It turns out that even artificial intelligence can make bone-headed moves.
Financial services company EquBot used IBM Watson, perhaps the most well-known artificial intelligence (AI) system in the world, to analyze thousands of companies on a daily basis to try to pick stocks that can beat the market. The AI system's recommended stocks are included in an exchange-traded fund (ETF), the AI Powered Equity ETF (AIEQ 1.38%) (AIEQ).
I wrote in January 2018 about the AI-powered ETF's top five holdings. At the time, the AIEQ had underperformed the S&P 500 since the ETF began operations in mid-2017. How did the AI system's picks fare against the S&P 500 last year? Not that great.
The S&P 500 index trounced the AIEQ, losing 6% compared to the ETF's 2018 loss of nearly 16%. The AIEQ didn't look quite as bad including dividends but still trailed the S&P 500 in total return with a 7.7% loss compared to a 4.6% loss.
I decided to dig into the AI ETF's stock holdings last year to try to figure out why it performed so miserably. I uncovered some important lessons that all investors can learn from the AI system's mistakes.
Image source: Getty Images.
What I found
The holdings owned by the AIEQ are available on the ETF's website. However, you can only view the current stocks owned and not historical changes in its positions. As luck would have it, though, I had downloaded the AIEQ holdings on Jan. 16, 2018, and again on May 15, 2018. I was able to compare the detailed lists against the stocks in the AIEQ as of Jan. 2, 2009.
Back in January 2018, the ETF held positions in 70 stocks. Fast forward to January 2019 and only 16 of those original 70 stocks were still held by the AIEQ. And it's actually even worse: Two of the initial stocks were sold during the year and bought back later.
Interestingly, of the top 10 AIEQ holdings from January 2018, only two remained in the ETF's top 10 as of January 2019: Alphabet (GOOG 0.98%) (GOOGL 1.04%) and Amazon (AMZN 2.47%). Amazon was an especially good stock to hold onto, as it delivered a 28% gain last year. Alphabet ended 2018 with a small 1% decline, but that was much better than the S&P 500 index's performance.
But I noticed a small issue. While Amazon was still one of the AIEQ's top holdings throughout 2018 (at least based on my snapshots of the ETF's holdings), the ETF owned nearly twice as many shares in early 2018 as it did by early 2019. The AI system actually sold a big chunk of one of its biggest winners of the year. Its position in Alphabet increased, however.
Also, the AIEQ completely sold 6 of its top 10 holdings from January 2018. As it turned out, two of those six went on to outperform the S&P 500, and a third stock beat the AIEQ's performance for the year.
Big mistakes investors make
Investing legend Jack Bogle once said that investors' two greatest enemies are "expenses and emotions." I don't think the AI system that powers the AIEQ has emotions, but it appears to trade more frequently than a highly emotional human investor would. And frequent trading definitely drives expenses higher.
My Motley Fool colleague Matt Frankel was asked recently what was the worst investing mistake he'd ever made. Matt's response was that he sold a winning investment too early. That's my worst investing mistake, too. Based on what I saw from the AIEQ's dramatic changes to its portfolio, the AI system succumbed to the same mistake.
I would even tweak my response a little to the question that Matt was asked. My worst mistake is selling too early, period. I continue to be awed by David and Tom Gardner, founders of The Motley Fool. Both hold on to their winning stocks even when they're not winning.
David, for example, has owned Amazon since the 1990s. But between December 1999 and September 2001, Amazon lost more than 90% of its market cap. And David Gardner held on to his shares. Why? He bought a company that he believed had great long-term prospects. Those prospects didn't change, even though Amazon's share price did.
This underscores what I think are the greatest investing mistakes for the AI system powering the AIEQ. The artificial intelligence system appears to be buying stocks rather than businesses. And it isn't taking a long-term view. If it were, it definitely wouldn't have such a ridiculously high amount of turnover in its portfolio.
Different lessons from an AI system
A couple of years ago, I became fascinated with artificial intelligence and decided to build my own AI system. It was an artificial neural network that "learned" from financial data spanning more than 50 years. My goal was to see what the AI system would say about whether I should be invested in stocks (I used the S&P 500 as a proxy) or in cash based on available financial data.
To be sure, the complexity of this system was not even in the same league as the AI system used to power the AIEQ. But the results that I obtained from my experiment were intriguing. The AI system that I developed always recommended being in stocks. And its confidence in stocks increased more when the S&P 500 fell.
I dubbed the system a "Buffett-bot." Warren Buffett famously stated that investors should "be fearful when others are greedy and greedy when others are fearful." That's basically what my AI system concluded was a winning strategy.
It won't surprise me if AI systems eventually outperform the market year in and year out. But I think the way they'll do it won't be by trading in and out of stocks frequently like the current AIEQ is doing. Instead, I suspect that the market-beating AI of the future will recommend being in stocks most of the time, if not all of the time.
My hunch is that it will, like Warren Buffett, be greedy when others are fearful. I think it will select the stocks of companies that have solid business models and great long-term growth prospects like Amazon and Alphabet have right now.
Basically, this winning AI will just do what David and Tom Gardner have been doing for years.
Check out all our earnings call transcripts.
Editor's note: A previous version of this article didn't include total return comparisons between the AIEQ ETF and the S&P 500.
