom but now was something more. They’d never really formally talked about who got the master and who got the guest room. It had all started with another of their arguments and Keegan just needing to be alone for a night, not even wanting to breathe the same air as him. And then she’d never gone back. It was easier than having that conversation out loud, which would have made it real, and maybe permanent.
Sitting in bed, she opened her work account on her Watchlet and messaged Noritz: There’s no fighting change. We have to embrace it. Agree to program.
FBI Academy
Quantico, Virginia
“We all ready?” Keegan asked the obstacle course instructor. A giant of a man with jet-black hair and a streak of silver through the middle, he had a long mustache and a milky white prosthetic eye, as government healthcare did not cover the augmented implants that allowed a wearer to see. He wore black fatigues that he tucked into his black boots.
“Course clear. We’re green,” said the ins
tructor. The man nodded at Keegan to begin.
Looking out at the quarter-mile sprint obstacle course, its end just visible in the evaporating morning dew, Keegan instinctively rolled her shoulders, unconsciously stretching her body at the memory of having to run it herself years back and, even more, reliving that mental pressure to burn through it so fast so that no man would question her right to be there. It had been generations since J. Edgar Hoover had died and women had finally been allowed to attend the Academy just ten days later. But it still mattered.1 The same course for everyone—rope climbs; wooden walls, some you had to climb over, others with openings designed to simulate a window to go through; the worst were the monkey bars that individually spun. If you really annoyed them, the instructors might even grease them with canola oil. While bringing in as diverse an agent pool as possible was an FBI recruitment talking point, physical fitness was the one place where discrimination was allowed. If the rest of America was losing the war against obesity and inactivity, then the advantage of being fit and healthy only increased. And any perceived advantage was as important as any real edge.
Keegan grew up watching videos of two-legged robots flipping and jumping through warehouses.2 She knew that a humanoid robot could work its way through a set of obstacles like this. That wasn’t what intrigued her. What Keegan most wanted to figure out was how TAMS would deal with the cognitive, rather than the physical, part. Machines could handle labeled data and single, structured tasks easily enough. Where they often crapped out was comprehending things from context only and all the multitasking that went into generalized learning and interaction.3 Basically all the things that humans found easy.
“Good morning, TAMS,” Keegan said. At the mention of its name, the system tilted its head, the physical signal that it was now awaiting command. It was a response she expected after spending the past several days reviewing its operating manuals and testing reports. What she had not had time for yet was to configure the bot’s user interface to her own preferences—or, as she thought of it, “the right way.”
“Good morning, Agent Keegan,” it replied, the eyes lighting up slightly as it spoke.
“TAMS, in one-on-one communication, no longer use my name,” Keegan ordered. “Saying it every sentence is wasted words and time.” These critical first steps shaped how the technology would fit her, rather than the other way around.
“Confirmed,” the system replied. The robot’s face was blank, yet Keegan realized that she’d imagined it was making a slightly miffed expression at being ordered around. Transposing emotion onto the bot—she’d have to watch out for that.4
“So, TAMS, do you know where we are?”
“Yes.”
“Where?”
“Latitude 38.53—”
Keegan quickly interrupted again. “End. Unless numbers requested, provide textual answers.” Computers “thought” in terms of numbers, which could quickly overwhelm a person.
“We are at the FBI Academy obstacle course in Quantico, Virginia.”
“Correct. Do not tilt you head when awaiting verbal or other commands.” She felt herself lapsing into the robotic affect she had used when she was a wrangler in the Marines.
“Understood.”
With an exaggerated blink to activate a targeting command on her AR glasses, Keegan tagged the obstacle course’s finish point—a rope climb that would test upper body and core strength, as well as mental toughness at the end of the run. “Then get your ass moving, machine. Go!”
Keegan thought it might take the system a few seconds to compute her order, but the bot was a blur of movement. Made sense. An AI was as much a big prediction machine as anything else. TAMS hadn’t been there sitting idle, but building scenarios and likelihoods of potential commands that it would receive based on its context.
The other thing that was notable was that the machine didn’t go directly to the finish point that Keegan had identified for it, but took off toward the route of the course. It had gauged that the humans wanted it to follow the rules of the environment.
As TAMS took off at a run, Keegan focused first on the feet of the system. The system design was human for interface but had clearly also been influenced by studies of nature. While it had two legs, the bot’s spine flexed slightly while it ran. This both balanced its movement as well as gave it more of a recoil from each step—like a cheetah. She’d have to look closer, but they also likely mirrored the animal’s retractable claws too, giving it a grip into softer surfaces, like that of a soccer player’s cleats.
