by P. W. Singer
As they left the room, Noritz gave Keegan a look that communicated she better do whatever the hell the deputy director planned and keep her mouth shut as she did.
The walk began silently, leaving the detention cell area and moving out into the mall’s atrium. It was all bathed in a pale undersea light due to the green film that had built up on the skylights above. The atrium’s planters were empty and a lack of any trash inside spoke to the austerity of the highly automated DSDF.
Bosch began again, speaking directly to Keegan. At that level, a deputy director didn’t care about the minor difference in ranks between Noritz and Keegan. They were both simply movable parts in the bureaucracy to him. Noritz’s face stayed the same, but Keegan knew him well enough to know that was killing him inside.
“Agent Keegan, I have a proposal for you, which will take you off this case but potentially be far more important to the Bureau.”
She nodded warily. Reppley had been her collar, so she already felt possessive toward the case. It also had the promise to be her biggest yet at the Bureau.
“What you and poor Mr. Reppley met today is a product of the defense industrial complex, which Congress in its infinite wisdom has given to us to test out in a new setting. As I understand from your service jacket, you served in the Marine Corps and worked with unmanned systems there.”
“Yes, sir. I was an MP, but in the Corps, we all fight. Every squad has a designated squad systems operator. That was my task; basically the robot wrangler for the unit, handling ground systems to hunt IEDs and launching microdrones to scout convoy routes. But nothing of this type of sophistication.”
“Well, among the field agents in the Washington Field Office, that still makes you one of our top experts. I assume that is how our algorithm made the match. Correct, Dr. Modi?”
“Yes, sir,” Modi replied. “We matched a variety of factors, from experience to past case history, but that stood out.” As he spoke, the robot from the room appeared from the detention center hallway. Keegan noted that its steps were actually quieter than a human’s on the old shopping mall’s floor tiles, likely the same rubber material as its fingers muffling the sound.
She pegged the system at 5 feet tall exactly, 9 inches shorter than she was. As she expected, its human form on top was matched by two legs; anything else likely would have been perceived as too monstrous, moving from the uncanny valley into the scary. That didn’t mean its designers had bound it to human limits, though. The hip and knee joints were both circular, akin to large ball bearings, that would allow pivots in multiple directions, but pulled by animal-like tendons likely controlled by bio-inspired AI algorithms.13 Similarly, it had feet and toes, but the feet had another ball bearing joint, where the human foot would have had an arch, and the toes splayed out at an angle to address the balance issues that had plagued the early unmanned systems. It was a meld of evolution and engineering.
“Ah, here it is,” said Bosch. “Agent Keegan, meet TAMS, your partner in this little experiment. TAMS, introduce yourself.”
The robot’s head turned toward Keegan and Noritz with a faint whir of servos. “I am a Tactical Autonomous Mobility System—TAMS for short,” said the robot. “It is a pleasure to meet you.”
“Quite the technology, sir,” said Noritz. He stepped closer to the robot, peering into its eyes. Keegan’s assessment was far less romantic: It should be a quadruped. The best bots she’d worked with in the Corps had four legs; they could lose one and still move. The only downside of the doglike designs was Marines got more attached to them.14
Bosch began to walk again and the four followed, the machine just slightly behind. “TAMS was originally developed for a joint DARPA and HSARPA program at the end of the last decade. Then the contractors got involved. However, for all the excitement and cool marketing clips on YouTube, it never made it across the proverbial ‘Valley of Death’ in contracting, from prototype to program of record. Contrary to all the movies we grew up with, it turns out that Terminator robots weren’t the future of war.”
“Yes, sir. We figured that out pretty quickly. Bigger bot is just a bigger target with more to break,” said Keegan. “We needed them to do things and go places that humans couldn’t, not just recreate something the military could already force a warm body to do.”
“However,” said Bosch, “while the military may not have wanted it, there is no such thing as a sunk cost in government programs, and TAMS here has some important fans on the Hill.”
