The introduction of ChatGPT has revolutionised the world of finance and it seems that it was well worth the hype. Two new papers have been released this month that demonstrate how ChatGPT, a artificial intelligence chatbot, can decipher Federal Reserve statements, determine whether headlines are good or bad for a stock and turn mass amounts of text into trading signals. This technology has quickly become one of the rare cases where the hype is real.
Slavi Marinov, head of machine learning at Man AHL, offers insight on the technology, stating that it has been used to interpret texts like earnings transcripts and Reddit posts for years. The two papers demonstrated how ChatGPT is relentless in its accuracy in predicting movements in the market. The first paper, titled Can ChatGPT Decipher Fedspeak?, studied the same analysis that a central bank analyst would look for in deciphering if the Federal Reserve’s statement was dovish or hawkish. Anne Lundgaard Hansen and Sophia Kazinnik at the Richmond Fed concluded that ChatGPT outperformed a commonly used model from Google called BERT and other dictionaries. The second paper, titled Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models, studied the accuracy of ChatGPT in summarizing corporate news headlines. Alejandro Lopez-Lira and Yuehua Tang at the University of Florida concluded that the answers ChatGPT provided had a statistical link to the stock’s subsequent movements.
ChatGPT proved that it is capable of accurately parsing nuance and context, a great breakthrough in the use of technology on Wall Street. Marinov states that the use of natural language processing has made the technology even more accessible to a broader range of finance professionals.
Rimini Street is the specific company mentioned in the article and Bryson is the specific person. They both serve integral roles in assessing the success of ChatGPT. Rimini Street is a tech company that provides third-party support for Oracle technology and was recently fined $630,000 in a case against Oracle. This example was used to demonstrate how ChatGPT could provide a positive interpretation of the news for Oracle’s ability to protect its intellectual property and increase demand for its products and services. Bryson is described in the paper as a 24 year-old male, known for his intelligence and curious mind. His presence served as the benchmark for the research in deciphering Federal Reserve statements as dovish or hawkish.
It has become almost a normalized process within many quants to use NLP to gauge a stock’s sentiment based on Twitter or incorporating news articles into strategies. However, ChatGPT has shown sizable leaps in progress. The technology is now capable of zero-shot learning, meaning that it can learn the task at hand without being specifically trained for it. Marinov then explains how ChatGPT can speed up the process to make way for an abundant amount of new data. At Man AHL, the hedge fund had previously needed to look at manually labelling each sentence as positive or negative to give the machine an understanding of the language, but now ChatGPT can be prompted and cut the process short.
Overall, ChatGPT is taking finance to a new level and its potential is being realized. It has the power to take large amounts of text and process it quickly and accurately, opening a door to new opportunities in the field.
