Machine Learning

Ethan slipped into the back of the conference room in Building 5 without being noticed. Fifty researchers and administrators, jammed into the room lab-coat-to-suit, all faced the projection stage. Today, of course, it would be set for maximum display. The CEO of the company was here, his six-foot-three frame looming over the crowd. Beside him, invisible to Ethan in the crush, would be tiny Anne Gonzalez, R&D chief. For five years a huge proportion of the Biological Division’s resources—computational, experimental, human—had been directed toward this moment.

Anne’s clear voice said, “Run.”

Some people leaned slightly forward. Some bit their lips or clasped their hands. Jerry Liu rose onto the balls of his feet, like a fighter. They all had so much invested in this: time, money, hope.

The holostage brightened. The incredibly complex, three-dimensional network of structures within a nerve cell sprang into view, along with the even more complicated lines of the signaling network that connected them. Each line of those networks had taken years to identify, validate, understand. Then more time to investigate how any input to one substructure could change the whole. Then the testing of various inputs, each one a molecule aimed at the deadly thing near the center of the cell, the growing mass of Moser’s Syndrome. All this hard work, all the partnering with pharmaceutical companies, in order to arrive at Molecule 654-a, their best chance.

So far, no one had noticed Ethan.

The algorithm for 654-a began to run, and in a moment the interaction combinations produced the output on the right side of the screen. Only two outputs were possible: “continued cell function” or “apoptosis.” The apoptosis symbol glowed. A second later, in a burst of nonrealistic theatrics, the cell drooped and sagged like one of Dalí’s clocks, and the lethal structure at its heart vanished.

Cheering erupted in the room. People hugged each other. A lab tech stood on tiptoe and kissed the surprised CEO. They had done it, identified a possible cure for the disease that attacked the bodies of children, and only children, killing half a billion kids worldwide in the last five years. They had done it with molecular computation, with worldwide partnerships with universities and Big Pharma, and with sheer grit.

Someone to Ethan’s left said, “Oh!” Then someone else noticed him, and someone after that. Ethan’s story was company-wide gossip. The people at the front of the room went on burbling and hugging, but a small pocket of silence grew around him, the embarrassed silence of people caught giggling at a wake. Laura Avery started toward him.

He didn’t want to talk to Laura. He didn’t want to spoil this important celebration. Quickly he moved through the door, down the corridor, into the elevator. Laura, following, called out, “Ethan!” He hit the DOOR CLOSE button before she could reach him.

In the lobby he walked rapidly out the door, heading through the rain toward his own facility. Buildings of brick and glass rose ghostly in the thick mist. MultiFuture Research was a big campus, and he was soaked by the time he reached Building 18. Inside, he nodded at Security and shook himself like a dog. Droplets spun off him. What the hell had he done with his umbrella? He couldn’t remember, but it didn’t matter. The important thing was to get back to his own work.

He didn’t belong at a celebration to defeat Moser’s Syndrome.

Too late, too late. Way too late.

Building 18 was devoted to machine learning. Ethan’s research partner, Jamie Peregoy, stood in their lab, welcoming this afternoon’s test subject, Cassie McAvoy. The little girl came with her mother every Monday, Wednesday, and Friday after school. Ethan took his place at the display console.

That end of the lab was filled with desks, computers, and messy folders full of printouts. The other end held child-sized equipment: a musical keyboard, a video-game console, tables and chairs, blocks, and puzzles. The back wall was painted a supposedly cheerful yellow that Ethan found garish. In the center, like a sentry in no-man’s-land, stood a table with coffee and cookies.

“The problem with machine learning isn’t intelligence,” Jamie always said to visitors. “It’s defining intelligence. Is it intelligence to play superb chess, crunch numbers, create algorithms, carry on a conversation indistinguishable from a human gabfest? No. Turing was wrong. True intelligence requires the ability to learn for oneself, tackling new tasks you haven’t done before, and that requires emotion as well as reasoning. We don’t retain learning unless it’s accompanied by emotion, and we learn best when emotional arousal is high. Can our Maip do that? No, she cannot.”

If visitors tried to inject something here, they were out of luck. Jamie would go into full-lecture mode, discoursing on the role of the hippocampus in memory retention, on how frontal-lobe injuries taught us that too little emotion could impair decision making as deeply as too much emotion, on how arousal levels were a better predictor of learning retention than whether the learning was positive or negative. Once Jamie got going, he was as unstoppable as a star running back, which was what he resembled. Young, brilliant, and charismatic, he practically glittered with energy and enthusiasm. Ethan went through periods where he warmed himself at Jamie’s inner fire, and other periods where he avoided Jamie for days at a time.

MAIP, the MultiFuture Research Artificial Intelligence Program based in the company’s private cloud, could not play chess, could not feel emotion, and could only learn within defined parameters. Ethan, whose field was the analysis of how machine learning algorithms performed, believed that true AI was decades off, if it were possible at all. Did Jamie believe that? Hard to tell. When he spoke their program’s name, Ethan could hear that to Jamie it was a name, not an acronym. He had given MAIP a female voice. “Someday,” Jamie said, “she’ll be smarter than we are.” Ethan had not asked Jamie to define “someday.”

