At the height of artificial intelligence (AI) chatbots, everything seems to indicate that the true potential of this new technology has not yet been revealed. And it is that beyond the multitalents they have, such as writing cover letters or playing with code, these chatbots could still have the odd ace up their sleeve.
According to Alejandro López-Lira, a professor at the Department of Finance at the University of Florida, United States, large AI language models such as OpenAI’s GPT-3, on which the AI chatbot ChatGPT is based, can be useful for the forecast of stock prices.
Homework, cover letters… stock forecasts?
— CNBC's Fast Money (@CNBCFastMoney) April 13, 2023
University of Florida professor @alejandroll10 says ChatGPT *might* just be a top stock picker… someday.
He breaks down his fascinating research paper on what could be the future of trading: pic.twitter.com/tTVLxkAnob
Specifically, as part of an experiment, López-Lira and her colleague Yuehua Tang used ChatGPT to analyze more than 50,000 headlines from a provider of public equity data from the New York Stock Exchange, the Nasdaq, and a small-cap stock exchange to determine whether They were either good or bad for a stock, and surprisingly they found that ChatGPT’s ability to predict the direction of the next day’s returns was much better than conventional stock barometers and more accurate than random forecasts, they say in a recent paper not yet. peer-reviewed.
“The fact that ChatGPT understands information intended for humans almost guarantees that if the market doesn’t respond perfectly, there will be predictability of profitability,” López-Lira told CNBC.
Current limits of linguistic models
However, the experiment also showed the current limits of linguistic models when it comes to concrete calculations. For starters, ChatGPT never really analyzed the target prices or did any calculations. Moreover, according to various reports, currently, linguistic models have a strong tendency to make arithmetic mistakes.
On the other hand, López-Lira predicted that if everyone in the financial industry started parsing headlines using linguistic models like ChatGPT, their advantage could diminish completely given enough time, wiping out any potential gains. In the next five years, López-Lira told CNBC, his kind of experiment would stop working.
“As more and more people use these types of tools, the markets will become more efficient, so expect the predictability of returns to decrease,” he told CNBC.
