Chasm Waxing: A Startup, Cyber-Thriller

by B. Michaels

  “How was that? I feel like I owe him a royalty for the idea of using Bibles to train the deep learning algorithm.” Josh’s tone was noticeably rosier.

  “You don’t owe him a royalty, silly. He didn’t do anything. You owe me a royalty; it was my idea to use them for your ground truth.” She giggled. “He’s great. But a little of my dad goes a long way. Somebody always ends up healed, except my mom...” Becca’s voice trailed off.

  “What?” asked Josh.

  “Never mind.”

  Becca and Josh continued chatting for a good portion of Becca’s trip. She got back to the office it was a little before 3:00 p.m. It was odd that Samantha, Ali, and Saul were all gone. Nobody left early from Gamification Systems. Especially, since everyone was preparing for next week’s Gecko demo.

  Becca sat in her chair, opened up her Mac laptop, and began to work on a section of source code that dealt with non-player characters. Source code was the actual collection of instructions written in a distinct programming language. Because Becca’s code interacted with the Unreal Engine, her source code was written in C++. She wanted to make sure this section of code was rock solid, just in case the need arose for the G-Master to spawn a character in the Gecko demo.

  To perform her work, Becca needed to inspect some of Saul’s G-Bridge code. Her code sent messages to Saul’s code. Becca browsed to Gamification’s GitHub repository. She located the right module of Saul’s code. Unexpectedly, she saw many changes in the various files that comprised the module. Saul updated the files—just five minutes ago. Becca carefully examined the modified files. She ran a diff command to see the exact changes. The only thing different about the code was newly added comments.

  In source code, comments were only meant to make the code more understandable to humans. The computer completely overlooked the comments when it compiled or executed the code. The same comment was attached to a number of method declarations—‘Make sure works with Velocity.’ An odd feeling enveloped Becca.

  Out of the corner of her eye, she saw Ali. As Ali moved past her open office door, she greeted him. Ali looked as if he’d just met the President. “Did you see the guys from Velocity Games? They’re here to meet with Samantha and Saul. I shook their hands in the reception area, when I was running to my car. John Vincent and Corey Lawson! Those guys are the gods of gaming.”

  Vincent and Lawson were the two Co-Founders of Velocity Gaming Studios, located in Bethesda, Maryland. Velocity was the first gaming company to fully incorporate all aspects of virtual reality gaming. Other companies simply ported over prior games to VR worlds. Velocity built VR-optimized games—from the ground up. Fog of War had broken all the video game sales records and won numerous awards.

  “No,” replied Becca. “Where are they meeting? Why aren’t they meeting in our conference room?”

  “I don’t know, when I got back, they were gone. Obviously, they’re meeting somewhere in this building. I do know that I want to play me some Fog of War tonight.” Ali entered his office. As he took his seat, he looked at Becca. Then, he pretended to shoot at her, with the rat-a-tat-tat of an imagined machine gun.

  “Great,” said Becca, loudly. “A software engineer went postal today after he met the Co-Founders of Velocity Studios. No one knows if the two incidents are related. By the way, let’s run through the demo tomorrow morning. I want to double-check that your code for spear phishing is working.”

  “Ok.” A few minutes later, Ali cursed loudly.

  “What’s the matter?” yelled Becca.

  “None of my music works in A-Tunes. Nucleus’ last update applied the music Blockchain to all of my songs. It broke everything.”

  Becca chortled and walked to Ali. “Is it possible that you own pirated music?”

  “Whatever,” said Ali, in a snit.

  “Music labels and movie studios are using the Blockchain to stop piracy,” Becca said. “They love the Blockchain because it allows them to control music distribution and royalties with smart contracts.

  “You can’t play your music unless A-Tunes locates a record of your purchase on the Blockchain. You can still use streaming services, like A-Radio, but now the hottest music is only available for purchase via the Blockchain. After they’re old, the albums move to the streaming services.”

  “Thanks for the tech update, Professor Buzzkill.”

