From Analytics India Magazine: 2024-06-03 08:35:36

Researchers at the University of Chicago’s Booth School of Business have shown that OpenAI’s GPT-4 can match or surpass human experts in financial statement interpretations. Utilizing a method called chain-of-thought, GPT-4 was trained to mimic human financial analysts, resulting in successful market analysis and future predictions.

The study revealed that GPT-4 achieved 60% accuracy in forecasting profit direction, outperforming most human analysts. The model was trained to identify patterns, calculate ratios, and synthesize data, leading to precise predictions. The researchers emphasized that GPT-4’s predictions were not based solely on training memory but on insightful narratives regarding a business’s potential performance.

Despite the impressive results, skeptics caution against overinterpreting the findings. ChatGPT’s accuracy is contingent on the questions posed, and comparisons to outdated benchmark models have been questioned by some AI researchers. The ongoing debate revolves around the practicality of AI-driven stock market analysis vis-a-vis traditional approaches.

In a separate experiment, ChatGPT demonstrated a 26.9% net return amidst a 20% drop in the S&P 500 index, showcasing AI’s potential and limitations in stock market analysis. Additional research from Virginia Tech and JPMorgan AI Research indicated that while GPT-4 shows promise, it might not yet match the proficiency of human financial analysts. Thus, traditional methods still hold sway in financial analysis.



Read more at Analytics India Magazine: OpenAI’s GPT-4 Shows Prowess in Picking Stocks