by P. W. Singer
“And that is where you and TAMS come in.”
“How?”
“During disaster, people can lose that indefinable but real force of faith. They need to believe that their future is going to be better and, having experienced the worst, that they have not lost control. Rather, they can shape what is to come. A reset, of sorts.”
And, of course, you know just what everyone else needs.
“While I was not directly engaged with the flood response, your and TAMS’s actions during it have become significant to this greater project. Besides increasing attention on your work among White House and Justice Department officials, the cloud activity has been notable. The unexpected combination of this calamity and your actions have made TAMS a potential icon. The machine is definitively trending, with symbology analysis finding interpretations at a pivot point between themes of fear and those of human-machine partnership and, ironically, human compassion.”
“It’s proving its usefulness,” Keegan said, trying to dodge the whole matter of people putting their emotions and beliefs onto machines. It was bad enough for a single person to do it, let alone an entire nation.
“This goes beyond any single machine’s utility. Like it or not, you are now the steward of not just TAMS’s training but a larger change for us all. I cannot overstate how crucial your work has become.”
Keegan thought about Noritz’s comment about the magnifying glass. Now it was more like being a bug under it on the sunniest of days.
“As such,” Shaw continued, “my offer of aid stands. In exchange . . .”
And there it is, thought Keegan, the other part of the deal.
“I remain highly interested in how TAMS develops, and your relationship with it, Agent Keegan. I hope you will share any insights you gain along the way.”
Keegan knew that phrasing it as a “hope” was artifice, wrapping in politeness a command that was expected to be followed.
“Sure,” she replied. “And if you learn anything that could be of use about the flood, I hope you’ll let me know.”
She wasn’t a gajillionaire, but she wasn’t going to just roll over for one.
FBI Satellite Facility
Reston, Virginia
“Hail, hail to the hero of the day,” Noritz said, making a quick bow in front of her cubicle.
Keegan looked at TAMS, which was recharging beside her cubicle, and then back to her boss.
“Which one of us do you mean?” she asked.
“Does it matter? You’re the FBI’s new super-team . . . which means we need to talk,” Noritz said, indicating there was something more to be discussed without the bot listening.
She could have instructed TAMS to turn itself off, but anyone who had grown up during the first generation of home-assistant AIs knew not to trust any recording device even when powered down.7
“TAMS, I could really use a drink of water. Go get me a cup of water . . . in an orange mug.” This was something else that Modi didn’t get about the idea of a machine as partner. Even the most junior agent would balk at being sent off on an obvious snipe hunt.
“OK,” TAMS said, and went down the hallway.
“Cruel, but effective,” observed Noritz. “Think it will find it?”
“Eventually. There’s gotta be one somewhere in the facility. If not, it’ll probably 3-D print one. What’s up?”
Noritz slid a chair over from an empty cubicle, setting it up to face her. “How are things going?” he asked. This was the Noritz that Keegan liked working for. She could see him back in the day, two state troopers pulled up alongside one another, their cars facing in the opposite direction, shooting the shit.
There were a lot of ways to answer his question. Too many. Which meant she didn’t share them.
“Fine.”
“You handling TAMS?”
“Yeah, it’s under control.”
My family, well, that’s another matter, she thought. The agent-spouse-parent “identity triangle dilemma,” they had called it in the briefing last year. Noritz had been surprisingly cool during that session, sharing with the office how he felt like he was letting down his kids as a father every time he worked a late evening. But she wasn’t going to unload that on him either. For all that had changed in the Bureau, she still felt pressure in front of a male boss to only show her work side, to demonstrate those other identities weren’t holding her back.8
“Good to hear that. I know I was hard on you before, but it was for a reason. If things change, I need to know,” he said. “ ’Cause things are about to get far more challenging for us.”
“Yeah?” said Keegan, noting how Noritz used the word “us,” not “you.”
“People are seeing what we’re doing with TAMS,” he explained. “They’ve decided they want more. In case you hadn’t noticed, there’s not much good news in this town right now.”
“ ‘People?’ Who?” she asked. “Like ‘people’ in the Bureau or ‘people’ elsewhere?”
Or maybe even a “person,” she thought.
“That’s above your pay grade,” he said. “Let me worry about that.” He was trying to get some of his power back. It was understandable enough. First the deputy director had ignored him, and then he wasn’t invited to the White House meeting. “What you need to worry about,” he continued, “is casework. Bad guys. Arrests.”
“Sir?”
“No more filling sandbags for you and the bot, though that played perfectly. You’re back on the real job . . . with your robot and its mug.”
“I’m not sure it’s ready for an active investigation. You said this is like being under a microscope. Won’t that get us a lot more attention?”
“We’re past that point now,” Noritz said. “It’s been decided the Bureau needs a win, anything, and you’re going to be the one to do it.” He said it in a way that made it clear that arguing the point was not worth her time.
“So who’s the lucky criminal TAMS is going to get to pop its legal cherry on?” Keegan asked, knowing the language would piss him off.
Noritz sighed and rolled his eyes. “Ignoring your evident attempts to dodge this with an assignment to sensitivity training, it is the proverbial good news–bad news case. The good news is that it’s the first case that you and TAMS ever had, so your colorful language is not necessary.”
