Artificial intelligence chat robot ChatGPT may possess the ability to predict stock prices, better than random predictions, according to a recent study by Alejandro Lopez-Lira, a finance professor at the University of Florida. Although the paper is yet to be reviewed, Alejandro conducted the experiment for several months and found that ChatGPT can accurately judge the stock price trend for the following day by parsing over 50,000 news items for stocks listed on the New York Stock Exchange, Nasdaq, and a small stock exchange.
During the experiment, Alejandro and his partner Yuehua Tang entered headlines into ChatGPT, along with several prompts. ChatGPT was asked to pretend to be a financial expert with stock recommendation experience and answer “Yes” for good news, “No” for bad news, or “Unknown” for uncertainty, and then explain the reason in a concise sentence on the next line. The results showed that ChatGPT’s hit rate was significantly higher than random guessing.
While Alejandro was surprised by the results, he emphasized that the so-called “large language model” still cannot accomplish many financial tasks, such as predicting target prices or performing mathematical operations. However, ChatGPT’s ability to understand financial news headlines and judge stock prices could pose a threat to many high-paying jobs in finance. According to a report issued by Goldman Sachs, about 35% of financial jobs will face the risk of being automated due to the rise of AI.
Although hedge funds have shown interest in Alejandro’s research, he believes that as more institutions integrate the technology, ChatGPT’s ability to predict stock prices will gradually decline. “The more people use such tools to improve market efficiency, the more unpredictable the rate of return will become.” He speculates that the predictability of returns may be completely eliminated within five years after actual use.