  Becca continued, “I just read an article detailing how artists love the Blockchain too. Not only do they thwart pirates, but they can also control who listens to their intellectual property. And they can more easily pay creators and producers. If the song changes by one note—with remixes for example—it’s a new creation on the Blockchain. Everything starts over for that song.”

  Ali made a face. “You’re full of good news. Becca, you’re a hacker. Hack A-Tunes and make my music work.”

  Becca laughed, “I’ve already spent plenty of time with the FBI.” She returned to her office. 90 minutes later, Becca saw Samantha Powers and Saul Abrams.

  Samantha stopped outside of Becca’s office. “How’s it going?”

  “Good,” said Becca. “I feel excellent about next week’s demo for Gecko. I’ve got to talk to Saul about a couple of things, but everything is working. And Ali tells he’s 100% sure of his code changes. We’re going to run through the demo tomorrow morning.”

  “Ok,” replied Samantha. “Why don’t you and Ali run through the demo with me tomorrow afternoon. I want to see it work for myself. Will you also look at Lou’s PowerPoint for Gecko? I want you to do a sanity check on the technical details. We’re going to start with the PowerPoint, you’re going to do the demo, and then finish back up with the PowerPoint.” Lou Skaist was Gamification’s Vice President of Sales.

  “Sure,” replied Becca.

  “How’s the CyberAI stuff coming?”

  “Josh just told me it’s going well. Once we get this demo done, I can start to work with him more closely. I’m going to have to get Saul more involved.”

  Samantha nodded her approval and began to unlock her office door.

  “Why was Velocity here?” asked Becca. Becca was sure Samantha heard her. The CEO breezed into her office and shut the door. A few moments later, Becca’s phone buzzed with a text from Josh.

  ‘Are you available tonight???’ Becca replied, ‘tnite not best…who about 2moro?’ Then, noticing the misspelling, she texted, ‘grrr—how, not who! ’

  Chapter 13 – TextWorld

  6:15 p.m. (EDT), Friday, July 31, 2020 – North Laurel, MD

  Josh Adler’s Apartment

  Becca thanked her Uber driver and knocked on Josh’s door. She imagined the Founder and CEO of CyberAI lived in something bigger. Maybe the startup life was not as glamorous as she thought.

  “How are you,” said Josh. “Wow, you should wear a sundress more often!”

  “It’s better than a flannel?”

  “Much,” Josh replied, with a broad smile and a wink.

  “Thanks for being so understanding and stopping by to see the progress I’m making. I haven’t left this apartment since Tuesday. I want to knock this out as quickly as possible, so I can show you a proper first date.”

  Josh Adler’s three bedroom apartment was more like a computer lab, with a kitchen and couch. Flashing green, blue, and red lights flickered everywhere among the line of rackmount servers. Wires ran from all angles. Whiteboards, chock-full of diagrams and mathematical equations, adorned every wall. There wasn’t a personal picture or memento to be found.

  “Why isn’t all this stuff at your office?”

  “My dad gave me most of this equipment—when we were still talking—and I didn’t want to make it a part of the company. Sometimes, when I’m in a zone, I need to be in my own place, with my own stuff.”

  Josh brought Becca to his central work area. Becca scooted a chair next to him. “Look at this. Becca, the computer is learning! I�
��ve trained the neural network on Bibles and commentaries. I’ve also included some ancient texts, like Philo and Josephus.”

  Josh pointed to the two NVIDIA DGX-1s in the server rack. “These boxes processed the text and created the neural network, using my deep learning algorithm. Because the math of deep learning is relatively straightforward, GPUs are much faster at creating neural networks than CPUs. My neural network is still a work in progress. It’s not perfect. I want to ask General Shields if I can get time on NSA supercomputers. But the results are still astonishing.” Josh maximized a window on his computer screen. The monitor displayed a bar chart.

  “I’ve gone from 83% recognition of cyber-events to 91.5%. And I’ve only been working on this for a week! My last demo for General Shields was such a disaster, because month-over-month, the results improved less than one percent. Now I can show him substantial progress. And it’s all due to you.”

  Becca smiled. “Josh, you did this. Not me. These results are epic. You’re getting close to your 10X metric of 95%.”