“Piss boy from Union Station?” she asked.
“That’s how you clean up your language?” Noritz asked. “Yes, the Reppley case. So you’re already familiar with the particulars. There’s only one problem.”
“He wet himself again?”
“Of a sort,” Noritz replied. “He got shanked.”
“Don’t look at me.”
“I know that. You only tried to rip his finger off. Somebody got the word out that he was Sons of Aleppo.”
“He wasn’t.”
“Either way, it wasn’t appreciated and he bled out in the shower.”
“So we investigating his murder?”
“No, Richmond Field Office has it. It has, however, been noticed that you and TAMS were the ones to extract the one lead. You’re being tasked to pull that thread further.”
“The relative? The one that may have gotten a payout via the gaming site?”
“Yes. A certain Michael Harris Simpkins.”
“Where is he?”
“If we knew that, he’d be in a Dizz-Diff already. You’re to find him.”
“I assume I don’t need to ask which one of us you meant by the ‘you’ in ‘You need to find him’?”
“You learn fast.”
FBI Domestic Special Detention Facility
Reston, Virginia
It oinked and shook its body with a tremor that started at its rubber spring-mounted curlicue tail and ran all the way to eyes that rolled back and closed in what looked like ecstasy. Designed to look like a cartoon pig wearing a wooden barrel, the eight-wheeled delivery bot was from the sandwich shop that had remained open in the mall complex, kept in business
by nearby federal workers. The bots were supposed to be cute, but Keegan found them grotesque, especially how, for large orders, they would travel nose-to-tail on the sidewalk.
“Oink, oink. I have a delivery for Lara Keegan,” said the robot.
They were three hours into the search for Simpkins and it was not going well. The quick pull of data by TAMS in the interrogation room had made it all look too easy to the watching crowd of FBI leaders. TAMS’s search algorithms instantly scooped up the easy stuff, but now it was all dead ends. Simpkins was a ghost—no criminal history, no registered place of work or residence, and no social media presence to glean for clues. The only data they could pull was a digital spending trail, funneled only through virtualized accounts, run by offshore holding companies that were not responding to warrant requests. For instance, Simpkins had an online gambling account, but he could have been making bets from anywhere in the world. Keegan then played a hunch that maybe he wouldn’t be able to resist doing it in person. Pulling betting records and facial recognition databases from physical casinos, which had been among the first to track their customers, though, revealed nothing.9
All the network and system outages that had followed the flood in Washington didn’t help. Experts had warned about the physical fragility of the federal and commercial database systems, but doing something about it was a different matter. Now, it was as if somebody had dropped a massive rock in the middle of the pond, and the ripples kept radiating out again and again, knocking different networks offline.
Keegan pulled off her viz and glared at the delivery pig bot. “I didn’t order anything.”
“Oink, oink. I have a delivery for Lara Keegan,” it repeated. Between the set messaging system and the wheels instead of legs, it was pretty low-grade tech, even for a sandwich shop. But the older systems were actually what allowed it to be cleared for use in the federal building. Low autonomy, limited sensors, and a weekly pastrami payoff to the security office greased any counterintelligence concerns.
“I said I didn’t order anything,” she repeated. “Get out of here.” She gave the pig bot a gentle kick under its chin. Hopefully that would help it get the message—or at least reboot.
“Oink, oink. I have a delivery for Lara Keegan.”
She looked over at the door and then down the hallway, to see if somebody was messing with her. No one.
“TAMS, track the order of this. Find out who sent the delivery.”
“I did,” TAMS immediately replied. “Your posture and respiratory rate indicated lower blood sugar levels.”
She looked at the pig bot and then to the orange mug on her desk. “Fuck me.”
Hooking back into the on-site network in the office had meant a whole new set of updates had uploaded onto TAMS. Apparently one of the other TAMS program testers had set their machine for predictive food runs. All she could hope was that his machine (and she was certain it was a he who had trained their TAMS this way) would start randomly wandering off in search of orange mugs.
“New rule: don’t order food for me without asking,” she commanded.
“OK.”
The discussion of food did make her notice just how hungry she was, the thought of it instantly opening what felt like a cavern in her stomach.
Swiping open her pay app, she tapped her Watchlet on the pig’s snout. It squealed in delight and the faux-wooden barrel split open.
“So, what’d you decide to get me?” she asked TAMS, reaching into the barrel to pull out the delivery pack. The pig bot oinked its glee at a successful delivery and then spun on its axis and left.
“A Reuben sandwich, jalapeño-flavored taro chips, and a can of root beer,” the robot answered.
She salivated at the mere mention of the sandwich and wondered if TAMS had detected that response. “How’d you come up with that?”
“While it is not ideal, according to your current weight and body-mass index . . .”
Great, now she was being body-shamed by a robot.
“. . . the order matches observed preferences for food type, but avoids repeating a past meal.”
So TAMS had seen her eat a sausage with chili and therefore concluded her preference structure was messy pork products of midwestern origin. Keegan unwrapped the sandwich and took a massive bite, savoring the corned beef, Swiss cheese, sauerkraut, and Russian dressing mixing together in a perfect fusion of tang, spice, and grease. Thank God it didn’t see me order a salad, she thought.
