Can AI Pick Winning Stocks?

Dear ChatGPT, “Have AI-powered stock picking models performed well?” … Let’s come back to this later.

Let’s first define some terms. In the AI investment discussion, there are:

1) AI ETFs (exchange-traded funds – the new and improved cousin of the mutual fund) that resemble a basket of companies operating in the AI space

2) AI-powered stock picking models

Let’s first look at AI ETFs where two of the biggest and best examples come from Mirae Asset:

• The Global X Artificial Intelligence & Technology ETF (AIQ) is a fund that “seeks to invest in companies that potentially stand to benefit from the further development and utilization of artificial intelligence (AI) technology in their products and services, as well as in companies that provide hardware facilitating the use of AI for the analysis of big data.”
• The Global X Robotics & Artificial Intelligence ETF (BOTZ) is a fund that “seeks to invest in companies that potentially stand to benefit from increased adoption and utilization of robotics and artificial intelligence (AI), including those involved with industrial robotics and automation, non-industrial robots, and autonomous vehicles.”

How have they performed when benchmarked against the most popular technology index, the Nasdaq-100:

Investing is AI companies seems like a good place to be and AI technologies may prove transformative, but it might not follow that this sector at present valuations will prove a good investment. Recall, for example, that in the 2000-2002 period, the Nasdaq lost 78% after the internet buildout with its attendant hype.

Of greater interest to me are the newer AI-powered stock picking models. Can AI break the market code and pick winners? It would certainly seem plausible that stock picking could be one of the best applications of emerging AI technologies.

Two of the best examples of these models are from Qraft and Amplify.

• The Qraft AI-Enhanced U.S. Large Cap ETF (QRFT) seeks to utilize AI processes to identify data patterns at a scope, scale and speed not readily achievable by humans alone. By continuously learning from expanding data sets, QRFT targets outperformance.[i]
• The Amplify AI Powered Equity ETF (AIEQ) is built on the IBM Watson™ platform. “Leveraging the power of artificial intelligence (AI), the unbiased and data-driven approach revolutionizes security selection by harnessing up to 10 years of historical data and then applying this analysis to recent economic data and news articles to transform security selection.

How have they performed when benchmarked against the most popular technology index, the Nasdaq-100:

We might also consider quantitative funds which have been practicing AI-like investing for decades. Among the “Quants,” Renaissance’s flagship Medallion Fund is the closest thing we’ve ever seen to an investment approach that broke the market code. Before the 5 & 44 fees, this fund generated 62% annualized returns from 1988-2021. Jim Simons, an esteemed mathematician, brought together a team of quantum physicists and number theorists to develop a process that saw one losing year and gaudy returns in most years, even in years when the broader market dropped. The Medallion Fund story is well told in Gregory Zuckerman’s The Man Who Solved the Market: How Jim Simons Launched the Quant Revolution Kindle Edition.

If you learned of these returns and read Zuckerman’s book, you might leave feeling bullish about AI’s ability to harness data and advanced mathematics to win in the markets.

To gain balance, you should then take the cold shower that is Roger Lowenstein’s When Genius Failed: The Rise and Fall of Long-Term Capital Management. The investment teams at Renaissance and LTCM are similarly structured with mirroring casts of brainy characters. Renaissance performs ok and then goes stratospheric, and the performance remains otherworldly for a long, long time. LTCM looks brilliant in becoming the most interesting hedge fund in the world and then it collapses in 1998, creating a small global financial panic.

Taken together, these two AI-adjacent examples, alongside a broader review of the mixed results Quant funds have delivered investors, leave us lacking great confidence in AI stock selection.

Back to my original prompt: ChatGPT, “Have AI-powered stocks picking models performed well?”

“Long-term: sustained outperformance is difficult due to market randomness, overfitting, and changing market regimes. AI models struggle during regime shifts (e.g., sudden interest rate changes and geopolitical events) because they’ve trained on historical data that may not reflect new realities … public or noisy data leads to weaker signals and higher risk of overfitting. Competition: AI models in markets are part of a zero-sum game. Once a pattern is discovered and exploited, it tends to disappear … consistent outperformance from AI stock picking has been elusive.”

Can AI Pick Winning Stocks? The answer for now is a hard “no.”

[i] While QRFT is small in terms of assets under management, it has been included due to its relatively strong performance.