  “Yep. Thanks, Becca. I meant that without you, I would never have selected the training material that I used. I might have picked Shakespeare or Plato, but definitely not the Bible. Anyway, this is only the half of it. Look at this—actually, it’s cooler in VR. Hold on.”

  Josh grabbed two VR head-mounts and four hand-held wireless controllers. They both put on the VR goggles and gripped their controllers. Josh properly positioned themselves among the room’s external VR cameras and sensors. “You remember that the CyberAI software scans the web and social media for cyber-threats? Then it incorporates that information into our predictive modeling?”

  “Yes,” replied Becca.

  “The bar chart you just saw included the data from the Internet.”

  “Ok, I think I’m following you. Go on.”

  “I was curious to see if the neural network could determine meaning beyond cybersecurity. So I relaxed the parameters of the search, and told my query engine to ingest everything—no parameters.”

  Josh clicked a button on his VR controller. Becca and Josh’s immersive VR world sprang to life. The VR headset completely blocked out the real world. In front of them, a large word cloud of variably sized and differently colored words appeared. Some words included, ‘#,’ to denote that they were hashtags.

  “Welcome to TextWorld,” said Josh. “In TextWorld, these are the most important English words and categories for the last 24 hours. My neural network produced this list.”

  “You mean these are the trending words?” asked Becca.

  “Exactly.”

  In the virtual reality space of TextWorld, Josh stepped forward and grabbed the word, ‘Sports.’ Becca watched the virtual Josh move in front of her. Her eyes traced the whole of his virtual body. As Josh reached out for Sports, the old word cloud dissolved, and a new word cloud appeared. It contained a list of sports related words and topics. “Let’s see what the computer predicts about—”

  Becca burst forward and grasped, ‘Football.’ “I love football!”

  The word cloud now entirely consisted of American football terms. Becca selected, ‘Cowboys.’

  “Eww!” said Josh, a diehard New York Giants fan.

  Now, a three column grid appeared in front of Becca and Josh. From left to right, each column contained the labels of; ‘Past,’ ‘Present,’ and ‘Future.’ Under the labels, in each column, were minimized but readable web pages—individual articles.

  As Becca looked further left, in the Past column, she could see older articles related to the Dallas Cowboys. In the center, Present column, she saw current posts. Most covered the Cowboys training camp. It was occurring this month in Oxnard, California.

  “So, you’ve created a Virtual Search application?” asked Becca.

  “Sort of. But look in the Future column.” The Future column resembled the first word cloud. It was much less content rich; there were no pictures or videos. The Future column only contained words and text files with bold-type headings.

  Josh clutched the word, ‘Predicted Record,’ from the Future column. Everything dissolved. Future predictions related to the Dallas Cowboys appeared. “Based on information available right now, my AI is predicting that the Cowboys will finish with 11 wins and five losses. It also sees them winning the division title. Dang it! The Giants need to get rid of that stinking coach.

  “TextWorld, display current Vegas odds for 2020-2021 NFL season,” ordered Josh. A web page from the Washington Post appeared in TextWorld. “Look, according to today’s Vegas odds; the Cowboys are not favored to win the NFL Eastern Division. So, if I trusted the predictive analytics of my deep learning algorithm, I’d make this bet. Vegas odd’s makers are not bullish enough on the Cowboys. I can make money.”

  “Wow, go Cowboys!” said Becca. Josh groaned. “The question is, do you trust your deep learning algorithm?”

  Josh smiled from ear to ear. Those dimples, thought Becca, even in VR.

  “That’s where this gets mind-boggling,” said Josh, excitedly. “TextWorld, display file backtest.”

  Becca now saw a massive grid. It looked like an enormous spreadsheet. Her eyes first focused on the multiple different colored numbers—there were black, green, and red numerals. Then she widened the focus of her perspective. With her larger point of view, Becca noticed that everything fit within three large rows, spread across her entire field of vision.

  To their far left, each row began with a year-to-date price chart. From top to bottom, the labels were, ‘DXY,’ ‘WTI,’ and ‘FB.’

