by Rahul Bhagat
Martin paused and thought for a moment. This was a good time to break the news. “Listen, Robert. I have something to talk about.”
“Your early retirement?” Blair asked.
Martin was taken by surprise. “How did you know?”
“Come on, Martin. I’ve known you for more than twenty years. You think I don’t know what’s going on in your life? I’ve been waiting for you to tell me about her,” Chief Blair said.
Martin turned red; he cleared his throat. “I meant to tell you, Robert. It was just that I didn’t want to jinx it. You know what I mean.”
“I know. I’m not holding any grudges.”
“I think I’ve met the woman I want to spend the rest of my life with,” Martin said. “We’re planning to move to the country. Maybe start a hobby farm. Get away from this mad world.”
“I’m happy for you, my friend. Not many of us can afford early retirement,” Chief Blair said with a smile. “Now we’ll have to look for your replacement.”
There was an incoming call. It was Officer Padma Laxmi from the accident site.
“Stream her on the Wall,” Blair instructed his digital assistant.
Martin turned sideways to look at the Living Wall. It displayed a dark granite wall with pink veins. The wall flickered and then turned pitch-black. A moment later, a close-up shot of a young woman appeared in the middle of the Wall. Her uniform was crumpled, and her eyes were puffy. She shifted from foot to foot and started talking rapidly about the accident. She quickly recounted the chain of events from the night before, skipping the minor incidence with the janitor bot.
“This guy, Charlie Doud, says he’s in charge. It’s supposed to be an NTSB investigation or something,” Laxmi added sullenly at the end.
Blair looked at Martin and rolled his eyes. “Yes, an AV accident is a federal investigation, treated similar to a plane crash. Have they IDed the victims yet?”
“Yes, sir, they have.” Laxmi looked down to consult her notes. “I have their names. The girl is Paige Callaghan, twelve years old, and the woman is Julie Marse, sixty.”
Martin couldn’t hear the last part. A single-seater quadcopter appeared behind Padma Laxmi and drowned out her voice. But for some reason, his heart had started beating hard. He got up and walked over to the Living Wall.
“What did you say?” he asked. “What were the names?”
“Sorry, sir. Did you want me to repeat the names?” Laxmi shouted over the sound of quadcopter.
“Yes.” Martin raised his voice.
Laxmi looked down again. “Paige Callaghan and Julie Marse.”
Martin gulped and took out his phone. He fumbled as the phone slipped in his sweaty palms. He dialed Julie’s number. Ring… ring…
There was no response. He called again. Same result.
Martin looked at Blair, a thousand things racing through his head. He was not able to hold on to any single thought. Things were happening too fast. He desperately wanted a drink.
“Martin. Martin. What’s going on?” Blair was standing face-to-face with him.
He snapped back to reality.
“It’s Julie, my Julie,” he said with a blank expression.
Martin listened in a daze as Blair urged him to take the emergency quadcopter. Numb, he moved mechanically and made his way out of Blair’s office, to the elevator, and the roof.
EIGHT
THE QUADCOPTER, HIGHLY maneuverable thanks to its four tilting rotors, was the vehicle of choice for emergency personnel. Fire, police, hospitals—they all had it. These were compact, lean machines that were perfect for zipping between tall towers, conveniently removed from the crowd on the streets.
The quadcopter Martin took to the accident site was an extended-range version. It was also roomier. When it arrived at the accident site, its powerful propellers kicked up wet leaves from the ground and made them swirl in the air. The flying machine hovered then gently lowered itself onto the asphalt. As soon as the rotors powered down, Martin unlocked the glass shield and stepped outside.
A police officer approached the quadcopter. Martin recognized Laxmi from Chief Blair’s office. She looked at him with a pained expression and mumbled a few words of condolence. Martin nodded mechanically, his mind in a haze; he desperately wanted to see Julie. They walked to one of the parked ambulances.
