The crowded and often frenzied trading floors that were once an integral part of capital markets in the last century are now gone; they’ve been replaced, for the most part, by machines that make decisions in fractions of a second. In the U.S., 80% of all stock trades are reportedly machine-driven; however, this near-total dependence on machines can open the door to new risks and errors.
To be clear, a heavy dependence on computers is not new to the financial sector. Banks and other financial institutions have adopted a number of computer technologies; they’ve used automation to speed manual processes while reducing costs. Today, high-frequency trading technology makes it possible to bypass humans in the investing process. This practice is often referred to as algo trading, a reference to the mathematical algorithms that drive these trading decisions. By allowing users to take advantage of instantaneous profit opportunities, algo trading can help traders beat their competitors to the punch. While using machines to execute trades based on algorithms approved by humans makes perfect sense in theory, the history of algo trading suggests we should use caution in adopting human-free decision making.
The Risk Of Instability
The infamous flash crash of 2010, a trillion-dollar stock market plummet that unfolded in just minutes, is a prime example of the risks associated with algo trading. While machine trading was not necessarily the cause of that sell-off, it certainly aided and abetted the crisis. In this case, during a stock slowdown, machines believed the best bid for some stocks was $0 and made bids as low as $0.01, with disastrous results. Some companies’ stock prices dropped as much as 37% as a result.
There’s no question to me that algo trading has a place in our global economy. That said, there are also certainly risks involved when humans leave the picture, and even more so once we hand the reins over to AI.
From Algorithms To AI
It’s important to understand the key differences between algo trading and AI trading. In algo trading, algorithms are based on predetermined data and relationships. In other words, a known set of conditions (i.e., data points) triggers specific actions, like buying, selling, shorting and so on. With AI trading, computers can analyze huge amounts of data and determine which conditions should trigger certain actions, with no human assistance.
As with most AI-based systems, these conditions could potentially include not only financial data, but also the tone of media coverage, social media chatter and some factors that may seem strange and totally irrelevant. For instance, AI-based credit scoring includes the ratio of smartphone photos taken with the front-facing versus the back-facing camera. In fact, AI algorithms can be so complex that many managers, regulators and investors may have difficulty understanding how they work. Even developers may be unable to explain how an algorithm obtains a particular result.
Despite this so-called “interpretability” problem, I haven’t seen enthusiasm among investors dampen over the past decade. As with algo trading, however, I believe AI has yet to become investing’s silver bullet.
The Machines Are Here To Stay
In spite of these problems, there is no question in my mind that algo and AI trading have transformed the financial sector and will surely grow in importance. Just like the successful programs humans have developed to win at chess and poker, AI trading will likely become more successful and accurate over time. For this reason, I believe it’s imperative that financial companies institute training programs to help their employees work with AI.
The good news, according to a recent survey by a customer experience company, is that over 70% of U.S. employees hold a positive outlook on workplace technologies that use AI. These employees should receive the training they need so that they can use AI, in trading and other applications, to enhance their efforts and make them more productive. This doesn’t necessarily involve learning how to develop algorithms, but it does mean gaining a high-level understanding of key capabilities of data science (classification, clustering, regression analysis and optimization), and how they can be leveraged for business and financial value. To put it simply, employees should understand enough about data science to effectively work with actual data scientists and make use of their capabilities.
In this new era of AI adoption, the time has now come to consider how humans can be reintroduced into the trading process to act as safety valves. Given the already-complicated process of dealing with regulators, investors and decision-makers, I believe adding AI into the mix requires a level of familiarity that most managers lack. Most importantly, human experts on the ground with AI-aided trading technologies could minimize those catastrophic losses we’ve seen before. This means that humans should learn to both smartly coexist with and guide AI toward the future of trading.
By Mohit Joshi
The race to net-zero emissions will forever change the way many companies do business. The immediacy, pace, and extent of change are still widely underestimated. Early movers can seize significant advantage. In this report, coauthored with the WEF Alliance of CEO Climate Leaders, authors explore how other companies can take a similar path by identifying, creating, and scaling green businesses.
The current debate over ESG and sustainable investing is noisy and sometimes rancorous, and the temptation is strong to just tune it out until it’s better resolved. But, in the end, leaders must resist this urge and accept that it’s a relevant discussion.
Looking at today’s consumption levels, sustaining our current growth trajectory would require the ecological resources of 2.3 planets by 2050. This number is significantly higher for mature markets. The US, for example, would need five planets to sustain present-day consumption levels; Germany would need three.