“From, let me guess, the Bay Area?” Noritz said, trying to get into the conversation.
“More important,” answered Bosch. “The land of the Appropriations Committee. The result is that we are now examining TAMS’s suitability in federal law enforcement agencies. In addition to the Washington Field Office, eleven other Bureau field offices will be deploying the system on a provisional basis. This, as I am sure you have deduced with your keen human powers of observation, is where we hope that you come in.”
“You want me for a burn-in,” said Keegan quietly.
“A what?” asked Noritz.
“Why don’t you take this one, bot?” She looked at Modi, who nodded his head in assent.
“TAMS,” Modi said. “Tell Agent Noritz what a ‘burn-in’ is.”
“A burn-in is defined as ‘the continuous operation of a device, such as a computer, as a test for defects or failure prior to putting it to use,’” answered the machine.15 It wasn’t much of a trick to recite a dictionary definition, but one thing stood out: Keegan noticed that the machine had taken on a different intonation than in the interrogation room. Keegan surmised it had selected one researched to be best suited for conversing with FBI agents. At least, she thought, it didn’t have that annoyingly sexist female voice that all the old personal assistant AIs once defaulted to.16
As the robot spoke, they paused in their walk to listen, stopping in front of a high-end kitchen store that had not yet been converted into office space. From the inside, its windows still displayed promotional “Going out of business” signs.
“Yes, a burn-in,” said Bosch. “A highly appropriate term. Let me be clear. This test is about finding answers, not about creating a win for some funder on the Hill. What I want is data. I want feedback from the field on whether this technology is good for the Bureau . . . or not.”
At the last two words, Bosch made eye contact with her. Was that a signal that, despite what he just said, this test had a right answer, one that ensured the Bureau’s human agents had a future, unlike everyone else?
“We really do think, though, that TAMS will help you,” said Modi, quickly adding in what he thought to be a helpful caveat, “as an observation and decision aide. What an AI system can bring to an agent are the very same things it brought to a soldier in the field or a stock trader on Wall Street: machine-speed collection, collation, and analysis of information.17 It isn’t about replacing, but freeing up the human.18 Think of it not as artificial intelligence but augmented intelligence. There’s no way any of us can keep up with everything in the feed, the cloud, as well as whatever surrounds you. In turn, you need that data operationalized.”
“That’s true,” Noritz said, weighing in. “Keegan is perhaps my best field agent, sir, but even she is struggling to manage it all. For instance, she had to drop off the net at Union Station.”
“Too much feed gets in the way. Those designers and programmers haven’t been shot at before, and it shows,” said Keegan, irritated that Noritz would say she was struggling. “Sometimes you have to turn it all off and really see what’s in front of you. That’s how I was able to ID the suspect.”
“Exactly,” said Bosch emphatically. “That’s why we very much hope you will accept this assignment. And while we’re testing TAMS units in other field offices, you’re here with us, and that means your trial counts in a different way. We need to know if TAMS just gets in the way or, worse, endangers the agent or anybody else. If that’s the case, that this machine can’t perform properly, then
we get our own allies on the Hill to bury the program. So get to know it, and show it how to do the job as you see fit. And stay out of the spotlight.”
Keegan made eye contact with Noritz, and could see how much he wanted her to say yes.
“But remember, Agent Keegan,” said Modi, “TAMS is a learning system. You can study its capabilities, but you’ll be the one who can really show us what it can, and can’t, do. In the end, if it fails, it will be in part because of you, that you taught it to. Do you understand that responsibility?”
Before Keegan could answer, Bosch interjected. “Agent Keegan, there’s a moment when we fully comprehend what responsibility means in the Bureau. For some people, it’s a firsthand ordeal that, if they come through it, is transformational. Others learn it at lower stakes, in the form of a story. When I was in my second year in the Bureau, I had a supervisor who once worked on the security detail protecting J. Edgar Hoover. Mr. Hoover had a lot of enemies and just a few friends. I’m sure you are aware?”
