A lot has been written about AI and its impact on the financial industry in recent weeks, and an article in Financial Times tries to separate realistic fact from overblown fiction. While AI such as ChatGPT would mostly likely fail a CFA exam and doesn’t have a great track record of accurately predicting the stock market, it is true that AI could learn enough to potentially replace a junior analyst.
Torsten Slok of Apollo wrote a couple recent papers positing that ChatGPT can understand and interpret speeches from central bankers (ie, “Fedspeak), noting that the GPT models’ performance “surpasses that of other popular classification methods.” And Alejandro Lopez-Lira and Yehua Fang at the University of Florida found that while previous versions of “large language models” weren’t successful at predicting equity prices, ChatGPT does a much better job, outperforming other types of analysis systems that institutions may already be using. They noted in their paper that “future research should focus on understanding the mechanisms” that allows LLMs to predict stock moves so that “researchers can develop more targeted strategies for improving these models and maximizing their utility in finance,” the article quotes.
But a major hurdle for all the researchers is that financial markets offer a pretty limited database; all that can be analyzed is what has already happened in securities, and the further you go back the more rudimentary the data. Markets are also much more dynamic than most other areas that AI is starting to be used. While it’s unlikely that an investment firm would choose to turn over all of its operations to a computer, the tools are getting better and more accessible. Utilized smartly, AI could allow for much more efficient operations and become easier for the average employee to implement. Those kinds of advances will undoubtedly have an enormous impact on the financial industry, the article concludes.