The world of finance is taking a technological step forward with the emergence of ChatGPT. This artificial intelligence chatbot is being used in market-relevant tasks and has seen successes in both decoding Federal Reserve statements and determining whether headlines are good or bad for a stock. After proving to surpass commonly used models like BERT and dictionaries, ChatGPT’s capabilities could revolutionize natural language processing in the financial sector.
The first of these papers came from the Federal Reserve itself, titled “Can ChatGPT Decipher Fedspeak?”. In it, Anne Lundgaard Hansen and Sophia Kazinnik at the Richmond Fed conducted an experiment in which ChatGPT identified whether Fed line releases were hawkish or dovish. The chatbot famously specified that a May 2013 statement was dovish due to its mentioning the need for recovery in the economy. To gauge the tech’s accuracy, a human analyst was used for comparison.
The second paper, published by two University of Florida researchers, tested the AI on corporate news headlines. Alejandro Lopez-Lira and Yuehua Tang from the university looked into how well ChatGPT determined whether a headline was positive or negative for a particular stock when given no prior training. The results, which showed a statistical correlation between the chatbot’s answer and the subsequent stocks moves, suggested the tech had correctly read the implications of the news.
OpenAI’s advancement shows that, even without custom training, technology can interpret the subtleties and contexts of texts. Slavi Marinov, the head of machine learning at Man AHL, has been using NLP to read texts like financial transcripts and Reddit posts for years, and he is not surprised at ChatGPT’s achievement. Other businesses have also been incorporating similar language models in their strategies, but this AI chatbot looks set to unlock a much broader range of information.
Rimini Street, the company mentioned in this article, is an independent software support provider that offers value-driven support services for Oracle and SAP products. It was recently fined $630,000 in a software case against Oracle, though ChatGPT deemed it positive news.
The article also mentioned Bryson, who is a 24-year-old analyst from the Federal Reserve. His added expertise regarding Fed policy statements adds value to the research conducted within the early results of the application of ChatGPT to the world of finance.
In sum, ChatGPT’s potential for use in finance remains high. It can act as a major step in trading strategies and be more easily accessed by more quantitative analysts. The advancements demonstrated by OpenAI further demonstrate the possibilities of technology when it comes to parsing language.