“Yes, sir.”
“So my guy’s the junior agent on the detail, fresh-faced and all that, and they’re on the road, in New York City at some midtown hotel. The detail is staying two rooms down from Director Hoover’s suite. Never the one beside it; that’s for the associate director, what my old job was called back then.”
Keegan had heard all the rumors about the FBI’s first associate director, Clyde Tolson. Where was Bosch going with this?
“So the security detail is sitting down to eat room service, big steaks and ketchup, or whatever unhealthy crap they ate back then, when the fire alarm goes off. But no smoke. No fire trucks. Nothing. They gotta assume it could be anything. Maybe a Mob or Commie hit on the director, the alarm a distraction. So they bang on the door, but no answer. They gotta make a decision. So they use the master key to let themselves in, prepared in that instant to lay down their lives for the one man in the country who stands between Democracy and tyranny. But when they get inside, there’s no Bulgarian hit squad or mafioso assassin with piano wire. It’s just the director and Tolson in pink dressing gowns, painting each other’s toenails, while a Nina Simone record blares on the hi-fi.
“Now, you see, things like that were frowned on back then. This was a big fucking deal. Hoover’s blackmailing a good 10 percent of Congress at the time just for being gay and here he was. And so, my guy’s got another choice to make. They may have something now on the director. But J. Edgar sure as hell had something on every single one of them, otherwise he wouldn’t have let them anywhere near his security detail—some affair they were having he could rat to their wife, some crooked real estate deal that would put their father in jail, whatever.
“So, they’re easing their way back out, quiet like under the cover of the fire alarm, discretion being the better part of valor. But then, they actually do smell smoke. There really was a fire. Somebody passed out in bed with a lit cigar three floors down. So they bang on the door a second time. No reply. This time, they kick it in, but wait a beat to give the two of them time to at least put their nail brushes away, give ’em at least that. Then, they politely inform the director that he’s gotta change quicker than Clark Kent and they bundle the two of them out of there, down into the parking garage, and then out the building.
“A few hours later, they’re set up at another hotel and life seems to be back to normal, no one saying a thing about it. But then the associate director comes to my old boss, as he’s the junior guy on the team. Says he left something back at the first hotel, and would my guy be willing to go back and get it for him. You know what it is?”
Keegan didn’t answer, knowing he wanted to say it.
“It’s the nail polish. He’s either testing out his loyalty now or he just really likes that color. Either way, my guy’s got another choice that he did not expect he’d ever have to make as a G-man. So, the thing is, my old boss would never end the story by telling what happened next or even what color the nail polish was. Every damn time, you know what my old boss would say at the end?”
“No, sir.”
“ ‘That’s when I learned that, in the Bureau, you’ll always have career-defining decisions to make. And every single time, the Bureau trusts you’ll make the right one for it.’”
Ballston Neighborhood
Arlington, Virginia
Keegan stood outside the door to her condo, just listening. She rested her forehead gently on the thick wood, closed her eyes, and slowly exhaled. Finding a moment to collect herself before she passed between the two worlds she lived in had become ritual long ago. Leaving in the morning was often early, and done in the dark. She became an FBI agent in those moments, really before she was fully awake. Coming home was sometimes too jarring. She’d learned she had to first catch her breath and give herself time to become a mother—and a wife, whatever that meant now. For a few hours, at least before returning back to the darker places, this was to be where the better angels of her nature prevailed.
Beyond the door was Haley, her four-year-old daughter, and Jared, her husband of seven years. That she knew. But which Jared would it be tonight?
Haley’s laughter could be heard through the door, which was likely not a good thing.
Still propped against the door, Keegan swiped down, the various apps displayed on her Watchlet spinning around her wrist like an old-school slot machine until they landed on what she was looking for, a personal app she’d added. It showed a camera view from a mantislike insect bot that lived atop a bookshelf.