The immediate, more modest goal was for MAIP to learn what others felt, so that MAIP could better assist their learning.

“Hello, Cassie, Mrs. McAvoy,” Jamie said, with one of his blinding smiles. Cassie, a nine-year-old in overalls and a T-shirt printed with kittens, smiled back. She was a prim little girl, eager to please adults. Well-mannered, straight A’s, teacher’s pet. “Never any trouble at home,” her mother had said, with pride. Ethan guessed she was not popular with other kids. But she was a valuable research subject, because MAIP had to learn to distinguish between genuine human emotions and “social pretense”—feelings expressed because convention expected it. When Cassie said, “I like you,” did she mean it?

“Ready for the minuet, Cassie?” Jamie asked.

“Yes.”

“Then let’s get started! Here’s your magic bracelet, princess!” He slipped it onto her thin wrist. Mrs. McAvoy took a chair at the back of the lab. Cassie walked to the keyboard and began to play Bach’s “Minuet in G,” the left-hand part of the arrangement simplified for beginners. Jamie moved behind her, where she could not see him. Ethan studied MAIP’s displays.

Sensors in Cassie’s bracelet measured her physiological responses: heart rate, blood pressure, respiration, skin conductance, and temperature. Tiny cameras captured her facial-muscle movement and eye saccades. The keyboard was wired to register the pressure of her fingers. When she finished the minuet, MAIP said, “That was good! But let’s talk about the way you arch your hands, okay, Cassie?” Voice analyzers measured Cassie’s responses: voice quality, timing, pitch. MAIP used the data to adjust the lesson: slowing down her instruction when Cassie seemed too frustrated, increasing the difficulty of what MAIP asked for when the child showed interest.

They moved on, teacher and pupil, to Bach’s “Polonaise in D.” Cassie didn’t know this piece as well. MAIP was responsive and patient, tailoring her comments to Cassie’s emotional data.

It looked so effortless. But years of work had gone into this piano lesson between a machine and a not-very-talented child. They had begun with a supervised classification problem, inputting observational data to obtain an output of what a test subject was feeling. Ethan had used a full range of pattern recognition and learning algorithms. But Jamie, the specialist in affective computing, had gone far beyond that. He had built, “by hand,” one complicated concept at a time, approaches to learning that did not depend on simpler, more general principles like logic. Then he’d made considerable progress in the difficult problem of integrating generative and discriminative models of machine learning. Thanks to Jamie, MAIP was a hybrid, multi-agent system, incorporating symbolic and logical components with sub-symbolic neural networks, plus some new soft-computing approaches he had invented. These borrowed methods from probability theory to maximize the use of incomplete or uncertain information.

MAIP learned from each individual user. When Cassie’s data showed her specific frustration level rising to a point where it interfered with her learning, MAIP slowed down her instruction. When Cassie showed interest in a direction, MAIP took the lesson there. It all looked so smooth, Ethan’s and Jamie’s work invisible to anyone but them.

At the end of the hour, MAIP said, “Well done, Cassie!”

“Thank you.”

“I hope you enjoyed the lesson.”

“Yes.”

“See you on Monday, then.”

“Okay.”

Mrs. McAvoy took Cassie’s hand, exchanged a few pleasantries with Jamie, and led Cassie out the door. It closed. In the corridor, the motion-activated surveillance system turned on.

Jamie beamed at Ethan. “That went really well! Maip—”

“I don’t want to come here anymore,” said the image of Cassie on the surveillance screen.

“Why not?” Mrs. McAvoy said.

“It’s no fun. Please, Mommy, can we never come here again?”

Silence in the lab. Finally Ethan said, “I guess we need to work more on the ontology of social pretense.”

Jamie looked crushed. “Damn! I thought Cassie liked coming here! She fooled me completely!”

“More to the point, she fooled MAIP.”

“All the subagents worked so well on yesterday’s test kid!”

“There’s no free lunch.”

Jamie had a rare flash of anger. “Ethan—do you always have to be so negative? And so fucking calm about it?”

“Yes,” Ethan said, and they parted in mutual snits. Ethan knew that Jamie’s wouldn’t last; it wasn’t in his nature. There they were, yoked together, the Apollo and Cassandra of machine learning.

Or maybe just Roo and Eeyore.

The first time Ethan had heard about Moser’s Syndrome, he’d been chopping wood in the backyard and listening to the news on his tablet. Chopping wood was an anachronism he enjoyed: the warming of his muscles, the satisfying clunk of axe on birch logs, the smell of fresh wood chips on the warm August air. In a corner of the tiny yard, against the whitewashed fence, chrysanthemums bloomed scarlet and gold.

“—coup in Mali that—”

Also, if he was honest with himself, he liked being out of the house while Tina was in it. His year-old marriage was not going well. The vivacity that had originally attracted Ethan, so different from his own habitual constraint, was wearing thin. For Tina, every difference of opinion was a betrayal, every divergent action a crisis. But she was pregnant, and Ethan was determined to stick it out.

