Tales from Tech Street: Booms and Busts in an AI Financial Takeover

By Yolanda Lannquist, Ethan Shaotran and Arohi Jain

 

The following are three fictional, short stories about AI & finance set in the future. 

 

Occupy ‘Tech Street’

Johnny was a Harvard graduate. He had a degree in Economics, and when it came to getting a job, he had no trouble finding one at Amerigo Securities, the business of his father’s close friend.

As an equity research analyst, his daily job consisted of looking at the performance of companies, industry trends, and compiling information into reports with his team. In this too, he had no trouble. And when it came to analyzing companies, there was no one better than him.

This was all before the ‘Tech Street Giants’ of the 2020s. On that fateful day in February, Goldman Sachs and Google co-announced the joint release of a new AI-powered investment advisory service, the Intelligent Trading and Analysis program, or ITA. ITA took the world by storm, and soon after, many other top financial services firms partnered with largest AI companies: JP Morgan with Baidu, Bank of America with IBM, and Charles Schwab with Microsoft.

These firms’ stock prices soared, and the market results backed it up. AI algorithms were more adept at collecting, analyzing, and finding trends among numerous data from across the economy, and autonomously trading with high profitability. There was no way a human could operate with such efficiency and compete at such a scale. Across all financial asset classes, AI machines were more capable at analyzing and drawing insights from more data than teams of humans ever could. Finance professionals, long comfortable in their uncontested roles, were witnessing the limits of their cognitive capabilities compared to fast, analytical, and unbiased machines that would never forget, never become tired, and never decide on impulse or ‘gut feelings’.

The media showcased the mergers, acquisitions, talent transfers and new in-house AI capabilities in the investments industry. Millions of Americans poured their money in hopes of growing their riches. And they did.

ITA and other big programs quickly weeded out smaller firms like Amerigo. Johnny felt tensions rapidly growing in the company, and surely enough, one by one, his co-workers were laid off. The fact that his boss was a family friend was the only reason he was still left, and even then, he was nervous he might be called in for a “talk” at any moment.

Programs like ITA, funded and developed by large corporations like Goldman Sachs, Baidu, and Google, dominated the financial markets with their popularity and success. These large banks enjoyed compounding big data, network and scale effects; already bolstered by brand names and media buzz, even more clients joined, adding to the stock of data on customer preferences, habits and account history. This data exhibited increasing returns to scale; it contributed to better predictability of algorithms, more adept at personalizing asset allocations for clients’ needs and maximizing their returns, eventually pushing the top firms over a ‘tipping point’ towards market domination and eliminating the competition.

Quietly, ITA and other programs continued to get better and smarter. Amerigo could no longer compete, and both Johnny and Amerigo recommended their last trade.

Photo by John Taggart/Bloomberg

 

 

Hacking the AI Race

A year had passed since the first ‘Tech Street Giants’ had appeared. Major investment houses around the world had adopted their own “smart” AI investment platforms, while the companies that were slow, unscrupulous, or hesitant in delegating responsibility to algorithms fell behind. Only a small handful of top investment firms remained, each with proprietary AI algorithms leading and executing virtually all aspects of investing, including research, sales & trading, investor relations, investment management, risk and tax optimization. These divisions reported to a ‘Master Algorithm’ set to optimize rewards for clients and firms’ proprietary funds.

The human executives still leading ITA started to notice a convergence in performance among a couple other top banks. It became hard to outperform markets; a buy or sell was similarly copied by the other firms. Their performance mirrored the others, with very little difference. While the top 3 ‘Tech Giants’, including ITA, were aware of their growing oligopoly, they also seemed to be almost perfectly interchangeable.

Antitrust regulators started to investigate these top three banks. Why were the banks performing so similarly? Were the algorithms similar? After all, top talent had interchanged among the firms. Were they trained on the same, equally accessible global economic data? The banks held firmly onto their IP claims, refusing to release information and open up their data sets and algorithms. Unable to uncover and audit the deep learning, ‘black box’ algorithms, the antitrust regulators had no evidence to back up their suspicions.

