Hot Dogs and Corn Flakes
Marson took another swig from the thermofoam cup and raised his hand. His tea tasted a bit acrid, but at least it was keeping him warm in the frigid classroom.
“PRISM, ECHELON, Tempora?”
“Yes, all early intelligence collection programs. But it goes back further. Anyone else?”
Parmisi smiled. “Very good, Miss Copeland. STELLARWIND, implemented after the 9/11 attacks, was the first to tie in financial records with global data surveillance. Better, but still not quite there. Someone else?”
A thick silence hovered over the tiered seats.
“Anyone heard of SHAMROCK?”
Parmisi watched the curious faces. “No? Would you believe 1945 was the effective beginning of omni-surveillance? In the limited technology of the mid-twentieth century, of course.”
Marson asked, “Did they even have computers back then?”
“Indeed they did. Very basic computational machines, not suited for complex analysis. SHAMROCK was a simple program created for the same reason we are here today: the greater good of society. Trying to prevent another terrible war, every telegraph message sent in or out of the United States was recorded for later analysis.”
Desoto muttered, “Telegraph?”
“Telegraphy,” Copeland said, rolling her eyes. “An early electronic symbolic-text communication. Jeez.”
“It was the first effective binary communication method. It’s in your glossary if you care to read it,” Parmisi added, applying a hard stare at Desoto. “By the close of the program in 1975, SHAMROCK was recording over a million messages annually—equivalent to a few hundred megabytes of raw data per year.”
A snickering spread through the small auditorium.
“Funny, is it? Go ahead and laugh, but understand the source data was in the physical medium: paper and film. It pales to even the simple PRISM system of the NSA—thank you, Miss Copeland, for mentioning that—which by 2013 was collecting two gigabytes of messages per minute. A petabyte yearly.”
Parmisi highlighted the relevant program names on the screen. “Our humble beginnings in the Mission Data Repository in Utah.State of the art in 2015, though its capacity was limited to a few exabytes. This allowed a year’s data to be stored from PRISM and other SIGINT programs worldwide—hardly enough to produce worthwhile results for our purposes.”
The nodding of heads made him return the motion, happy to be understood.
“Sir, how much is collected now?”
Parmisi grinned as he turned to his desk and picked up his coffee mug. Looking up at the long list of surveillance programs on the screen, he spoke to the class behind him.
“You six are the best of our baseline analysts. This is why you were promoted to the POLM teams. But none of you are cleared for that level of knowledge—yet. Although I might mention that when I came here in 2038 the DELPHI-ORACLE network was analyzing 29 zettabytes yearly. Feel free to apply Moore’s law over the fifteen years since then. Which is why we now have twenty-three cloudweb storage centers spread over five continents.”
He took a long sip as computations chattered around the auditorium. When he heard someone say “yottabytes” he clanked his mug on the desktop to regain their attention.
“The simple truth is, it’s hard to say where it really began,” he said, turning to the students. “High-level secrecy, along with regular disposal of potentially embarrassing documents, tended to erase much history. The British Tempora program, after integration with the NSA’s XKeyscore database browser, certainly started the possibility of effective POL analysis.”
Parmisi sensed he was moving too fast when he saw a few yawns. He slowed down his presentation and suppressed a yawn himself.
“The POL—ah, that is, pattern of life—analysis only reached effectiveness when processing power became sufficient to cross-index the collected data. It’s easy to aggregate information and archive it, but,” he paused a second for effect, “what good is a molecular storage card if you can’t analyze anything on it?”
Nudging the flat yellow square on his desk, well stained with brown rings, he said, “Makes a good drink coaster, that’s about it.”
He continued, listening to the amused response. “Popular turn-of-the-century entertainment had no shortage of speculation about self-aware artificial intelligences, even systems that could predict crime. No doubt this spawned research to produce a working POL system to equal, or even surpass, the fantasy versions. Science fiction has often driven science fact.”
A sharp click came from the controller in his hand and the slideshow on the big screen changed to a bulleted list.
“You will note the 2010s American television show Person of Interest. Has anyone had a chance to view this production in the archives? It’s listed in your historical reference guide.”