“—tropical storm off the coast of North Carolina, and FEMA is urging—”

Thunk! Another fall of the axe on wood, not a clean stroke. Ethan pulled the axe out of the log. Tina came out of the house, carrying a tray of iced tea. Although her belly was still flat, she proudly wore a maternity top. The tea tray held a plate of his favorite chocolate macaroons. They were both trying.

“Hey, babes,” Tina said. Ethan forced a smile. He’d told her at least three times that he hated being called “babes.”

He said, “The cookies look good.”

She said, “I hope they are.”

The radio said, “Repeat: This just in. The CDC has identified the virus causing Moser’s Syndrome, even as the disease has spread to two more cities in the Northwest. Contrary to earlier reports, the disease is transmitted by air and poses a significant threat to fetuses in the first and early second trimester of pregnancy. All pregnant women in Washington and Oregon are urged to avoid public gatherings whenever possible until more is known. The—”

Ethan’s axe slipped from his hand, landing on his foot and partially severing his little toe in its leather sandal.

Tina shrieked. In his first stunned moment, he thought she’d screamed at the blood flowing from his foot. But she threw the tray at him, crying, “You took me to that soccer game last week! How could you! If anything happens to this baby, I’ll never forgive you!” She burst into tears and ran into the house, leaving Ethan staring at the end of his foot. A chunk of toe lay disjointed from the rest, bloody pulp surrounded by chocolate macaroons. Vertigo swept over him. It passed. The newscaster began to interview a doctor about embryonic damage, nerve malformation, visible symptoms in newborns.

Ethan shifted his gaze to the axe, as if it and not a maybe-living-maybe-not molecule were the danger to his unborn child. An ordinary axe: silver blade, hardwood handle, manufacturer’s name printed in small letters. Absurdly, a sentence rose in his mind from decades ago, a lecture from his first tech professor when he’d been an undergraduate: Technology is always double-edged, and the day stone tools were invented, axe murder became possible.

Then the pain rushed in, and he bent over and vomited. After that, he pushed the chunk of toe back into place, wrapped his shirt around it, and applied pressure.

If anything happens to this baby, I’ll never forgive you!

They divorced eighteen months later.

Social pretense was not a problem with one of Jamie and Ethan’s other research subjects, eleven-year-old Trevor Reynod. He barreled into the lab, shouting, “I’m here! Freakish! Let’s go!”

“My man!” Jamie said, giving him a fist bump that Trevor practically turned into an assault.

“Jamie! And Dr. Stone Man!” That was the kid’s name for Ethan. Ethan didn’t object, as long as Trevor stayed well away from him. Trevor suffered from ADHD, although most of the suffering seemed to belong to the tired-looking mother who trailed in after him. A member of some sect that didn’t believe in medication, she refused to allow Trevor to be calmed down by drugs, but computer games were apparently allowed. Ethan suspected that these thrice-weekly sessions were an immense relief to her; she could turn Trevor over to someone else. Mrs. Reynod poured herself some coffee and slumped into the easy chair in the corner.

Trevor pummeled the air and danced in place, knocking over a pile of blocks. Jamie got the bracelet onto his wrist (“Your superpower ring, dude!”) and settled both of them in front of a game console as carefully wired as Cassie’s keyboard. Trevor’s data began to flow down Ethan’s display. MAIP was silent during Trevor’s sessions, adjusting his game in response to his frustration or satisfaction levels but not instructing him. Trevor did not respond well to direct instruction.

The game involved piloting a futuristic one-man plane, ridiculously represented as a bullet-shaped soap bubble. Its flight simulator was state-of-the-art, similar to the one used to train USAF jet pilots, who might eventually have MAIP incorporated into their training sessions. While flying over various war-torn terrains, Trevor had to shoot down alien craft to avoid being vaporized and to dodge falling stars that appeared from nowhere. Jamie’s role was to fire at Trevor from the ground. He almost never hit him, which allowed MAIP more control and Trevor merciless mockery.

“Ha! Missed me again!”

“You’re really good, Trev.”

He was. Like most attention-deficit kids, Trevor could muster enormous powers of concentration when the activity actually interested him.

They followed their plan of transitioning Trevor from the shooting game to one teaching math in the last fifteen minutes of the hour. Trevor’s levels of arousal and engagement fell, but not as far as they had the previous week. This was a new version of the math game, punchier and more inventive. In effect, Trevor was beta-testing Math Monkeys, while Ethan and Jamie gained learning-algorithm data from him.

The session was a success. After Trevor left, shouting about his victory over the math monkeys, Jamie said, “Did you catch that? Maip tried a stutter-and-recover strategy on him! We didn’t program that!”

“Not in quite that form, anyway.”

“Come on, Ethan, she figured out for herself how to apply it! She learned!”

“Maybe.” He would have to do the analysis first.

But Jamie danced around the lab in an exuberant imitation of Trevor. “Freakish! She did it, Dr. Stone Man! You did it! Go, Maip!”

Ethan smiled. It felt odd, as if his face were cracking.

