The latest research in the field of AI has opened up a new world of possibilities. A chatbot developed by OpenAI called ChatGPT can effectively interpret nuances, context, and language in order to predict stock market moves. The ability to turn reams of texts into useful trading signals has long been undertaken but ChatGPT has the potential to take this process a step further.
Two recent papers have been published using ChatGPT to test tasks relevant to the market. One paper tested the technology’s ability to understand Federal Reserve statements and determine if they are hawkish or dovish, and the other paper tested its ability to determine if news headlines were positive or negative for a particular stock. On both tasks, ChatGPT was able to outclass commonly used models from Google, such as BERT, and even classifications based on dictionaries.
Slavi Marinov, head of machine learning at Man AHL, has noted that this technology uses natural language processing in order to read texts such as earnings transcripts and Reddit posts, something that has been implemented in Wall Street strategies for years.
Therefore, ChatGPT has the potential to reach a new level in terms of accurately parsing and understanding language and context.
In terms of the individuals involved with this technology, Anne Lundgaard Hansen and Sophia Kazinnik are two researchers at the Federal Reserve who tested the ability of ChatGPT to decipher Fedspeak. Their paper, called Can ChatGPT Decipher Fedspeak?, showed that ChatGPT came closest to humans in understanding if the central bank’s statements were dovish or hawkish. The chatbot was even able to explain its classifications in much the same way as a human analyst.
Alejandro Lopez-Lira and Yuehua Tang from the University of Florida have published Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models. For this paper, the chatbot was given corporate news headlines that it had not seen before and asked to predict whether it would have a positive or negative effect on the stock. The answers given by the chatbot showed a statistical link to the stock’s subsequent moves, indicating that it was able to comprehend the implications of the news.
ChatGPT appears to be a welcome addition to the world of finance. Marinov believes that the use of this technology could potentially speed up processes when manually labelling data as it eliminates the need for certain tasks.
On Wall Street, AI tools like ChatGPT is already helping quants to make better-informed decisions but with ChatGPT, it looks like this technology could enable a broader community of finance professionals to access new information and make future trading decisions even more accurately.