After a few seconds of foot shuffling a hand crept up. “I watched a couple. A computer with omnipresent input predicts when crimes might occur. The system’s creator then dispatches his team to help correct or prevent the crime.”
“Mister Desoto is correct. One amusing plot device the show employed was an inability of the computer to name the crime or establish if the so-called ‘person of interest’ was the victim or the perpetrator. Although this was conducive to suspenseful plotlines, in reality it’s rather useless for the societal greater good. Another person of interest is…”
He chuckled at his own joke as his green laser danced back and forth on the screen, underscoring a name.
“Phillip K. Dick, the famous science fiction author, wrote the short story ‘Minority Report’ almost a hundred years ago. It proposed using a clairvoyant team—‘precogs’—to predict crimes. Their visions were interpreted by a computer and sent to a future-crime squad, which then arrested the pre-perpetrator. Later they discovered that the precogs see many possible futures, not one true future. This made the entire program unreliable and a waste of resources.”
Parmisi smirked. “I think you’ll all agree that our current methods are somewhat more trustworthy.”
The room’s yellow fabric walls muffled the laughter as an analyst sitting in the second row spoke. “But it’s similar, isn’t it? Their computer predicted crimes based on analyzed pattern data.”
“Similar, Mister—” Parmisi blinked into his glasses to read the name projected onto his retina. “Ah, Mister Wilson. Yes, well, similar in the prediction, I suppose, though not in the method. Unless you know of a room of mutants somewhere around here, making predictions based on psychic observations?”
Everyone laughed except Wilson, who reddened, slumped into his swivel chair and remained silent.
“No, we don’t have a crystal ball to gaze into for answers. Fortunately, it’s not needed. Why?” His finger panned across the room, stopping on Wilson. “What important data points are collected for analysis?”
“Uh, we, uh…voice calls, text messages, emails.”
Parmisi waved the answers away. “Yes, yes, we know that. What else?”
Wilson sat up, thinking hard. “Traffic and toll cameras, travel habits, GPS tracking logs, parking monitors.”
“Very good.” The finger moved again. “More. Copeland.”
She coughed. “Ahem. School records, tests, grades, subjects studied.”
The finger remained, unwavering. “More.”
“There’s, um, work history, medical history, vacation preferences?”
“Yes. Good.” The finger moved on. “You. What else? Speak up so we can hear you.”
The analyst cowering beside Marson jerked in her seat and the hinge made a loud squeak. Marson laughed, and the finger shifted sideways one seat, zeroing in on him.
“Mister Marson, you have something to impress us?”
Marson flinched. “Um, yes, sir. Entertainment broadcasts viewed, reading preferences, hobbies.”
The finger dismissed him with a flick and returned to its former victim.
“Miss Jiang, you’ve been rather quiet so far.”
She licked her suddenly dry lips. “Favorite foods, restaurants?”
“Yes, excellent.” The finger released her for the moment. “What reflects our pattern of life better than our favorite food, our favorite color, the artwork we prefer? These are deeper connections to our psyche. More so than our neighborhood or the model of car we own.”
Parmisi looked around. “Who said that—Copeland? Yes, music, of course. A personal and revealing datapoint. Sexual preferences, hair and clothing styles, choice of pet, preferred intoxicants. You should understand how these are indexed to obtain the patterns used to predict behavior.”
“But why? Doesn’t the Oracle calculate the predictions for us?”
“The DELPHI-ORACLE system, Mister Nuboto. I’ll thank you to refer to it correctly, there’s no room for inaccuracy here. The D-O does make the associative pattern matches, but the human element reviews the outputs and makes the adjustments. You are the latest human elements, that’s why you’re here, suffering through history lessons. Important lessons about what we do.”
Parmisi sat back on his desk and picked up his mug. “Let me tell you a story. Before any of you were born, there were thousands of government agents, called detectives, spread around the world. They worked long hours to deduce who had committed crimes. Murder, robbery, arson—that sort of thing.”
“You mean after the crime was committed?” Jiang’s voice quavered with disbelief.
“Yes, after the fact.”
“But the victims of murder, they are, were, they’d already be dead. What good does it do to locate the perpetrator then?” Jiang’s face twisted with the terrible thought.