It was dark and cool inside. There was a young paramedic sitting next to the gurney. On the stretcher lay Julie’s body. She looked quiet and peaceful, as if in a deep sleep. It was hard for Martin to believe that she was never going to wake up again. He wanted to shake her and ask her to get up and prove all these people wrong.
The young paramedic excused himself and stepped outside. He closed the door of the ambulance behind him and left Martin alone with Julie. A diffused bluish-white light permeated the interior. Emotions overwhelmed Martin. He sank down to the ground and fought back tears. A feeling of hopelessness and utter emptiness engulfed him, and he couldn’t stop from crying. He broke down, sobbing in hiccups.
For the next few minutes, Martin let his emotions run their course. But then he stopped, took out a napkin, and wiped tears from his eyes. The expression on his face had hardened. He stood up and looked at Julie’s face. She lay peacefully, looking like an angel. Martin extended his arm and cleared twigs and strands of hair from her face, then he bent down and gave her a gentle kiss.
When he stepped outside the ambulance, Laxmi was waiting for him. He wanted to visit the crash site, so they walked to the edge of the highway. There, standing by the twisted guardrail, Laxmi showed him the site below. A group was gathered around the upside-down AV. People in different-colored uniforms milled around.
Martin decided to go down into the ravine. He went alone, insisting that Laxmi head home. He took his time navigating the uneven, slippery surface.
At the bottom, he asked for Mr. Charlie Doud and was directed to a man in a brown coat, who was inspecting something inside the vehicle. Martin waited for him to finish his task. Eventually, the man pulled his head out of the vehicle, and Martin introduced himself. Charlie treated Martin with a lot more respect than Laxmi had and took him around to introduce him to the rest of the team. They stopped their work and gathered around Charlie, who gave an overview of the operation. Martin asked Charlie if he had any guesses about the cause of the accident.
“I can’t say. It’s too early,” Charlie said. “Seconds before the crash, the AV braked hard and desperately tried to avoid the plunge. It hit the guardrail, flipped over, and landed on its head. That’s what we know so far.”
He asked Martin to have a look inside the vehicle and pointed at the chair facing outward.
“Was it in the manual mode?” Martin asked him. The Julie he knew would never manually operate an AV.
“No, we don’t think so,” Charlie said.
“Then what else could it be?”
Charlie looked up at the sky and said, “Nature is inherently unpredictable. We are still trying to divine weather, but even with all our fancy computers and sophisticated algorithm, we can’t forecast with one hundred percent accuracy. We don’t know why that AI driver decided to jump in the ravine. The problem with these deep learning systems is that we can’t really tell how they arrive at a decision. It’s a black box, and that makes our job harder.”
Deep learning systems? Martin wasn’t sure what Charlie meant and made a mental note to ask Natalie. “If the issue is with the AI driver, then doesn’t that put other Lott vehicles at risk?” he asked.
“It does,” Charlie said. “We’ve already issued an industry-wide recall for this model. Lott Enterprises has put all vehicles with this driver in safe mode; they’ll revert it to the last stable version while we look for the bug. You know, these systems are robust and thoroughly tested, but no one can guard against black swan events.”
“Black swan events?” Martin frowned.
“One in a million events, like a major earthquake,” Charlie said. “There is no way to predict them. I
t could be one of those things. This is going to take a lot of meticulous work, and even then…”
Someone came behind Charlie and said something in his ear. Charlie excused himself and stepped away to speak with him privately.
While Charlie was busy, Martin decided to check the wreckage. He went around the perimeter and looked at the smashed-up vehicle. He wondered if Julie had suffered in her last moments. Probably not. Death must have been quick, he tried to reassure himself. The feeling of utter emptiness he had felt in the ambulance came back and engulfed him.
“Detective, did you know one of the victims?” Charlie asked from behind.
Martin snapped out of his brooding and looked at Charlie. Then he looked at the guy Charlie was talking to. Word traveled fast. “Yes. She was my girlfriend,” Martin said.
“I’m very sorry to hear that, Detective,” Charlie said solemnly.