At midnight, Ethan let himself into the modeling lab in Building 6. The place was empty, even the most die-hard geek having gone out on a Friday night for beer and company. “Lights on low,” Ethan said. The lab complied.

He’d told himself he wasn’t going to do this again. It only made everything harder. But he could not resist. This was the only place that felt meaningful to him now—or at least the only place where meaning felt natural, like air, instead of having to be manufactured moment after effortful moment.

The lab contained, in addition to its staggeringly expensive machinery, three “rooms,” each with the missing fourth wall of a theater stage or a furniture showroom. The largest was an empty, white-walled box, used to project VR environments ranging from an Alpine village to the surface of the moon. The two furnished rooms represented living spaces with sofas and tables, onto which could be projected the VR programs: changing a chair from red velour to yellow brocade, setting out bottles on a table. Old stuff, but it was the starting point for the real challenge of modeling three-dimensional “reality” that could move and be moved, touch and be touched. This lab, already a huge profit-maker for MultiFuture Research, was usually the first one shown to visitors.

Some of the programs, however, were private.

Ethan slipped on a VR glove and put his password into the projector aimed at the smallest room. It sprang to life and Allyson was there, sitting on the floor, holding her stuffed Piglet. This was the Allyson he’d brought to the lab near the end of her illness, when it was clear that the doctors’ pathetically inadequate measures could not help her. Four more months, they said, but it had been only two. Ethan was grateful that Allyson had gone so quickly; he’d seen children for whom Moser’s Syndrome took its slower, crueler time.

Tina had not been grateful. By that point, she had barely been Tina.

Allyson had loved Winnie the Pooh. Kanga, Roo, and Eeyore had been her friends, but Piglet had been more: a talisman, an icon. Once she’d told Ethan that she hated Christopher Robin, “because his Piglet can talk to him and mine can’t.”

The 3-D model of Allyson raised her head and looked up at Ethan. It was a tremendous technical achievement, that mobile action on a holographic projection. Right now, Ethan didn’t care. When he’d brought Allyson here, late at night on another Friday, she’d already begun to lose weight. Her skin had gone as colorless as the sheets she lay on at home. Her hair had fallen out in patches. Ethan had known this was his last chance; the following week Allyson had gone into the hospital. When Tina had found out what he’d done, she had raged at him with a ferocity excessive even for her. Although it should have been a warning.

The model of Allyson—or, rather, the voice recorder in the computer—said, “Hi, Daddy.”

“Hi, baby,” Ethan said. And she smiled.

That was it. Ten seconds of Allyson’s short life, and an enormous expenditure of bandwidth. He hadn’t kept his daughter in the lab longer than that; she’d looked too tired. Ethan hoped that the Biological Division’s Molecule 654-a could cure Moser’s Syndrome. But for him, there was only this.

He called up the overlay programs, one by one. Allyson’s skin brightened to rosy pink. Her hair became thick and glossy again, without bare patches. Her little body grew sturdier. Her eyes opened wider. “Hi, Daddy.”

“Hi, baby.” He reached out with the VR glove and stroked her cheek. The sensation was there: smooth, warm flesh.

Over and over he played the enhanced, miraculously mobile model. Throughout, Ethan kept his face rigid, his hands under control, his thoughts disciplined. He was not Tina. He would never let himself be Tina.

No one, not friends or colleagues, had known how to treat Ethan after Allyson, after Tina. “Call us,” friends had said while Ethan awaited Allyson’s diagnosis, “if anything goes wrong.” And later, after Tina, “Call us if you need anything.” But there is no one to call when everything goes wrong, when you need what you can never have back.

“Hi, Daddy.”

“Hi, baby.”

When he’d had his fill, the fix that kept him from becoming Tina, he closed the program and went home.

On Monday, Laura Avery waylaid him as he walked from the parking lot to Building 18. This being October in Seattle, it was still raining, but at least Ethan had remembered his umbrella. She had one too, blue with a reproduction of a Marc Chagall painting, which seemed to him a frivolous use of great art. Laura, however, was not frivolous. Serious but not humorless, she had made important contributions during her months at MultiFuture Research, or so he’d been told. The company, like all companies, was a cauldron of gossip.

“Ethan! Wait up!”

He had no choice, unless he wanted to appear rude.

She was direct, without flirtatious games. Ordinarily he would have liked that. But this was not ordinarily, and it never would be again, not for him. Laura said, “I wondered if you’d like to have dinner one night at my place. I’m a good cook, and I can do vegetarian.”

“I’m not vegetarian.”

“I know, but I thought I’d just show off my fabulous culinary range.” She smiled whimsically.

It was an attractive smile; she was an attractive woman. When they’d first been introduced, Laura had glanced quickly at his left hand, and her smile had grown warmer. He’d taken off his wedding ring the day after Tina had left him, long before she’d killed herself. Later, after someone had undoubtedly told Laura about Ethan’s story, Laura had grown more circumspect. But the warmth had still been there; he hadn’t needed MAIP to read her face. Now, a year after Tina’s death, this invitation—had someone told Laura it was exactly one year? Was she that coldly correct?

