In a downtown Toronto skyscraper a block from the Hockey Hall of Fame, a small hedge fund is hoping to find an advantage in the financial markets. Castle Ridge Asset Management is betting on Wallace, a specialized supercomputer that powers hedge funds' trading strategies using artificial intelligence.
Hedge fund insiders have long wondered if AI could help them beat the market, but the AI trading efforts they've launched have often been disappointing, and marketing plans to attract client money have been slow. It was nothing more than However, his launch of ChatGPT in November 2022 reinvigorated this new breed of AI-driven hedge fund players.
Founded in 2015 by current CEO Adrienne de Valois Franklin, Castle Ridge is a relatively small company in the hedge fund world, with about $190 million in assets under management and little in the way of market creation. We are developing a business in an unknown town. Defeat the hedge fund. Still, Valois-Franklin believes the investment fund's approach of using AI to predict movements in financial markets could make it a dominant player in the multitrillion-dollar hedge fund industry.
Valois-Franklin, a former investment banker with little experience in quantitative trading, said Wallace's main selling point over rival AI-powered hedge funds is that it uses an evolutionary process likened to breeding to create a proprietary model. He argued that the ability to continually improve Speaking to MarketWatch, Valois-Franklin described Wallace as if he were a multi-manager hedge fund whose hypothetical portfolio managers are constantly “battling each other to see who is best suited to this environment.” expressed in. But the hedge fund chief points out that, unlike human portfolio managers, Wallace doesn't need sleep or pep talks.
Simply put, the Wallace Evolution process involves machines creating thousands of differently weighted virtual investment portfolios every day, which are then tested and ranked according to their suitability to current market conditions, Valois-Franklin said. Stated. Wallace picks the best-performing portfolios in eight-hour cycles, prioritizes them, and “breeds” them.
“Every day, Wallace creates thousands of copies of itself, each becoming a virtual portfolio manager with different characteristics. By toggling specific weightings and patterns up or down, or on or off, each portfolio manager is better or worse suited to the market environment. The better it is, the more likely that portfolio manager will breed,” Valois-Franklin said.
“It's like a flock of birds.”
Castleridge has had some success in generating investment returns. Since Wallace's founding in 2017, the investment fund has generated 12.4% annualized net returns compared to his SPX on the S&P 500.
The stock returned 12.1% in the same period, according to documents reviewed by MarketWatch. It will go up against quantitative hedge funds with vast resources, such as Two Sigma and DE Shaw, which are looking to expand into machine learning and AI.
Alex Bogdan, Castleridge's chief scientific officer, told MarketWatch that Wallace's evolutionary approach compares to neural networks used in systems such as ChatGPT, which are modeled after the human brain. He argued that it allows for a deeper level of understanding.
In Bogdan's view, these evolutionary processes represent the future of AI, allowing machines to go beyond simple imitation. Bogdan said neural networks, most notably in the form of large-scale language models (LLMs) like OpenAI's Chat GPT, simply “mimic the responses humans would make when given the same input.” I explained that it is only.
In contrast, Wallace's “genetic algorithm” works by combining its individual knowledge to build its own understanding, “getting smarter over time.” “GPT is a smart algorithm,” Bogdan says. “We have had enough of imitation. What we need is understanding, not cleverness.”
Early research into AI first began in the mid-20th century, driven by advances in computer science during World War II. In a 1961 experiment, British scientist Donald Mitchie succeeded in developing a machine made of matchboxes that could solve the game of “fives and crosses” and compete against human opponents.
Mitchie's machine, called the Matchbox Educible Note-and-Cross Engine (Menace for short), uses matchboxes to represent all 304 states of the tic-tac-toe game, and each small box has a relative advantage. It contained the beads shown. each position.
The matchbox machine moves hands based on the number of beads in each box until it finally solves the problem, with each winning hand awarded a bead and each losing hand penalized with the removal of a bead. It was a system where you could A simple paper and pen game. The system Wallace uses is based on a subfield of his AI called “evolutionary computing,” which seeks to solve complex problems using continuously adapting algorithms.
Similar to Mitchie's machine, Wallace plans scenarios, selects the most successful ones, and consolidates those winning strategies. But unlike Mitchie's Matchbox Engine, Wallace operates in a complex world of financial markets where parameters are constantly changing.
Castle Ridge's success is based on the AI machine's ability to adapt to changing market conditions. So unlike Mitchie's Matchbox Machine, which quickly solved the game it was designed to play, Wallace must constantly adapt his strategy.
“The system is not trying to determine what will happen in the market. It is trying to predict how players in the market will react every time news happens. From that perspective, the system is not trying to determine what will happen in the market. They're less interested in the basics of the cards they're dealt and more interested in what the other players at the table are saying,” Valois-Franklin said.
As a byproduct of this strategy, Castle Ridge said Wallace was able to predict a series of market events in advance of their official announcements, based on signals from the data it analyzed.
Valois-Franklin explains that Wallace views the market “like a flock of birds, ever-changing and changing,” capturing early signs of movement driven by insider knowledge.
These predictions also include Wallace's bet on Gilead Sciences.
The company pursued the acquisition of New Jersey biotech company Immunomedics in September 2020, and the cancer treatment company's stock price soared more than 100% after the deal was announced.
“As soon as an individual security starts to separate from the herd, that's one of the signals for Wallace to say, 'Look, why is this security acting independently relative to other securities?'” Independence of action often indicates knowledge imprinted in security. Often when people don't know something, they tend to act in lockstep with others. ”
Hedge fund staff are now spending their time “disrupting” Mr. Wallace's system by injecting “unknown unknowns” and feeding the AI with new data. In one case, the team provided Mr. Wallace with satellite images of Walmart.
In the parking lot, see if that information can help the machine predict consumer behavior.
In Valois-Franklin's view, this type of work may soon take up a large portion of the workday for people working at the world's top hedge funds. “It will replace some types of jobs, but it will expand production capacity in certain areas. I’m doing other things.”