NINE
MARTIN WAS GLUED to the Living Wall. He was having another chat with Natalie. Since Julie’s untimely death, she was the steady hand that helped him cope with his grief. Otherwise, he would have just bottled up his emotions and continued living a miserable life. Fatso, Natalie’s white Himalayan Siamese cat, jumped on the couch and curled up comfortably in Natalie’s lap.
“Natalie, what are deep learning systems?” Martin asked.
“Deep learning systems.” Natalie nodded knowingly and chewed on a fried zucchini stick. She took a moment to frame her response, and Martin could tell that she was about to go into her habit of describing it in detail.
Natalie said, “The human society started trying to develop artificial intelligence right after microprocessors were invented. But for a long time, we barely made any progress. Try as hard as we might, nothing worked. Then around the beginning of the twenty-first century, suddenly the floodgate of breakthroughs opened. Speech synthesis, no problem. Language translation, no problem. AI became so good that you couldn’t tell the difference between a human and an AI speaker. You know what made the difference?”
“What?”
“Deep learning systems,” Natalie said. “Researchers realized that we couldn’t build an AI the traditional way. There was no way we could have written software code for an AI brain that accounted for all possibilities in real life. There are just too many. So they tried something new. They decided to create artificial neurons as a hardware-software stack that behaved exactly like a human neuron.”
“Like in our brains?”
“Right. They replicated these artificial neurons millions of times and connected them to each other. Now they had the electronic brains of an AI. They started teaching this brain, just like we teach children, through training and repetition. And such systems came to be known as deep learning systems. There are no codes, no if-then-else logic. It’s all neural networks. That’s why figuring out an AI’s intention is no different from figuring out a person’s intent; it can only be done through interrogation.”
“That’s depressing. What happens if we can’t figure out the cause of the accident?” Martin asked.
“It’s possible. It has happened in the past,” Natalie said and grabbed another zucchini stick to munch on. “Lott Enterprises is just going to give everyone a sleek upgrade, free of charge, of course, and we’re all going to forget about this AI driver. Secrets of another failed product buried forever.”
“I worry about the future of humanity,” Martin said and let out a loud sigh. “So much faith in machines, and we don’t even know how they work. They have no accountability to us; this is blind trust.”
“Oh, come on, Martin. Technology is not that bad,” Natalie said. “Look where it has brought human civilization.”
“To the point where our lives depend on an erratic electronic brain that decides to go rogue for no apparent reason.”
“They’ll find the bug,” Natalie said.
“Huh… the bug. But just a minute ago, you said they may never find the cause.”
“You’re right,” Natalie conceded. “We can’t really pinpoint how deep learning systems arrive at a decision. Nature has inherent randomness. Quantum mechanics is all randomness. But science has been able to not only control this random behavior, but also harness it, predict it.”
“Has it? Really?” Martin rolled his eyes.
“Yes, sir. Have you heard of Poisson distribution? Named after a French guy who came up with this graph that was amazingly good at predicting random events.”
“Really?”
“Let’s say you want to predict the number of emails you receive each week. It’s totally random, right? But if you keep track of your emails and chart it week after week, guess what’ll happen. It will start to look like a Poisson distribution, a bell-shaped graph.”
Martin scratched his head.
“And if you know the overall trend, then guess what,” Natalie said.
“What?”
“You can start predicting the probability of specific events. For example, we know that AV crashes are rare, but they do happen. We have been keeping records of accidents for decades, and they do form a bell shape. With Poisson distribution, we can tell whether there is a higher or lower probability of a random event happening. We’ve been using it for all sorts of things, from predicting presence of space junk to the number of sonic booms in a neighborhood.”
“So are you saying this crash was statistically predictable?”
“Not this one per se, but yes, we do have a good idea how many AV accidents we’ll have this year,” Natalie said.
“I suppose it was a high year,” Martin said with a dejected tone.
