ChatGPT, an artificial intelligence chatbot created by OpenAI, is revolutionizing the world of finance. In two new papers, researchers from the Federal Reserve and the University of Florida deployed the AI in tasks to decipher whether the Fed’s statements were hawkish or dovish, and to determine whether headlines were good or bad for a stock.
The results of the new studies highlighted the advances of the technology known as natural language processing, which is able to parse nuance and context to deliver trading signals from reams of text from news articles to tweets and speeches. By interpreting the language in Fed statements and headlines, ChatGPT was able to outperform commonly used models like BERT, even outperforming the human analyst.
Slavi Marinov, the head of machine learning at Man AHL, commented that the hype surrounding ChatGPT is genuine. He explained that prior to the chatbot, firms had to manually label each sentence as positive or negative for an asset to train the machines in language interpretation. Now, with the advances demonstrated by ChatGPT, the whole process is sped up and more accessible to a larger community of finance professionals.
Gillian Hadfield, a professor at the University of Southern California and the author of the book Rules for a Flat World, remarked that ChatGPT is a game changer for artificial intelligence. Alejandro Lopez-Lira and Yuehua Tang, the authors of the University of Florida study, confirmed the potential of the AI by demonstrating that its answers had a statistical link to the stock’s subsequent movements.
Man AHL, where Marinov is based, partnered with Bloomberg LP, the parent of Bloomberg News, last month to make their natural language processing model available to the public. With this and the new findings of the ChatGPT technology, it is clear that the future of finance has just gotten one step closer.