“I know, I know,” Parmisi consoled. “Hard to believe, but true. That’s where the genesis of the D-O program began. When crimes were rampant, psychological profiles began to be built on the offenders. Many thousands of criminal profiles were created over the following years. Our current database holds—” He stared off for a second as he called up the data on his display. “Almost a half million psych-profiles. More than enough for seven sigma accuracy in forecasting patterns of life. Eight nines, as they used to say. We don’t make mistakes.”
Parmisi gave a short wave of his hand. “Those numbers are classified, by the way. Along with most everything else you will learn here.”
Everyone nodded while he sipped his heavily sugared coffee. Jiang was still shaking her head. Kids, Parmisi thought, never read the history texts. He knew he hadn’t when he was their age.
“Back in D-O phase one,” Parmisi lifted his mug in salute of olden days. “Before they got the AIs working together, and the infrastructure reached the needed processing power, things were different.”
Marson said, “What was the accuracy then?”
“Not good. But even when the indexes reached the golden point—that’s what they termed it at the time—the POL predictions were still years away. What sold the project back then was its ability to identify culprits within hours of a crime’s commission.”
Parmisi hopped off the desk and clicked up a new slide. Red overlapping circles filled the screen.
“Example—and here’s where your set theory class comes in handy. Miss Copeland? Let’s pretend that you are mad at…hmm…Mister Desoto, who dumped you for Mister Wilson. To demonstrate your displeasure, you decide to bash Desoto with a brick when he comes home from work.”
Copeland turned and smiled evilly at Desoto, who began fidgeting in his seat.
“This first group is Desoto, found dead in his living room. These data points describe his last hours of life.” Parmisi’s laser pointed out the various subsets.
A new slide overlaid the first with a group of yellow circles.
“Here you see—Mister Marson, let’s say. You will note a few overlaps. A restaurant, a grocery store. Connections, but no more than random chance provides. But here…”
Orange spots joined red and yellow ones, forming complex patterns. Parmisi’s green dot circled in on an orange oval.
“Miss Copeland. There you are, recharging your car two miles from Desoto’s house, though you hadn’t used that station in the past. And here, thirty minutes after Desoto’s time of death, you purchased three drinks in a bar halfway between Desoto’s house and your apartment. A bar you hadn’t visited before that night.”
“Coincidental,” Copeland said, with a shrug. “I was upset over the breakup and I went back to get some clothes I left there. The car ran low, so I stopped at a station. While it was charging I changed my mind about going to Desoto’s. Then I decided to stop for a drink because I was frazzed about the whole thing.”
Nodding at her, Parmisi said, “Yes, good. Nothing provable, coincidence explains it. You can see the problems our ancestors faced: insufficient facts to ensure accuracy, too few data points. Now, look at this.”
The click brought up a confused tangle of multicolored circles, layered so densely the screen appeared brown.
“Eighteen suspects were in the area at the time of death. Mr. Desoto is the purple layer, Miss Copeland is green. Use your tablets to zoom and parse the layers and tell me something interesting.” Parmisi leaned against the rickety metal desk and waited, stretching his legs one at a time.
A low mutter began as his students sorted and rotated the layers, expanding and panning the colored circles.
“Man, there’s a lot of data here.”
“Orange bought a drink at a store just a mile from Purple, I mean Desoto’s house.”
“Magenta was right there, next door! Then he—oh, he’s a plumber on call. Never mind.”
“Green’s car was recorded by a traffic camera,” Marson said. “One block from Purp—uh, Desoto’s house.”
“Hmm. That’s twenty minutes before his approximate time of death,” Jiang said.
Parmisi looked at her with a raised eyebrow. “And how did you determine the time of death, Miss Jiang?”
She peered at her tablet. “The lighting in Desoto’s house is programmed to shut off after fifteen minutes of no motion. The lights came on when he entered, then they turned off sixteen minutes later. So he was most likely attacked within the first minute of entering. Twenty minutes after the traffic camera recorded Green one block away.”
She looked up at the sound of applause to see the instructor clapping. “Very good, Miss Jiang. Very good indeed! Do you see how the six-fold increase in data allowed us to establish the time of death to the minute and then zero in on a probable suspect? For every order of magnitude we increase our descriptive data, we gain a predictive accuracy increase of a hundred times.”