No. She was an intelligent, appealing woman aware enough of her appeal to go directly after someone she liked. Why she liked him was a mystery; in Ethan’s opinion, there wasn’t enough of him left to like. Or to accept a dinner invitation.

“Sorry. I’m busy.”

She recognized the lie but hid any feeling of rejection. “Okay. Maybe another time.”

“Thanks anyway.”

That was it. A nothing encounter. But it left him feeling fragile, and he hated that. The only thing that had gotten him through the last year was the opposite of fragility: controlled, resolute, carefully modeled action.

After his encounter with Laura, he threw himself into work, trying to figure out why MAIP hadn’t detected Cassie McAvoy’s social pretense of enjoying her piano lesson. He found a few promising leads, but nothing definitive.

How far they still had to go was made clear by Jenna Carter.

Jamie was good with the children who came to the machine lab. Sometimes Ethan thought this was because Jamie, brilliant as he was, was still a child himself: enthusiastic, sloppy, saved from terminal nerdiness only by his all-American good looks. Untested, as yet, by anything harsh. Other times Ethan felt ashamed of this facile assessment; Jamie was good with kids because he liked them.

Not, however, all of them equally. While Jamie had no trouble with Trevor Reynod, he had to hide his dislike for Jenna, who wasn’t even a test subject, only the babysitter for her little brother Paul.

They came in after school on Tuesday. Paul, at eight years old their youngest subject, went straight to the small table where Jamie had set out a wooden puzzle map of the United States.

“Hey, Paul,” Jamie said. “How’s it going?”

“Good.” Paul had a thin face, a shock of red hair, and a sweet smile.

“Can I put the magic bracelet on you? Have to warn you, though, it might turn you invisible.”

Paul looked uncertain for a moment, caught Jamie’s grin, and laughed. “No, it won’t!”

“Well, if you’re sure—let’s see if you can put this puzzle together. Recognize it? It’s our country, all fifty states. Wow! That’s a lot of states! What a challenge!”

“I can do it!”

Jenna pushed forward. “He can’t do that! It’s too hard! He’s only in the third grade!”

“Yes, I can!” Paul picked up Maine and fitted it into the upper right corner of the wooden holder. “See?”

“That one’s easy, dingleberry! Anybody can get Maine!” She turned to Jamie. “Our mother said I was supposed to do the puzzles today.”

Paul looked up, outraged. “No, she didn’t!”

“Did too!”

“Did not!” Jenna grabbed her brother by the shoulders and tried to pull him out of the chair.

“Hey! Quit it! Dr. Peregoy!”

Jamie detached Jenna’s hands. “Paul, let Jenna try the puzzle. I’ll let you do the flight simulator.”

Paul’s mouth opened and his eyebrows rose: surprise, one of the basic facial-recognition patterns. The flight simulator was a treat usually withheld until the end of each session.

Jenna cried, “No fair! I want to do the flight simulator!”

“Maybe later.” Jamie slipped the sensor bracelet off Paul and onto Jenna, and pushed her gently onto the chair. “After all, your mother said you should do the puzzle, right?”

Jenna glared at him. “Yeah!”

“Then let’s see how fast you can do it.”

Jenna hunted for a place to fit Iowa. Paul ineptly piloted the transparent bubble. (“You have crashed the jet, Paul,” MAIP said.) Ethan wondered what Jamie was doing. Then he got it: Jamie wanted to see if MAIP could detect the fact that Jenna was lying. Ethan studied his displays.

MAIP worked with what was, basically, a set of medical data. It didn’t have the context to interpret what that data might mean. To detect social pretense—which it also couldn’t do yet—its algorithms used a subject’s baseline data, observed data, and contradictions among the ontologies of emotion. But MAIP hadn’t “learned” Jenna, couldn’t yet do cold readings without a subject’s baseline data, and had neither context nor algorithms to detect lies. So it was no surprise that MAIP didn’t recognize Jenna’s lies.

“Well,” Jamie said after the children left, “it was worth a shot.”

“Not really,” Ethan said.

“Mr. Negative.”

“MAIP didn’t even register social pretense for Jenna, no matter how much you led her into lying. We’re just not there yet.”

Jamie sighed. “I know.”

“What you just did was no better than a polygraph, and there’s a reason polygraphs aren’t admissible in court. Not reliable enough.”

“Yeah, yeah, you’re right. But there should be some way to do this.”

“We need to solve the problem of social pretense first, and with subjects that we do have baseline data for.”

Jamie said, “Maybe if we…no, that wouldn’t work. And—oh, God, I just thought of another problem. Jenna clearly knew she was lying, but what if someone has convinced themselves that they feel one thing but are actually feeling something different? Like, say, a woman who convinces herself she’s in love, even though all she really wants is to have babies before her biological clock stops ticking? She doesn’t really feel love for some poor schlump but thinks she does, to ease her conscience about trapping him?”

Was this a glimpse into Jamie’s personal life? If so, Ethan didn’t want to know about it. He said, more primly than he intended, “Oh, I think most people know what they really feel.”