“So, what’s on the menu for tomorrow?” she asked, taking another bite. “Let me guess. Pulled pork barbecue? Pastrami?”
“Tomorrow, it would not be an advisable choice to eat meat due to ongoing activities at the US Department of Agriculture Food Safety and Inspection Service.”
With a full mouth, she mumbled, “What?”
“The US Department of Agriculture Food Safety and Inspection Service network has reported a series of malfunctions at animal protein processing facilities in both the eastern and the Midwest region.10 A briefing for the press by the secretary of agriculture has been announced for 9 a.m. Eastern Standard Time tomorrow. This matches past episodes of salmonella and/or E. coli that resulted in the suspension of sales of animal protein products.”
Keegan’s chewing slowed, and she put the sandwich down. “Fuck,” she muttered and pulled out a napkin to spit out the half-chewed bite she’d taken. Giving the robot a glare, she tore open the bag of chips, and chewed, hoping the jalapeño flavoring had some kind of medicinal power.
“OK, let’s get back to work before you try to poison me further,” she said. As she mused on the problem with at least some part of her brain refueled, she thought about how they could attack the problem from different angles. They were trying to pull at threads when maybe the design of the quilt of data would reveal something they had missed.
She tasked TAMS to create a series of data visualizations. As they built, she stood and walked around the cubicle, as much to change her perspective as to stretch her back.
An image cloud built across her vizglasses of the people Simpkins might know or seek shelter with. While their person of interest didn’t have any active online profile, TAMS could populate the cloud with a network drawn from their known contact points. It showed his family members and former schoolmates, only up to tenth grade, of course, she noted. Nothing jumped there. There were crosses and connections between them as would be expected, but nothing stuck out. Then TAMS built a network map visualization of Simpkins’s monetary transaction history, envisioned as a ray of different colored lines connecting to various accounts, pulsing with each activity over the last year. They then compared his financial pattern to the network of profiles and their individual accounts and payments, each a similar ray of colored lines, looking for any connection points.
Nothing. She stretched to touch her toes, more in frustration than the need to relieve the tension in her lower back. Even viewed upside down, Simpkins’s activity had the same pattern. But there were no crosses, no payments going back and forth between him and the people in the network they were able to track. And that’s when it hit her—it shouldn’t have the same pattern. It was abnormal that his activity was so normal.
“TAMS, generate a list of his regular transactions. I want to look over them.”
The word cloud dissipated and the screen populated with rows of accounts that Simpkins regularly paid into. Without being asked, TAMS color-coded them in the ways it understood would help a human process the data. The breakdown had different colors for frequency—the weekly payments shaded in blue and the monthly payments in yellow, while one-time payments got a splash of orange. Nothing was in red, as the system knew that human brains couldn’t help but focus on that color.
Still nothing. There was no signal in the noise that either machine or human brain could detect.11
TAMS recategorized them by activity clusters, a new set of colors shading every payment on “entertainment,” like the gambling site that had first caught their eye—distingu
ishing it from the payments to other categories like “utilities.” That was what jumped out when Keegan read it. Making regular payments to the power company was quite the responsible thing to do for someone potentially wrapped up in a terrorism case.
“Pull out the payments to Southern Energy Exchange.” The moment she said it out loud, she smiled, thinking back to the little jokes they too used to bury in unit expense reports.
“Cancel that,” Keegan commanded. “Bring up anyone else who has made a payment to the Southern Energy Exchange. Then geolocate any who are males between the ages eighteen and seventy years old.” She thought about it for a second. “No, you know what, make it between eighteen and eighty-five years old. There’re still some dirty old dudes at that age.”
She took a long pull of root beer and sat back down as the screen transformed into a map flickering with red dots, popping up one after another. And on the edge of the Anacostia River, there was a pulsing cluster of them.
“OK, TAMS, time for you to learn something new.”
Anacostia River Waterfront
Washington, DC
The congestion on the roads into Washington was back to its old levels. With the floodgates restored and the river returned to its normal level, it was as if the traffic on the roads was trying to forget that whatever had gone wrong had run its course.
During the half-hour drive, Keegan kept thinking of calling Jared, but stopped herself again and again. Instead, she switched on the view from her bots in the apartment. He’d left them up, she guessed both to serve as visual evidence that he could be trusted and a reminder that they had no trust between them. The screen showed Haley drawing at the kitchen table with crayons, while Jared wore his rig, back at work. It was all so normal, as if here too nothing had happened outside their home. But it was an illusion. She also noticed he had the backpack on the couch next to him, ensuring that she saw he was keeping the cash.
Keegan cut the feed as they approached the river and took manual control of the SUV. She wasn’t confident enough in the machine’s mapping updates to let it drive itself anywhere near the river with all the recent changes. The coordinates of the pulsing red dot put it just beyond the now long-gone RFK Stadium, which edged the Anacostia Riverwalk Trail. It was conveniently located not too far from the Capitol and a parking lot used by everyone from commuters for the Metro to the employees of a nearby cluster of government buildings to anyone out to jog along the river.