“And this was all before psych-profiles were applied?”
“Yes, Mister Wilson, phase one employed simple physical patterning. Psych-profiles have been used since the 1970s, but the D-O was the first to statistically analyze each subject’s pattern of life while cross-indexing all the profiles.”
He clicked through a few slides, then left a program overview on the screen.
“Phase two began when the integrated AIs went online and began making inferences we didn’t conceive existed. Or thought possible.”
“Based on the profiles, right?” Wilson said, with a slight hesitation.
“It ties together; one facet amplifies another. Let’s look at a method the AIs use.” Parmisi looked around the small classroom. “Let’s say Miss Jiang stops at two different stores and buys a bottle of drain cleaner at each one. Why didn’t she buy both bottles at one store? The stock listing shows each store had plenty. Perhaps she’s covering her tracks.” He watched the group to make sure they were following his logic. “Using the lye to dissolve a body and dispose of evidence.”
“I bought one bottle first,” Jiang said, not to be outdone by Copeland’s earlier improvisation. “Then I thought I needed more because it was a bad clog. So I stopped and bought another bottle.”
“Certainly possible. Did you do any web searches on disposing of a body? Of course not, no one would be stupid enough to leave such an obvious trail. But did you search for the best brand of drain cleaner to use, or how to clear a drain by other means? Does your water usage record show an increase that day, as you tried to flush out the clog? No?”
Jiang raised her hands, helpless at what to say.
“No. So the next step is a Boolean search on known associates of Miss Jiang, priority-based on recent proximity, and see if anyone is missing. Maybe someone she knows has dropped off the feed. Or maybe not, but I’ll flag each associate for continuous monitoring, just in case one of them disappears.”
“But that doesn’t integrate pattern of life cross-analysis with the psych-profiles,” Desoto observed. “You’re just waiting around to see what might happen.”
Parmisi pointed at him. “Correct again. That’s where phase one was when it morphed into phase two. The AI inferences changed everything. In Miss Jiang’s hypothetical case, she is described in patterns only conceivable by the AIs using millions of interconnected data points. The complexity exceeds human comprehension. It’s when phase two really began.”
“Using those inferences, the AIs reverse-engineered the POLs, right?” Marson asked. “That’s when the first teams were organized?”
“Mister Marson, I am ecstatic that someone has read their history. One part at least. Yes, the first pattern-of-life modification team started a year after phase two began. Once it was proved that the D-O could predict the patterns that lead to a crime, it was a short step to taking ante-crime preventative action. In Miss Jiang’s case, a prediction confirms or denies her intentions. Either housekeeping…or murder.”
Copeland half-raised her hand.
“No place to be shy, Miss Copeland. What?”
“My grandfather told me that back when he was a boy there were big protests when the POLM teams began vectoring off crimes. Were you involved in any of that?”
Parmisi gave her a wry smile. “No. Believe it or not, Miss Copeland, I’m not as old as your grandfather, though I’m sure you’ve noticed me limping.” He leaned against the desk and absentmindedly rubbed his right thigh. “Let that be another lesson, always be prepared for surprises. Sometimes things don’t go as well in the field as you’d like them to.”
He was quiet for a moment, then he shook his head and stood up.
“Anyway…let’s be clear, we don’t ‘vector off’ crimes. We vector—adjust—POLs leading to the commission of crimes. Two very different concepts. We recognize that certain patterns, if followed, will lead to criminal behavior. It is our duty to modify them to prevent unfortunate future events from occurring.”
“But the protests did happen?” Copeland asked. “It was a civil rights issue?”
“Yes, and yes. But once the Oracle…the D-O, that is,” he smiled at his lapse, “could classify the modifications required to prevent criminal action from happening, the issue rapidly changed. It went from,‘Do we have the right to do this,’ to ‘How can we not do this if it prevents crimes?’” He shrugged. “The answer was indisputable, especially after the cause-effect test cases.”
“Test cases?”Nuboto said. “I don’t remember anything about those.”
“You wouldn’t. It wasn’t publicized, though world leaders were privy to the results back then. You’re vetted to see it now.” Parmisi glanced at the ceiling. “Dumpling, slide UN-11 please.”