Jamie gave him a strange look. “Really, Ethan?”

“Yes. But the point here is that MAIP didn’t know.”

Jamie picked up Texas and fitted it into the puzzle, his head bent over the small table, his hair falling forward over his face and hiding his expression.

December, and still raining. Ethan went to the modeling lab late on a Sunday afternoon. He was alone in the building; it was almost Christmas. Water dripped from his raincoat and umbrella onto the floor. “Lights on.”

“Hi, Daddy.”

“Hi, baby.”

Allyson smiled, and the recording ended. He clothed her in artificial health, pink cheeks, and lustrous hair, and started it again.

“Hi, Daddy.”

“Hi, baby.”

He stroked her cheek. Soft, so soft in his VR glove. But Allyson had not been a soft child. Not noisy and obnoxious like Trevor or Jenna, not hidden and falsely polite like Cassie. Allyson had been direct, opinionated, with a will of diamond. She and Tina clashed constantly over what clothes Allyson would put on, what her bedtime was, whether she could cross the street alone, why she drew butterflies instead of the alphabet on her kindergarten “homework.” Ethan had been the buffer between his wife and daughter. It seemed ridiculous that a five-year-old had to be buffered against, but that was the way it had been. Allyson and Tina had been too much alike, and when Tina had blamed not only Ethan but herself for exposing Allyson to Moser’s Syndrome, Ethan had not seen the danger. Tina, dramatic to the end, had thrown herself under a Metro train at the Westlake Tunnel Station.

Allyson would not have grown up like that. As she matured, she would have become calmer, more controlled. Ethan was sure of it. She would have become the companion and ally that Tina had not been.

“Hi, Daddy.”

“Hi, baby.”

The recording stopped, but Ethan talked on. “We’re having trouble with MAIP’s ability to attune, Allyson.”

She gazed at him from solemn eyes. Light golden brown, the color of November fields in sunshine.

“‘Attune’ means that two people are aware of and responsive to each other.” And attunement began early, between mother and infant. Was that what had gone wrong between Allyson and Tina? He and Allyson had always been attuned to each other.

Ethan reached out both arms, one in the VR glove and one bare. Both arms passed through the model of Allyson that was made only of light. The gloved hand tingled briefly, but it still moved through the child as if she did not exist.

For a terrible second, Ethan’s brain filled with thick, tarry mist, cold as liquid nitrogen. He went rigid and clamped his teeth tightly together. The mist disappeared. He was in control again.

He turned off the recording, wiped the rain droplets from the floor, and left.

Zhao Tailoring didn’t open until 10 a.m. on Mondays. Ethan, who’d been there at 8:30, waited in a Starbucks, slowly drinking a latte he didn’t want. The Seattle Times lay open on the table, but he couldn’t concentrate. At 9:50 he threw his cup in the trash, left his unread paper, and walked back across the street to the tailor shop. He huddled under the roof overhang, out of the rain.

Tailoring was not part of his life. Ethan bought clothes haphazardly, getting whatever size seemed the best fit and ignoring whatever gaps might present themselves. The window of Zhao Tailoring held Christmas decorations and three mannequins. The plastic-resin woman wore a satin gown; the man, slacks and a double-breasted blazer; the child, a pair of overalls over a ruffled blouse. They looked bound for three entirely different events. The sign said ALTERATIONS * REPAIRS * NEW CLOTHES MADE. At 9:58, an Asian woman unlocked the front door.

“Ethan! What are you doing here?”

Laura Avery, under her Marc Chagall umbrella. Ethan felt his face go rigid. “Hello, Laura.”

“Are you having tailoring done?” Her voice held amusement but no condescension.

“No. What are you doing here? Why aren’t you at work?”

Her brows rose in surprise at his harsh tone. “I had a doctor’s appointment across the street. Nothing serious. Are you having a suit made?”

“I already said I wasn’t having tailoring done. Please stop asking me personal questions.”

Surprise changed to hurt, her features going slack in the blue shadows under the umbrella. “Sorry, I just—”

“If I wanted to talk to you, I would.”

A moment of silence. Ethan opened his mouth to apologize, to explain that he was just distracted, but before he could speak, she turned and stalked away.

“You come in, yes?” the Asian woman said.

Ethan went in.

“You want nice suit, yes? Special this week.”

“No. I don’t want a suit. I want…I want to buy the mannequin in the window.” Incongruously, an old childish song ran through his head: How much is that doggie in the window?

“You want buy what?”

She didn’t have much English. The person who did was late showing up for work. “You come again, twelve o’clock maybe, one—”

“No. I want to buy the mannequin…the doll.” They had finally agreed on this word. “Now. For a hundred dollars.” He had no idea what store mannequins cost.

She shook her head. “No, I cannot—”

“Two hundred dollars. Cash.” He took out his wallet.

They settled on two-fifty. She stripped the overalls and blouse off the mannequin, and, to his relief, she put it in a large, opaque suit bag. Ethan watched its stiff plastic form—hairless, with a monochromatic and expressionless face—disappear into the bag. He put it in the trunk of his car, pushing from his mind every bad B movie about murderers and wrapped-up bodies.

