The Reason AI-Managed Funds Are Missing Out On the AI Stock Boom

AI managed exchange-traded funds have, so far, under-performed those managed by humans. And that is due, in part, to the fact they missed the recent stock market rally driven by investor interest in AI.

The problem seems to be that because they have to be trained, the AI-driven funds respond too slowly to recent developments or the implications of future changes. The AI funds are prisoners of history without having yet developed the discernment of many human investors, even in AI-oriented venture funds, to identify and trade into new trends. That could change as the funds and those who program them gain more experience. Or not. JL

Bob Henderson reports in the Wall Street Journal:

13 exchange-traded funds have put AI in charge of managing their portfolios. Almost all have missed out on this year’s tech-led market rally and are lagging behind the S&P 500, a a sign of the technology’s limitations in stock picking. AI-powered ETFs exploit AI’s aptitude for finding patterns in data. A portfolio manager chooses a trading strategy. The AI is then trained on historical information and trades within that strategy.AI doesn’t adapt quickly enough to paradigm-shifting events to outperform professional portfolio managers reliably over time.  "For now AI is limited to plagiarizing history.”

 

It is early yet, but so far artificial intelligence doesn’t make much of a portfolio manager. 

At least 13 exchange-traded funds have put artificial-intelligence applications in charge of managing their portfolios. Almost all have missed out on this year’s tech-led market rally and are lagging behind benchmark indexes such as the S&P 500—an irony given the investor enthusiasm regarding AI subjects and likely a sign of the technology’s limitations in the ultracompetitive world of stock picking.

Take the WisdomTree U.S. AI Enhanced Value Fund, ticker AIVL, which has $385 million in assets. The fund has generated a total return of 2.2% this year, while an ETF tracking the Russell 1000 Value index, AIVL’s benchmark, has risen 4.5%.

The biggest reason for AIVL’s underperformance? Its AI’s refusal to buy Facebook parent and AI stalwart Meta Platforms—whose shares have soared more than 140% since December, in part driven by the firm’s investment in AI-related technologies.

“There was a really, really, really sharp price increase, and it’s just viewing it as overvalued,” said Chrissy Bargeron, client portfolio manager for equities at Voya Investment Management, which developed the WisdomTree fund’s algorithm. 

AI-powered ETFs, much like chatbots such as ChatGPT, make predictions by exploiting AI’s aptitude for finding patterns in mountains of data. A portfolio manager chooses a trading strategy. The AI is then trained on a vast quantity of historical information and directed to trade within that strategy by using the tactics that would have worked best.

Poor performance might be why parallel waves of enthusiasm for both AI-themed stocks and actively managed ETFs haven’t won investors over to the AI-powered club. The 13 members identified in a Wall Street Journal analysis have about $670 million under management—a minuscule sliver of the $7 trillion ETF market—and have seen outflows this year and last of more than $300 million.  

AI-powered funds promise to skirt the costs of human behavior and to bring hedge-fund-like technology to the masses. And AI can, in some cases, time stock trades better than individual investors, according to research by Eric Ghysels, a professor of economics and finance at the University of North Carolina, Chapel Hill. But AI doesn’t adapt quickly enough to paradigm-shifting events such as 9/11 or Russia’s invasion of Ukraine to outperform professional portfolio managers reliably over time, said Ghysels.

“Maybe one day it will, but for now AI is limited to plagiarizing history,” said Ghysels.

The human brains behind the bot-powered funds tend to blame their underperformance on the strategies with which they have saddled their AIs. AIVL is seeking value opportunities in a challenging environment for value stocks, said Voya’s Bargeron.

Fans say performance will improve over time because the AIs are continuously retrained with incoming data, though the evidence on that score isn’t entirely consistent with that idea. 

The oldest exchange-traded fund driven by artificial intelligence—known as the AI Powered Equity ETF, ticker AIEQ—was launched in 2017. It runs on International Business Machines’ Watson supercomputer and bases its bets on an analysis of millions of news articles, social-media posts, analyst reports and financial statements. It is up 9% this year, trailing the broader market largely because it spread its bullish bets across too many stocks. It missed out on the surge in the “Magnificent Seven,” a group of big tech stocks that powered markets in the first five months of 2023.

AIEQ beat the major indexes for its first few years, but then got clobbered in 2022 by the stock rout triggered by the Federal Reserve’s aggressive interest-rate hikes. AIEQ’s total return since inception is now about 44%, compared with 93% for the SPY ETF, which tracks the S&P 500 index, over the same period. 

“This is machine learning, not machine knowing,” said Art Amador, chief operating officer and co-founder of EquBot, the AI fintech firm that developed AIEQ’s technology in partnership with IBM. The next time the Fed does what it did last year, Amador said, AIEQ’s AI will say, “Boom! I recognize this!”

 

That learning curve presents a dilemma for investors, said Jack Butler, 60 years old, who runs an irrigation-products business in Encinitas, Calif. On the one hand, Butler said, it might make sense to get into AI-powered funds early, before they evolve enough to eliminate the market’s opportunities. On the other hand, the hallucinations and other errors made by programs such as ChatGPT are part of what is holding him back.

“I think mistakes are going to be made early on, and I don’t really want to be part of those mistakes,” he said.