TAMS quickly made it to the first obstacle, a low wooden log set just above waist level on a human, about chest level on the smaller machine. It was an ease-into-it first waypoint for trainees. There was no pause and the machine didn’t go below or around the log; it understood the more advanced rules of the environment’s prescribed behavior. Its legs pushed it upward, using both the flex of its joints and a slight expansion of the legs firing like small pistons. The momentum carried it over the log, but at the apogee it touched the log with one hand and pivoted. With hips and shoulders that were fully articulated and a waist that could turn 360 degrees, the robot’s moves looked oddly organic—it had not just been building likely scenarios of what it might be asked to do, but was mirroring clips it had reviewed of human trainees on the course. What was not humanlike was how quietly TAMS moved. It was not just that there was no grunting or breathing, but there was also a lack of noise at the contact points. As it made it to the next obstacle, a set of metal bars to swing on, the rubber grips on its hands muffled even that.
Keegan and the instructor ran alongside the robot as it went, Keegan tracking the time and TAMS’s power levels on the AR glasses.
“I did it in a minute fifty-two back in my class,” Keegan said. “What’s the record now?”
“Not bad,” said the instructor. “Had a trainee do it in a minute twenty-one just a couple months ago; she was a Ranger, and before that ex-pro CrossFit. Total badass.”
As they moved toward the end of the course, TAMS began the final rope climb. When it gripped the rope, it seemed flummoxed, as it swung back and forth in a more random, unpredictable manner. It tried to balance, kicking its feet out in the air like a thrashing swimmer, but that spun the rope the other way. After a beat, it gave up and ignored the spinning, pulling itself up one handhold at a time. It was markedly slower progress than the rest of the rhythmic, gymnast-like movements throughout the course.
TAMS reached the top of the rope and slapped it with its hand; again most likely having scoured video in the cloud, it had figured out that the way you marked your time was to slap the top of the post.
“Time!” said the instructor, standing below TAMS.
“A minute six.” He tried to sound matter of fact, but Keegan could sense he was a bit angry that the course record had been broken.
They stood beneath the hanging rope and waited. After a few seconds gawking, the instructor started to laugh.
“Shit,” said Keegan, realizing the limits of the programming. TAMS was still at the top of the rope, locked in place, its hands gripping tightly and its feet dangling in the air. This was on her. She had not told TAMS what to do after it achieved the objective, nor authorized it to go beyond. The A in its name stood for “Autonomous,” but autonomy for a machine was really just about how much leash the human was willing to give it.
“TAMS. Join us here.” Then, worried it might just jump down and get damaged on its very first day, she added, “Use the fast-rope technique.” Whether the machine had pulled the meaning from a clip of a firefighter dropping down from a fire station’s second floor or of Navy SEALs fast-roping onto a cargo ship’s fantail, the effect was the same. With almost no pause, the robot slid down, wrapping its legs around the thick rope to arrest its descent.
“System status report,” Keegan said. The robot’s diagnostic system projected the bot’s internals onto her AR glasses; the layers of circuitry, servos, and batteries were revelatory. The machine was
as much a feat of art as engineering, the product of generations of genius coming together. Keegan circled TAMS like a sculptor weighing whether to carve away another sliver of clay. As she examined it, individual systems reported their relative temperatures and performance. The arms were slightly overheated from hanging on the rope, while the legs had returned to normal parameters.
Now, it was time to see how it adjusted. Keegan turned and walked back toward the start of the course. Unordered, TAMS followed, walking slightly behind her, like a servant staying out of the way. Clearly, “join” and human movement trumped the target designation.
“TAMS. I don’t want you dogging my heels, where I can’t see you. Stay to my side.”
The machine adjusted its pace. Interestingly, it chose to go to her right. Chance selection or had it decided based on an observation that Keegan was right-eye dominant?
Keegan asked it a more important question. “TAMS. How much faster can you run that obstacle course with a second try?”
A moment of silence, just the soft footfall of metal on dirt.
“Four point seven seconds,” said TAMS.
“Not under a minute?” said Keegan.
“No.”
“Why not?”
“The route is not optimized for performance,” said TAMS.
“Optimal or optimized?” said Keegan.
“Optimized,” said TAMS, which made Keegan stop.
“It is not optimized?” said Keegan.
“What the hell does it mean?” asked the instructor. “Is it some kind of critic now?”