AI-Powered ETFs Underperform, Study Finds
A recent study published in Scientific American has cast doubt on the effectiveness of artificial intelligence (AI) in stock selection for exchange-traded funds (ETFs). Researchers Gary N. Smith and Sam Wyatt conducted an analysis of AI-powered ETFs, revealing that the majority of these funds have underperformed compared to the S&P 500 index.
The study examined ETFs launched since October 2017 that relied on AI for stock decisions. The findings paint a sobering picture of AI’s capabilities in the investment world, with over half of the AI-reliant funds having been closed due to poor performance.
In a comparison against the S&P 500, a widely recognized benchmark for stock market health, AI-driven funds fell short. Out of 43 funds using partial AI, only ten managed to outperform the S&P 500. The partly AI-driven funds showed an average annual return rate 5% lower than the S&P 500’s 12.4%.
The performance gap widened further for fully AI-driven funds. All 11 funds in this category lagged behind the S&P 500, with six actually losing money. While the S&P 500 boasted an average annual return of 7.6%, the fully AI funds recorded an average annual loss of 1.8%.
Researchers attribute these shortcomings to AI’s fundamental limitations. While AI excels at identifying statistical patterns, it lacks the ability to judge their plausibility or understand the real-world context of the data it processes. This deficiency makes AI unreliable for critical decision-making processes, including investment strategies.
The study’s findings underscore the continued importance of human judgment in investment decisions. As AI technology continues to evolve, the research highlights the need for these systems to develop a better understanding of context and meaning to become truly effective in complex financial environments.
This analysis serves as a cautionary tale for investors and fund managers considering AI-powered investment strategies, suggesting that current AI capabilities in stock selection remain limited and potentially unreliable.