His affectionate nickname for the interface spawned an amused titter. Ignoring it, he directed his pointer up at the screen.
“Here you see the original presentation, made in 2023—which explains the poor graphics. This first chart, on top, is the large control set, a baseline prediction compared to the current crime rate. The second chart for the test group has two columns. Left side, same thing, prediction level against crime rate. But the other one—”
Parmisi circled the column with the laser. He always enjoyed this part. “The right-side column is the results after POLM was applied to the test subjects. You may notice a slight difference.”
Turning to his desk, Parmisi lifted his coffee and let the slide sink in while he took a long, slow sip. The exclamations behind him made him smile as he walked closer to the screen and pointed to the top.
“Here, the control group. D-O predicts crimes, no action is taken, and in 97.8 percent of the cases a crime is later committed.” He tapped the lower edge of the screen.
“But when D-O predicts a crime and the suggested POL modifications are followed, 99 percent of the time no crime occurs. And that was back then. Today our accuracy is a hundred thousand times better.”
“Why wasn’t this made public?” Jiang shook her head. “Everyone would have understood.”
Parmisi turned, nodding. “Yes, I’m sure it would have quieted the dissention, at least somewhat. But if you make this report public you’d have to release the supporting data for peer review. That’s basic scientific protocol.”
“Was there an error?” Jiang asked. “Is that it?”
“The data is pure. What’s inside it causes the trouble. Anybody care to take a guess?”
He paused. No one ever tried, but he always asked.
“That’s okay, you’d probably be wrong. It’s pretty zingo. The data shows the D-O predicted the assassination of a U.S. president a month before it happened. Any speculation on which group the assassin was in?”
The soft whirr of the air conditioning provided the only sound. Parmisi let it linger for a few seconds.
“She was in the control group—no modifications made. How do you think the public would respond to the D-O project being aware of an impending assassination and not taking any preventative action? But the results did convince the prime ministers, kings, queens, shahs, and premiers, including the new U.S. president. The global funding initiative came together without any delays.”
“They didn’t know, did they?” Desoto asked. “Nobody looked at the actual data, just the averages on the charts. Only the AI knew about the assassin.”
“Kudos to Mister Desoto, he is correct. To the AI it was just another crime to track, no different from petty theft or a traffic violation. Back then the AIs hadn’t learned how to make moral and social considerations. Of course we now have subroutines to promote serious predictions to high-priority status for immediate POLM team response. It’s another example of why the human touch is always required, no matter how advanced the D-O grows. AIs act as though they care about us, but they are just opto-electronics. Even if they are fully sentient.”
No one said anything. It was always a shock when recruits found out what had happened, Parmisi thought. And another jolt when they realized the same AIs they grew up with, played games with, trusted as their teachers—those same AIs had let a president die without one nanosecond of consideration.
“Well, now that you are all awake after that little surprise, how ‘bout we watch a video to introduce you to the POLM teams you’ll be serving with? Then we can take a lunch break. Dumpling, roll PMT-1 please.”
No one laughed this time. The room lights dimmed as the video started.
The image jumped around and then settled. A man wearing sunglasses was looking at himself in a rearview mirror. He waved at his reflection, then looked around at the others in the car as he spoke.
Two women and another man, each dressed in green polo shirts and tan pants, put on mirrored sunglasses.
“Dumpling, pause video.” Playback froze, and Parmisi continued. “This view is from the team leader, always referred to as ‘Alpha.’ The reflective coating on their glasses prevents subjects from seeing the team member’s eyes and this adds to the perception of authority.”
Parmisi pointed to the line of data that had been scrolling across the bottom of the screen.
“Here you see the statistical feedback from the AI as it uses the real-time video to analyze the subject’s responses and body language. If it decides further mods are needed, the information is relayed to the team in situ. The AI also determines if a follow-up visit is likely, or whether other methods will be necessary to correct the pattern. Dumpling, resume video.”
As they climbed out of the car one of the women glanced down at her tablet.
“Bert Matten, mod case Delta Theta 1459,” she said.
The video went into fast-forward and Parmisi spoke while the group raced toward the house. “The color of the team’s uniforms promotes a relaxed and accepting appearance to the subject.”
