A joint research team from UCLA, the University of Edinburgh, and Sungkyunkwan University found that generative AI such as ChatGPT can help invest in stocks, but in the long run, it does not beat a simple investment strategy of buying stocks and holding them. In a recent paper published on the pre-release site "Archive," the researchers verified a generative AI based stock investment strategy for 20 years from 2004 to 2024. The researchers cited the 'verification method' as the biggest problem of previous AI investment papers, noting that many targeted less than 10 stocks and some evaluated only for three months. The team created a new system using 20 years of stock prices, news, and corporate disclosures, including companies delisted from the S&P 500. Most AI investment strategies lost competitiveness when the period was extended to 20 years, often performing worse than traditional techniques like "buy-and-hold" strategies. In a bull market, AI traded too cautiously and missed opportunities; in a bear market, it traded too aggressively and increased losses. The paper evaluated this as "too conservative in the bull market and too aggressive in the bear market."