Marilyn Mahjoub was fifteen minutes late for her first testing session. Waiting, Jamie paced, smacking a fist into his palm, dialing the energy all the way up to ten. “You know, Dr. Stone Man, we’d be so much farther along with Maip if all the fucking subfields of AI research hadn’t been—oh, I don’t know—slogging along for sixty or seventy years without fucking communicating with each other?”

“Yes,” Ethan said.

“It’s just such a…oh, by the way, I changed some of our girl’s heuristics. What I did was—are you listening to me? Hello?”

“I’m listening,” Ethan said, although he wasn’t, not really.

“You’re not listening. Maip listens to me more than you do, don’t you, Maip?”

“I’m listening,” MAIP said.

“Why is she so much more here than you are? And why is that kid so late?”

If there was a reason, they never heard it. Marilyn Mahjoub arrived eventually, in the custody of a sullen older brother. Her clothing embodied the culture clash suggested by her name: hijab, tight jeans, and crop top. She had huge, dark eyes and a slender, awkward grace. In a few years, she would be beautiful.

Like Cassie McAvoy, Marilyn played the keyboard. Unlike Cassie, she was good at it. Ethan could picture her in a concert hall one day, rising to cries of “Brava!” However, she did not take well to MAIP.

“Try playing that last section slower,” MAIP said in the warm, pretty voice that Jamie had given her. She was comparing Marilyn’s rendition, note by note, to the professional version in her memory.

Marilyn’s lip curled. “No. It shouldn’t be slower.”

“Let’s try it just to see.”

“No! I had it right!”

“You did really well,” MAIP said. “Can I please hear the piece again?”

Jamie nodded briskly; MAIP was acting to lower Marilyn’s frustration level by offering praise and neutrally suggesting a redo. Ethan studied the data display. Her frustration level was not lowering.

“No,” Marilyn said, “I won’t play it again. I don’t need to play it again. I did it right already.”

“You did really well,” MAIP said. “I can see that you’re talented.”

“Then don’t tell me to do it slower!”

“Mare,” said her brother, with much disgust, “chill.”

Jamie stepped in. “What would you like to play now, Marilyn?”

Her childish pique disappeared. Lowering her head, Marilyn looked up at Jamie through her lashes and purred, “What would you like to hear?”

Christ—twelve years old! Were all young girls like this now? Allyson wouldn’t have been. She would have been direct, intelligent, appealing.

Jamie, flustered (Ethan hadn’t known that was possible), said, “Play…uh, what else do you…what do you want to play?”

Later, after brother and sister had left, Jamie turned on Ethan. “What’s wrong with you?”

“With me?”

“You’ve been distracted this whole session, and you made me deal with that little wildcat by myself! Did you even hear me say that I added heuristics to Maip, matching emotion with postural clues?”

“No, I…. Yes.”

“Uh-huh. Get with it, Ethan! We have to get this right!”

Ethan said, “Don’t take your frustration with Marilyn out on me.”

MAIP said, “Jamie, you seem distressed.”

Startled, Ethan turned toward the computer. “MAIP has your data? Did you give your baseline readings to her?”

“No!” Jamie’s irritation disappeared, replaced instantly with buoyancy; it was like a dolphin breaking the surface of gray water. “Well, I gave her some data, anyway—but I think she applied the postural heuristics and the other new stuff and…I don’t know, you’ll have to do the analysis, but I think she actually learned!”

Ethan gazed at MAIP. A pile of intricate machinery, a complex arrangement of electrons. For some reason he couldn’t name, he felt a prickle of fear.

It was after 10 p.m. when the last researchers left Building 6. In Building 5, the Biological Division, lights still burned. Perez and Chung clattered out together, talking excitedly. Maybe they’d had another breakthrough, or maybe they just loved their work.

Ethan knew he didn’t love his work on MAIP, no more than a castaway loved his raft. Depended on it, was grateful for it, needed it. But love was nowhere anymore, unless it was here.

“Hi, Daddy.”

“Hi, baby.”

The mannequin from Zhao Tailoring wore one of Allyson’s dresses, which had still been hanging in her closet at Ethan’s apartment. The mannequin had jointed arms and legs. Ethan carefully adjusted it into a sitting position. It was a little too tall for the projection, and he had to wrap the bottom four inches of plastic with his raincoat. That was all right; when he projected Allyson onto the mannequin, it looked as if she had plopped herself down onto his coat. Maybe after playing dress-up, maybe just with five-year-old mischief. Ethan set the lights to low, put the stuffed Piglet into her arms, and added the projected overlays, one by one. Healthy skin, glossy hair, bright eyes.

“Hi, Daddy.”

“Hi, baby.”

Ethan’s knees trembled. Slowly he knelt beside her, the coat buttons lumpy under his calves. Lightly—so lightly, the VR glove on his right hand feeling her skin but not the hard plastic below—he used his left arm to hug his daughter.

“Hi, Daddy.”

“What the fuck?”