Playback speed slowed when Alpha walked into a living room following a short, balding man in a red-striped bathrobe. The man gestured to the sofa as he dropped into an old reclining chair. It made a grinding sound as the worn out mechanism tried to adjust itself to Bert’s hefty frame.
“You guys have a seat.”
“Thank you,” Alpha said. “But the others will remain standing. Bert, do you know why we are here?”
The feed switched to the woman on Alpha’s left side as Bert shook his head no.
“But you know who we are?”
“Sure, I guess. Yeah. You guys are palmers, right?”
Alpha smiled at him. “Actually, it’s P-O-L-M, not palm. But anyway…Bert, how are you getting along with your wife?”
Bert blinked a few times as he tightened the belt of his robe. “Okay, I guess. Yeah.”
Alpha nodded. “No problems at all then. Everything’s peaches and sunshine.”
“Well, been hitched a long time. I mean, everybody’s got problems, or they ain’t married, right?” With a nervous laugh he looked around at the other team members. “Am I right?”
“Sure Bert, sure. We understand. But nothing making you mad, angry, sad?” Alpha gave a quick, dismissive shrug. “Who knows, maybe we can help.”
Bert ran his hand over his wisp of brown hair and then rubbed the back of his neck for a second. “Angry? Naw. I mean, sure, I guess I wish she’d get a steady job, yeah. Sure.”
Alpha nodded and leaned back on the tattered fabric of the sofa, shifting his shoulder a few inches to avoid the sharp end of a spring.
“Plus she’s always going out with her friends, staying out all night. Coming home late, you know. I mean, I got to get up early for work, right? Always out till two, three something, I don’t know. You know?”
Alpha looked up at the woman recording the session and she nodded back.
“Sure Bert, absolutely,” Alpha said. “Thing is, Bert, we’re pretty sure you’ve been thinking about killing your wife.”
Bert coughed and sat up in the chair, which protested the unaccustomed quick action with a rusty groan.
“I haven’t tried to kill my—”
“Bert, take it easy, relax. Nobody’s saying you did anything except think about it. But here’s the thing, Bert, and I need you to listen closely to this.” Alpha sat forward on the couch and tapped on a hairy knee poking out of Bert’s robe. “We’re going to have to insist that you stop buying hot dogs and corn flakes cereal. From now on, you’ll buy hamburgers and puffed wheat instead. Understand?”
Bert’s large head bobbed up and down as he digested the words. “Puffed wheat, burgers. Okay, yeah, sure, I…I can do that. Yeah.”
Alpha mustered a warm smile. “There you go, Bert, see? Easy as that.” He twisted slightly on the sofa and glanced at the teammate behind him who had cleared her throat. “Beta, I miss something?”
She read the AI’s update off her tablet. “Stop listening to classical music, Bert.” Beta looked up and smiled. “You should switch to neo-techno, before it’s too late.”
“Okay Bert?” Alpha said. “Got all that?”
Bert nodded again. “Yeah, sure. No problem.”
“I hope so.” Alpha’s smile evaporated as he stood. “Because if we have to come back, it’ll get messy. And trust me, Bert—you won’t like that.”
“Okay, that’s lunch. See you in an hour.”
As the recruits filed out and Parmisi put his controller down, a voice called out from behind him.
“Hey, still going to the deli today?”
Parmisi turned around to the man weaving his way through the outgoing flow of bodies.
“Hey Roger. Yeah, lemme grab my phone.”
Parmisi pulled open a desk drawer as Roger walked up, watching the door swing shut behind the last of the students.
“How they doing?” Roger asked, jerking his head toward the doorway.
“Alright, I think.” Parmisi folded up his flexphone and slipped it into his shirt pocket. “This batch is pretty sharp.”
Roger nodded, looking at a stack of charts on the desk. “You think they ever figure out the promotions are just their own modifications?”
With a serious expression Parmisi said, “Did you?”
He spun and headed toward the door, slightly favoring his right leg.
Roger stood by the desk for a second, his forehead wrinkling. “What? Hey, what’d you say? Wait a…hey!”
He ran to catch up as Parmisi, hungry and grinning, pushed into the hallway.
M. Richard Eley mainly writes sci-fi, dabbles in literary and nonfiction from time to time, and is hard at work on a new novel.