It was not unique to U.S. markets. In China, a few programmers at a top bank became suspicious of the algorithms themselves. Could they collude across banks when no one is watching? A few brave employees voiced concerns that the algorithms were colluding to help maximize each other’s rewards. But the Chinese government didn’t see a problem; there were still multiple players and their businesses were getting bigger and stronger, competing on a global stage. It was perfectly normal to share data and collaborate, especially when it was shared with the government as well.

With these ‘Master Algorithms’ colluding against the rest of the market, anyone who knew the future moves of this AI trading network could take part in the fun.

That’s where I come in. I’m no economist, but I realized that these algorithms communicated by making micro trades to indicate what they were about to do. Knowing the advance signals, I adjusted my trading framework to turn foreknowledge into profits. It’s a small enough amount that no one sees the difference, but over millions of transactions it adds up. Why work in data science when you can make more going rogue?

From my mansion here in Mauritius, it’s hard for them to find me. A strong computer, VPN and data package from my burner phone’s hotspot is all I need to get the job done.

Photo by Kevin Ku on Unsplash

 

A Mirage of Riches

Johnny lost his high-paying cushy job at Amerigo. He reluctantly joined his father’s small drone manufacturing company, which was losing market share to Amazon’s new in-house drone production.

His dad hired Johnny to keep track of the company’s financials, a rather menial and boring task. Several years passed, and Johnny wanted more from the company and soon came up with the perfect idea.

That night, Johnny requested to dial into his former co-worker Chad’s MindSpace, a new social media app for simple mind-to-mind communication. Chad was now working at ITA as a programmer for its Master Algorithm of Portfolio (MAP). In preparation, Johnny flipped through a handy ‘How to Communicate with Humans’ app, which ingeniously encodes and analyzes MindSpace friends’ past behaviors to learn what makes them tick.

Chad accepted his HeadSpace dial-in. Johnny persuaded him into giving him access to MAP’s source code. Johnny took the code, which contained the reinforcement learning algorithm, modeled to optimize returns in the stock market.

Johnny rewired the program to reward itself when his father’s drone company’s earnings went up. He connected the MAP algorithm to a popular data processing cloud system used by many banks, manufacturers, and retail stores, allowing it to slightly alter the data processed by these important economic agents.

The results were mixed at first. The learning algorithm didn’t know yet which factors led to his father’s drone company’s financial performance. The algorithm was experimenting, but the impact wasn’t visible in the real world or in his company’s morning virtual meetings. Business went on as usual for a few weeks, and among water cool banter and a new office crush, Johnny almost forgot about his tampering with MAP.

Then he started noticing volatility in energy prices in the newspaper, and a few drone hardware manufacturers closed shop. Despite efficient autonomous vehicle grids, he noticed news stories of traffic jams and a piece on commuting time hitting record highs. These made sense to Johnny; higher fuel prices and traffic contributed to rising demand for more efficient drone deliveries.

Investors were starting to notice and new companies propped up in the drone delivery market, tempted by rising profits. With a substantial market share already, his father’s company led ahead, and earnings, along with its stock price, rose to new highs! A month later, sales at the drone company increased twenty-fold. Johnny’s dad assumed his competent leadership had finally paid off, and announced an elaborate family vacation in Finland that summer to escape the heat.

It was a fantastic vacation. The Baltic Sea felt so real. Never was he more grateful for his high-definition haptic virtual reality suit.

Then his simulated global economic takeover became all too real. Johnny grew bored of the virtual reality and removed his goggles and haptic suit. He only desired to simulate how monopoly by his dad’s company would pan out. Instead of riches with lush paradise vacations, car races on Mars and the Ottoman Sultan’s harem, he got public protests and government investigations. The world he had simulated in VR began spiraling in unfathomable ways. Without competition, his company’s drones took flight faster than safety checks. ‘Rent a drone’ terrorist attacks were cheap, omnipresent hacking made privacy expensive under a sky of drones, and billions of small devices impacted the environment more than all the pre-hyperloop passenger jets. His haptic suit could really convey the constant humming and heat.

Games were so elaborate these days. Why bother escaping reality if the same rules apply in VR? He left the living room couch to go meet up with his GitHub buddies for beers. He could revert MAP’s code tomorrow.

Photo by Simon Zhu on Unsplash