Open AI, a venture headed by ntelligence guru Elon Musk, is accelerating efforts to revolutionize financial trading. Two new research papers have demonstrated that its AI chatbot, ChatGPT, can accurately decipher “Fedspeak” from the Federal Reserve statements as well as determine whether certain news headlines have a positive or negative impact on stock prices. This technology is seen as a new way for Wall Street quants to read text material, such as earnings transcripts and Reddit posts, and make trading signals.
The two new papers, Can ChatGPT Decipher Fedspeak? and Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models, used the chatbot to assess performances close to that of humans. In the first study, two researchers from the Fed itself found that ChatGPT could understand if the central bank’s statements were dovish or hawkish, outperforming other models such as BERT. What’s more, ChatGPT also successfully explained its conclusions in ways that mirrored the interpretations of a 24 year-old analyst at the Fed. The second paper suggests that ChatGPT was able to successfully interpret corporate news headlines, particularly when it was given the opportunity to ‘pretend’ to be a financial expert.
Slavi Marinov, head of machine learning at Man AHL, which has long researched natural language processing and text reading, commented, “It’s one of the rare cases where the hype is real,”. Man AHL has used a similar process to find patterns and make predictions. Therefore, it is clear that the advances demonstrated by ChatGPT far exceed prior technologies.
In addition, ChatGPT offers the ability to speed up the process adopted by Man AHL, including the need for manual labeling of each sentence. ChatGPT is also able to further automate of its processes as it can ‘pretend’ to be an expert and interpret certain corporate news headlines.
Overall, the supply of information and insights driven by technology such as ChatGPT can offer significant advantages to the world of financial trading. From its breakthroughs recorded in the two studies, it is possible for financial pros to access information from different forms of text material, including news from the Federal Reserve and corporate news headlines, and turn them into trading signals.
