A recent report from the University of Chicago Booth School of Business stated that OpenAI's GPT-4 may be better than humans in financial analysis and prediction, and the long-short strategies based on its predictions can also outperform the market.
The findings could disrupt the financial services industry, which, like other business sectors, is racing to adopt generative artificial intelligence technology. According to the latest study, large language models can do a better job than humans at analyzing financial statements and making predictions based on those statements.
“Even without any narrative or industry-specific information, large language models outperform financial analysts in predicting earnings changes,” the study said. “Large language models show a relative advantage over human analysts when analysts are faced with difficult problems.”
The study used "chain-of-thought" prompts to guide GPT-4 to identify trends in financial statements and calculate different financial ratios. Based on this, the large language model can analyze information and predict future earnings results.
“When we used thought chaining prompts to simulate human reasoning, we found that GPT achieved 60% prediction accuracy, significantly higher than the accuracy of human analysts, who were closer to the 50% range in terms of prediction accuracy,” the study said.
The study's authors also say that large language models are able to identify financial patterns and business concepts even when information is incomplete, suggesting that the technology should play a key role in future financial decision-making.
Finally, the study found that if GPT-4's financial acumen is applied to trading strategies, more trading returns can be achieved, generally outperforming the market.
