Artificial intelligence(AI) is called to revolutionize every industry in the next few years. However, there is an industry that has been actively relying on AI for decades and that can provide a lot of inspirations when comes to adopting AI technologies: we are talking about Wall Street. Relying is the imperative world in the previous sentence, Wall Street doesn’t only uses AI but it actively relies on it for some of its fundamental business functions.
From the emergence of quantities trading to the popular raise on high frequency trading passing through many high profile episodes, AI has evolved to become an essential component of Wall Street’s operations. In that sense, there are many lessons that other industries can learn from Wall Street’s journey through the adoption of AI technologies. Here are some of my favorite lessons:
1 — AI Can Get Out of Control
The last two decades of Wall Street are full of stories of AI-based quantities trading systems spinning out of control and causing major losses in the market. Remember Long Term Capital Management(LTCM)? The quantities trading hedge fund was started by academic luminaries and future Nobel prize winners promoting the ideas of the Efficient Market Hypothesis (EMH). LTCM relied on AI, quantities trading algorithms to identify and execute trading stragies based on EMH-based arbitrage models. After a few successful years, LTCM AI systems started placing questionable bets on markets such as Russia and Japan ended up literally wiping out the fund’s assets and threaten to bring down the entire US financial system.
LTCM is not the only example of scary episodes caused by AI systems in Wall Street. Famous quantities trading houses such as Ken Griffin’s Citadel have been on the brink of failure more than once dues to AI-drive strategies spinning out of control. Those episodes highlight the importance of the continuous training and monitoring of AI systems.
2 — AI Regulation and Governance are Essential
Last year, the US Government published several different reports about the role of goverment in the adoption of AI technologies. From Wall Street’s experience, we know that compliance and regulation are extremely important to guide the sane adoption of AI platforms on mission critical solutions.
3 — Advanced Statistics can Carry Us a Long Way
The industry is obsessed with machine learning(ML) and AI technologies. However, from Wall Street trajectory, we can learn that advanced statistics can provide a tremendous amount of value across different industries. From the early days of quantities trading based on Ed Thorpe (Beat the Dealer, Beat the Market, Man of all Markets…) to current high frequency trading models, advanced statistics have been an efficient stepping stope towards the adoption of AI technologies.
4 — The Battle of AI Strategies
Sometimes, I feel that some industries are mistakenly looking at AI strategies as a zero-sum-game. We often see vendors promoting THE AI STRATEGY for sales forecasting or marketing optimization. If we look at Wall Street, however, we can learn that industries are going to produce many and often competitive AI strategies to solve the same problem. In Wall Street, the competition between different AI strategies has been a vehicle to improve the efficient of different market algorithms.
5 — Cruise Control AI
Wall Street has become a master at leveraging AI strategies in an autonomous fashion for long periods of time. More importantly, Wall Street AI strategies are regularly making independent decisions that are immediately translated into financial gains and losses. For instance, if an arbitrage AI strategy infers that a specific market is overvalued, it can make the decision of shorting a number of stocks in that market. If the market continues trending upwards, covering those shorts will immediately reflect on a financial loss that is only justified by the trust on the underlying intelligence of the AI model. That’s a lot of trust to place in an algorithms. Similarly to Wall Street, as AI evolves on other industries, we are likely to see more cases of this semi-autonomous, “cruise-control” AI.