She pressed down on the screen and the app expanded to a full-wrist view. She zoomed the shot to see what exactly Haley was doing.
Haley was making her Barbie doll dance along the edge of the kitchen table with a gray spiderlike bot about half the size and made of builder blocks.19 Whatever imaginary hijinks the two of them were up to, it was making her squeal with laughter, which always revealed the dimple on her left cheek.
God, she loved that girl. They certainly hadn’t mixed up the babies at the hospital. Haley had her blue-green eyes, Jared’s curly blond hair, her own mom’s earlobes, and Jared’s mom’s small pug nose. Haley also seemed to have inherited his side of the family’s build; she was already tall for her age, which was making it more and more difficult to pick her up and carry her to bed.
But as much as she loved Haley, she hated that doll. Both she and Jared had pleaded with Haley to get something else, something less cliché. But before a tense visit to her in-laws in Philadelphia, they’d promised her a doll for good behavior. Barbie had been what Haley had chosen after playing with an old version at her grandparents’ house. At least they’d been able to steer her to the astronaut version, though Keegan doubted that anyone with an actual astrophysics PhD would have chosen such an unsuitable hairstyle for a life of helmets and zero-gravity environments.
Keegan panned the camera to track what was going on in the rest of the room, to see where Jared was while Haley played.
She needn’t have bothered, as he was in the same tangle of blankets on the couch that he always was when working on the HIT. “Human Intelligence Task” is what they called it, a marketplace of virtualized micro-jobs that machines couldn’t do.20
She zoomed the shot in. The VR rig covered his face, so she couldn’t see his expression or any part of his hair for that matter. He kept it shorter now than when they’d first met; it was easier, with the helmet, plus it meant that the white hairs starting to pop up along his temple blended in with the blond and were almost imperceptible. What she could see was that Jared’s neck muscles were visibly tensed under the weight of the helmet. How many hours had he been hitting it today? Three? Ten? The incentive-based nature of it made HIT a perfectly addictive way to help and harm your family.
Keegan watched her husband for a few more seconds, how the weaving of his VR visor had a pattern to it, the slightly exaggerated movements of someone in conversation, when no one was actually in the room with them. The bot did not provide her any audio, but she had liste
ned in on enough of these sessions to know that he was pretty good at it. He had that blend of intelligence and deference that peppered his speech from years of good schools, making him helpful and engaging, but not too pushy.
They were lucky in that and a lot more, she knew. Thank God when Haley was born they’d been covered under Jared’s healthcare plan back then. There was no way Uncle Sam’s HMO would have paid for the polygenic scoring that allowed them to turn off the genetic trait that would have given her Type 1 diabetes.21
And Haley was a happy kid; they’d been able to shield her from most of the changes at home, though she occasionally still cried at night for their nanny. Letting Sandra go had been tough on her, mostly because it was hard for kids to figure out that difference between natural and paid family members.
Luckiest of all, they still had Keegan’s job, which meant they hadn’t had to downsize out of the only home that Haley had ever known. When she and Jared had started dating, her new job as an FBI agent was a source of curiosity and even cool factor to his law-school buddies. No mention was made of the salary, which was probably a source of amusement to people making seven or eight times as much. Now it was steady income coveted by those same lawyers, reduced to pyramid sharecar syndicates or micro-moment legal work, trying to run up their billable seconds for Mexican pharma companies, all to just cover the interest payments on their credit cards. And that was for those lucky enough to still be in the profession—unlike Jared.
He had been on the partner track for financial sector regulatory contract advisory work, a part of law that was boring as hell but lucrative. That all changed when the firm brought in a “human performance potential” advisory firm. It had initially been hired to find efficiencies, but the monitoring doubled as human training for the machine learning systems, their algorithms getting smarter hour by hour, client by client.22 Within three months he was fired, as were 80 percent of the junior staffers. The senior partners took a share of the algorithm itself, which would transcend the firm and potentially create generational wealth.