As more jobs become automated over the years, it makes sense to ask the question: Will A.I and robots take over the investment-making decision process from humans? Will they necessarily do a better job? Will financial analysts and human investors become obsolete? Will there be a role for the human investment manager? What are the risks of leaving A.I to autonomously make all investment decisions?
These are tough questions to answer accurately, but we can make some informed predictions.
When we talk about A.I investing, we have to differentiate it from quantitative investing. Quantitative investing has more to do with investing based on certain rules which come about from using past data on stocks, then finding certain patterns and assuming those patterns will repeat in the future. The rules are stationary and decisions are predictable until someone changes the rules. A.I investing / machine learning, on the other hand, is dynamic. It continuously absorbs data, learns along the way, and looks for new patterns and trends.
In many ways, the trend toward applying artificial intelligence and machine learning to investing is well under way. Many firms are doing it and there’s even an ETF that investors can buy, which has been around about a year, called A.I. Powered Equity. (As of this writing, its 1 year performance of 9.2% is slightly below the S&P 500 ETF performance of 9.44%.)
Why is A.I and machine learning the new trend in investing? The competition in the investment industry is fierce. Markets are more efficient than ever and information is more public and transparent than ever. Thousands of research analysts pore over financial statements to find mispricings which are few and far between. Investment firms scour data from financial statements, social media, and even drones in order to find an edge in making investments decisions. Competition has driven down margins and profitability. Therefore, firms are betting on A.I and machine learning to make better predictions and avoid the mistakes that human investors make, thereby giving them an edge over the competition.
Also, the cost of computing power and cloud-based storage is falling dramatically as the years go by, so investing firms can now collect more data than ever. The idea is that the more data firms can collect, the more data they can feed into their A.I and machine learning programs. This will supposedly lead them to make better predictions, and make firms a ton of money. But is more data necessarily better? And does it lead to superior performance? It’s probably too early in the A.I investing revolution to answer this question, but let’s list some potential advantages and disadvantage of investing using artificial intelligence.
AI Investing Advantages
- Able to process thousands of variables in a short amount of time.
- More cost effective than having humans scour over data by hand and testing it.
- Can analyze mass amounts of data, finding patterns, correlations, and insights hidden to the human observer.
- Can continuously learn, improve, and teach itself as more data becomes available.
- Operates and makes investing decisions without emotion or human bias.
- A.I processor chips, data, and cloud-computing becoming cheaper over the years.
- More cost-effective; One A.I system can be more productive and cheaper than many analysts.
AI Investing Disadvantages
- A.I needs mass amounts of data continuously fed to it; Financial data is sometimes flawed, noisy, or incorrect. This can lead to bad investing decisions by A.I.
- When outliers occur, a.k.a. “black swan” events in markets, A.I might not know how to react to these events or unfamiliar inputs that haven’t happened before.
- A.I and machine learning assume that markets will act rationally and correct mispricings; Markets are made up of humans that may not be rational. Therefore, mispricings may last for a long time, reducing profitability of A.I investing systems.
- If A.I systems across firms start to resemble each other, they may start acting in similar ways, thereby eliminating trading edges, or causing waves of indiscriminate selling when certain signals or triggers are hit. This could cause a severe market crash and massive loss of investor capital. (See “Black Monday” of October 1987.)
- As an A.I system grows more complex, it makes it harder to determine what triggers its investing decisions without human guidance. This could lead to unwanted investment results that are based on a black box approach.
What to conclude from all this? Can artificial intelligence / machine learning outperform and completely replace a well-trained investor? Will all financial analysts just need to switch to other careers? Is the Ben Graham and Warren Buffet style / mantra of investing dead? Do we need to say hail to our new A.I investing overlords?
The answer is most likely this: A.I can never completely replace a human being in making investment decisions. In the best case, an A.I system will work with an investment firm, with someone guiding the system, but no sound firm would give up the reigns completely. Markets are simply too noisy. Humans are impossibly hard to predict. A.I is based on rationality and the hope that markets will react as they have in the past. Markets are much more complex than chess, backgammon, or even the ancient Chinese game Go. There’s a strict set of rules to these games, and a limited amount of combinations. There are no rules for how financial markets will react and the combinations are infinite.
Amazon.com’s A.I predictive analytics may be effective in predicting what you want to buy, but it’s much different than trying to predict the whims and reactions of millions of traders and investors around the world and how they will move prices. Humans are just too flawed to predict. How would A.I deal with subjective analysis such as whether a management team is effective, or a leveraged buyout, a hostile takeover, or a company spinoff? There’s just some events that need deeper subjective analysis by a human.
A.I and machine learning will be a valuable tool for investors and firms, and perhaps increase profitability, so the firms that don’t adopt it may be left behind. But man and machine will work hand in hand. A.I will provide the insights, and humans, based on years of experience, will still be pushing the buttons on investment decisions.