“I can look that up,” Natalie said. She moved Fatso from her lap and grabbed a tablet.
Martin watched her concentrate on the task. Her eyebrows became knotted, and then she started mumbling something. “Doesn’t make sense…” Martin heard her say.
“What doesn’t make sense?” he asked.
“You know, Poisson distribution is predicting it to be a low-probability year, and this is already the third crash. It defies the pattern.”
“Ha, there you go. Your science is false,” Martin said triumphantly.
“Wait a second. We don’t really know all the factors in play. If you toss a die, each number has a one-in-six chance of showing up. But if you rig the die—make a side heavier or something—you can change that probability.”
“What are you saying?” Martin was suddenly attentive.
“I don’t know. Maybe one of the accidents was intentional. Maybe someone sabotaged it.”
“Sabotage.” Martin sat up straight. His mind started running through possibilities. “Can you give me a number?” he asked.
“What do you mean, ‘a number’?”
“The probability that Julie’s accident was intentional sabotage.”
“I don’t know, Martin. That’s hard to pin down.” Natalie tried to wriggle out of the situation, but Martin would have none of it. He became stubborn like a mule, and eventually, Natalie capitulated.
“It’s almost the end of the year. There should have been only one accident, and we have three. Two could be explained statistically, but three? I’ll give it a fifty-fifty chance that Julie’s accident was sabotage.”
*
Later that night, Martin stood in front of Julie’s grave. Between the trees, he could see city lights twinkling in the distance. He placed the bouquet of pink begonias on her grave and sat down.
Had someone stolen Julie from him? He pondered the question. He felt mad. After years, he was finally happy, and she’d been snatched away from him. Not fair. If that was the case, then he was going to make sure justice was served. He was going to find the guilty person.
He looked down at Julie’s grave and said softly, “No matter what, Julie. No matter what. I promise you that.”
TEN
ALTHOUGH IT WAS still early in the day, Police Chief Blair felt weary. He looked at Martin sitting across from him. He liked Martin; the man had dedicated his life to the
department. The hardworking, tenacious detective had spent countless hours chasing leads. That was one of the reasons his marriage had broken down.
“Martin, what is it you want to know about the Go Team?” Blair asked.
“Generally,” Martin said vaguely. “Who’ll be on it? How does it work?”
Blair placed his hands on his temples. This was not the Martin he knew. Since Julie’s death, he had become unusually quiet and withdrawn, and Blair had been pushing him to take some time off. So when Martin walked into his office and started asking questions about the investigation, Blair was surprised. This was a technical investigation, and Martin ran away from anything technical. So what had changed? Blair took a deep breath.
“A Go Team is formed whenever there is an AV crash. It’s primarily a technical investigation that’s led by the IIC, Investigator In Charge. In this case, that person is Charlie Doud.”
“I see,” Martin said. “Who sits on the board?”
“It’s made up of NTSB board members, various specialists, manufacturers, their engineers.”
“What about representation from the jurisdiction where the accident happened?” Martin asked.
“Yes, that too,” Blair said. “A person from the local police department sits on the board.” He had an uncomfortable feeling about the course of the conversation.
“Have you made any decision on that?”
“I have a shortlist, and I’m meeting with Mr. Doud shortly to discuss.”
“Can I represent?” Martin looked directly at Blair.
Blair tried to reason. “It can get hectic, Martin. We’re talking AV crash. There is a lot of media attention, and this is all technical investigation. This is the world you want to back away from.”
“Not anymore,” Martin said with conviction. “I owe it to Julie.”
“You have a personal stake. Shouldn’t you recuse?”
“She was only a girlfriend. I didn’t know her long,” Martin said.
Blair stopped himself. There was no sense in reasoning with Martin any further. This had become an emotional issue for him. But on the positive side, since this was not a murder investigation, police representation on the team was a ceremonial position, anyway. Most of the work was done by engineers. But then again, he didn’t want to give the impression that the department was not serious about it. The girl had been an heiress to billions of dollars.