Lights crashed on full; illusion crashed with them. Ethan jumped up. Jamie said, “What the hell are you doing? Laura called me; she saw you go into—”

“Go away. Leave me alone.”

He didn’t. But Jamie’s face, always so confident, turned a mottled maroon of embarrassment. “Hey, man, I’m sorry. I didn’t mean to—” Then confusion and embarrassment vanished. “No, I’m not sorry! Ethan, somebody has to level with you. You can’t go on like this. I know—we all know—what you’ve been through. As tough as it gets, yeah. But you have to…. This isn’t normal. That model isn’t Allyson. You know that. You have to let go, move on, accept that she’s gone instead of…. This is a perversion of technology, Ethan. I’m sorry, but that’s what it is. And also a perversion of Allyson’s mem—”

He didn’t finish the sentence. Ethan crossed the floor in a mad dash and knocked him down.

Jamie looked up at Ethan from the floor. He wasn’t hurt or even winded; Ethan was no fighter, and Jamie outweighed him by at least forty pounds. Ethan had merely pushed him over. Jamie got up, shook his head like a pit bull hurling away a carcass, and left without a word.

Ethan began to tremble.

His fingers shook so much that he could barely shut down the programs. He left the mannequin sitting in the middle of the floor, a lifeless hunk of plastic, and left his coat and the stuffed Piglet with it. He couldn’t bear to touch any of them.

Outside, in the dark and blowing rain, there was no sign of Jamie. Ethan lurched to Building 18. He had nowhere else to go. He couldn’t drive; he could barely see. The tarry mist was back in his brain, filling it, chilling him to the marrow. There had never been anyplace else to go, not for a year. It frightened him that he couldn’t feel the sidewalk beneath his feet, couldn’t hear the raindrops strike the ground.

In the AI lab, lights burned and the flight simulator was running. Jamie must have been working late. But Jamie wasn’t here now, and if Ethan didn’t do something—anything—he would die. That was how he felt—how Tina must have felt. Thinking of Tina only made him feel worse. He stumbled to the game console and squeezed himself into the small chair in front of it. His hands gripped the controls. At least he could feel them, solid under his fingers: the only solid thing in his world of black mist and tarry cold. Black mist as a train sped into Westlake Tunnel Station, as an unseen virus ate into nerve and tissue…

“You have just crashed the jet,” MAIP said. “Let’s try again!”

Train speeding forward at forty miles per hour…“Hi, Daddy”…keep going keep going don’t give in or you’ll explode you will be Tina…damn bitch how could she leave me like that not my fault Moser’s Syndrome not my fault…don’t give in….

“You have crashed the jet. But I know you can do this—let’s try again!”

Over and over he crashed the jet, even as MAIP made it harder and harder for him to fail. He smashed the jet into mountains, into deserts, into the sea. Again and again and again. Someone spoke to him, or didn’t. There was noise again, a lot of noise; there was destruction and death, as there should be, to classify reality, to match the ontology of everything he had lost—

And then, finally, he realized the noise was his own screaming, and he stopped.

Into the silence MAIP said, “You were very angry, Ethan. I hope you feel better now.”

He gave a little gasp, first at MAIP’s words and then because he wasn’t alone. Jamie stood beside him with Laura Avery.

She said gently, “Are you all right?” And when Ethan didn’t answer, she added, “Jamie called me. After I called him, I mean. I saw you carrying something into Building 6 and—”

Jamie interrupted. “When did you input your data into Maip?”

Ethan said nothing. The tarry cold mist had receded. No—it had vanished. He felt limp, drained, bruised, as if he had fallen off a cliff and somehow survived. You were very angry. I hope you feel better now.

“You didn’t, did you?” Jamie demanded. “You never gave your baseline data to Maip! She did a cold reading on you, extrapolating from free-form observation! We didn’t teach her to do that!”

“Be quiet,” Laura said. “Jamie, for God’s sake—not now.”

MAIP said, “Ethan, I’m glad you feel better. You were both angry and sad before. You were sad even when you smiled.”

Jamie drew a sharp, whistling breath. “Detection of social pretense! I’m sorry, Ethan, I know you’re upset and I said some things I shouldn’t have, but—detection of social pretense! From cold readings! She’s taken a huge step forward—she knows you!”

Ethan said, not to Jamie but to the complexity of machinery and electrons that was MAIP, “You don’t know me. You’re a nonlinear statistical modeling tool.”

Laura said, “But I’m not.” She put a tentative hand on his arm.

Jamie said, “Maip’s not, either. Not anymore. She learned, Ethan. She did!”

Ethan looked at the flight simulator, which flashed the total number of jets he had crashed. He looked at MAIP. He saw the mannequin, a pathetic lump of plastic that he had left in Building 6.

Ethan rose. He had to steady himself with one hand on the game console. Laura’s hand on his arm felt warm through his damp shirt. He didn’t, he realized, know any of them, not really: not Laura, not MAIP, not Jamie. Not himself. Especially not himself.

He would have to learn everything all over again, reassess everything, forge new algorithms. Starting with this moment, here, now, to the sound of rain on the roof of the building.