So far, that promise has not been realized. The only major fund run by AI – AI Powered Equity ETF – has underperformed and attracted low investor interest. Since opening in October 2017 the fund has returned just under half as much as the Vanguard S&P 500 ETF, or 35.3% versus 70.8%, with slightly more volatility: 24.2% versus 21.5%. The AI ETF – ticker AIEQ – has a beta of 1 to the S&P 500 Index after fees, with an alpha of negative 3.82%, which means it has lost 3.82% to the S&P 500 each year on average over the last five years. However, that alpha is not statistically significant, meaning it’s plausible that AIEQ has a superior long-run expected risk-adjusted return but just had an unlucky five-year run.
Another prominent AI product is HSBC Holdings Plc’s AI Powered US Equity Index, or AIPEX. Since inception in August 2019, the index has returned only 2.3%, compared with 44.8% for the Vanguard 500 Index. However, AIPEX has a 6% annualized volatility target, about one-quarter of the S&P 500’s 24.3% over the same period. AIPEX has hit its volatility target almost exactly, 6.1%, and it has a beta of 0.19 to the S&P 500 and an alpha of negative 1.8% (and like AIEQ, that negative alpha is not statistically significant). AIPEX includes a 50-basis-point index fee and holds most of its hypothetical capital in cash. Adjusting for those two things, AIPEX’s pure stock selection — the measure of the success of its AI – -has lost 6.8% per year to the S&P 500 over the last three plus years.
Nevertheless, AI is making strong inroads in investment management. The main area is processing “unstructured data” like news stories and text reporting. There’s no doubt that AI trumps humans at this; it can read everything, in all languages, and distill the useful information. It can process pictures and anything else that can be converted to bytes in a computer file. The amount of such data is growing rapidly, and the sophistication of algorithms to process them, so AI will continue to advance in this task.
Another area in which AI and ML have been used widely is trading algorithms — not deciding what to buy and sell, but choosing how to break up orders and feed them into a variety of trading platforms. These algorithms don’t have to be very smart, their main advantage over humans is speed. They can monitor hundreds of pricing data feeds continuously and make instant decisions.
But these ancillary functions were not what AI pioneers dreamed about. They believed AI could take over the entire investment decision process, and not just create signals and execute trades, but also interpret those signals and choose which trades to execute. Bryan Kelly, head of AI research at AQR Capital Management LLC (where I once worked), puts it this way:
“Machine learning has a real impact on systematic investment processes because it allows managers to metabolize information from more new sources faster, and in more expressive ways (due to greater model flexibility). But it’s important to remember that the central motivation of machine learning—squeezing as much usable information as possible out of data—has long been the modus operandi of quant investing, so I see ML as one further step in the evolution of quant investment methods.”
I think this represents the mainstream belief at the moment. AI is slowly being integrated into quantitative investing, particularly for signal extraction and trading, but is enhancing human research and decision making rather than replacing them.
There are two areas of hope for a larger ML role in investment management. The first is the “L” in ML. Each day of underperformance is another opportunity to improve. Perhaps ML is like a baby bird just finding its wings and will someday soar far above Earthbound humans. The second is that institutional investors are getting interested in using ML for asset allocation rather than security selection. Cross-market optimization is far more difficult than picking portfolios within asset classes. Most investors don’t even attempt it, they instead build the best stock, bond and commodity portfolios they can, and so on, and then combine them according to pre-selected allocations. AI is the only known approach to constructing a true global portfolio.
Investors should forget looking for a Skynet or Hal 9000 to run their money at the moment. The best firms are using ML where it has been proven to work—and perhaps thinking about other applications—but pure ML decision-making has lagged the market.
More from Bloomberg Opinion:
• Creative AI Is Generating Some Messy Problems: Parmy Olson
• AI Can Help Make Cryptocurrency Safer for Everyone: Tyler Cowen
• Our Future AI Overlords Need a Resistance Movement: Parmy Olson
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Aaron Brown is a former managing director and head of financial market research at AQR Capital Management. He is author of “The Poker Face of Wall Street.” He may have a stake in the areas he writes about.
More stories like this are available on bloomberg.com/opinion
