Recently, a study conducted by economics professor Gary N. Smith and student Sam Wyatt from Pomona College has sparked deep reflection on the performance of artificial intelligence (AI) in the stock market. Despite the hype surrounding AI continuously driving up the stock market, the reality is that many exchange-traded funds (ETFs) relying on AI for stock selection have not achieved the desired results.
Smith and Wyatt mentioned in an article published in Scientific American that they analyzed all publicly available ETFs relying on AI systems for investment decisions since October 2017. The results showed that the performance of these funds mostly lagged behind the S&P 500 index, which represents the 500 largest companies in the U.S. stock market. The study revealed that out of 43 funds partially relying on AI, only 10 outperformed the S&P 500, indicating significant issues with AI in stock selection.
To better understand the performance of these funds, Smith and Wyatt summarized that the annual return rate of funds partially relying on AI was 5% lower than the 12.4% of the S&P 500. Those completely reliant on AI and without human intervention performed even worse, with all 11 funds lagging behind the S&P 500, and six of them suffering losses during a generally bullish market. Overall, the 11 fully AI-driven funds had an average annual loss of 1.8%.
The researchers pointed out that AI is unparalleled in data correlation, but it does not understand the meaning behind the data. They stated, "The fatal weakness of AI systems is that while they can find statistical patterns, they cannot judge whether these patterns are reasonable or meaningless. Only when AI algorithms can understand the meaning of words and their relationship with the real world will they become reliable in important decisions, including investments."
Key Points:
🌟 Most AI-dependent ETFs underperform the S&P 500 index.
📉 Fully AI-dependent funds have an average annual loss of 1.8%, failing to profit during a generally bullish market.
🤖 AI can find data patterns but does not yet understand the actual meaning behind the data.