The finance professor at the University of Florida, Jay Ritter, conducted an experiment to test whether OpenAI’s ChatGPT could predict the direction of stock prices by analyzing financial headlines.
The experiment involved training the model on news headlines from the Dow Jones Industrial Average from 1986 to 2018 and then testing its ability to predict stock prices in 2019.
The results of the experiment showed that ChatGPT was able to predict the direction of stock prices with an accuracy of 59.1%, which is better than random chance.
However, the experiment had limitations, such as the relatively small sample size and the fact that the model was only trained on headlines from one index.
Despite these limitations, the experiment is still significant because it demonstrates the potential of large language models to make accurate predictions in the financial industry.
If this technology is further developed, it could have significant economic implications.
However, there are also concerns that the advantage of using LLMs in the finance industry could diminish if more people start using them, which could lead to a crowded market and less reliable predictions.