  Josh explained, “You’re looking at the results of a backtest I ran last night. These charts here…” He stepped closer to the three, price-time charts. The DXY chart was in the row closest to his head, the WTI chart was close to his chest, and the FB chart was in the row near his knees. Becca followed.

  Josh continued, “These charts record the closing prices of the US dollar index, West Texas Intermediate crude oil, and Facebook. All of them begin on January 1, 2020, and run through today. Looking at the charts; you can see that the dollar has gone up, oil has fallen, and Facebook’s stock price has been up and down. Recently, you can see that it hit new, year-to-date highs.”

  Becca examined the charts meticulously. Each price-time chart contained a line that touched the closing price for every day. There were other lines as well. She could make out the 50 and 200-day moving averages and key Fibonacci levels.

  “I backtested my deep learning algorithm on these three financial instruments—a currency, a commodity, and a stock. A backtest calculates what the algorithm would’ve predicted. Then it compares the prediction to what actually happened. If the predictions are right, you know the algorithm is working. Or, at least you know the algorithm would’ve worked in the past.”

  Josh placed his hand on the FB row. He started walking to the right. There were tons of numbers. “Every trading day has a meta-column. Josh shuffled to February 14, 2020.

  Becca followed. “Aww, you walked right into Valentine’s day with me.”

  Josh flashed his virtual smile again. “My VR-self is smooth like that,” he snickered. “The February 14 meta-column has three different numbers. In this sub-column, are the actual closing prices. This sub-column contains the price that the AI would have predicted. And this sub-column records the daily variance number.

  “Now, what you’d like to see is as little variance as possible. Ideally, you wouldn’t see any green or red differences. Everything would be a black, ‘zero.’ If there were no variances, you’d be a god. All your predictions would be coming exactly true. Now, come down here.”

  They moved to the meta-column dated with today’s date, July 31, 2020. After that, the meta-column only recorded predicted values. “My backtest shows that the deep learning algorithm would have been very useful in predicting closing prices for the next day, especially as
time advanced.”

  Becca looked to her left. She noticed that while there were more red and green variances in January; each month, more black zeros were visible. July contained the most black zeroes. “This month was the best month so far. Even where there are red or green numbers, the variances are decreasing.”

  “That’s precisely right. The algorithm is learning. It’s getting better and better at predicting the closing prices. My dad would kill for this software.”

  “I guess I should invest in Facebook stock,” said Becca.

  “Yes. Not only could you invest in the stock, but you could make a fortune in options. The algorithm gives you a high degree of confidence in tomorrow’s closing price.”

  “So, you are going to quit your job as CEO and trade on Wall Street?” asked Becca—half serious.

  “No, my dad has done that forever. I admire him. He’s a billionaire. That’s awesome. But I want to change the world. This is exactly what I want my algorithms to enable—discovery. I don’t want to watch closing prices and charts for a living. I want to create ever more valuable closing prices for my publicly traded AI corporation.

  “I want my company to predict your future interests. It could be something like a movie, a pair of shoes, your future husband—anything. Your discovery service will run in the background and be an extension of you—your extended intelligence.

  “It could do your chores. It could execute smart contracts for you on the Blockchain. It could purchase your groceries, because it talks to your refrigerator and knows you’re low on eggs. The possibilities are endless. But the bottom line is that you would do a lot less searching and a lot more discovering. The more the AI knows about you, the better predictions your discovery service would render. Advertisers would love it. Some would even just target ads to the discovery service.”

  “I don’t understand how it’s predicting the future with such a high degree of accuracy?” asked Becca. “It only has information that is available to us. I mean, you didn’t program it to figure out tomorrow’s closing price of Facebook.”

  “That’s just it, Becca. I don’t know how it’s doing it either. I didn’t program an algorithm to bet on football or determine closing prices. I just used a deep learning algorithm to create a neural network. That neural network was designed to understand and learn from English text. I can’t predict what the AI is going to learn. Deep learning offers a new way to program computers. My only job is to make the deep learning algorithms understand text better.”