The advent of modern technologies has brought forth a plethora of innovative ideas, designs, and advancements that hold the potential to transform the next century in stark contrast to the previous one. The rise of smart technology, particularly smartphones, tablets, and laptops, has ushered us into a digital realm in which we are all to some extent immersed. While this technological revolution has been groundbreaking, it is important to recognize that progress never stands still, and it appears that the wave of Artificial Intelligence (AI) is here to stay.
According to the Oxford Dictionary, "artificial intelligence" refers to the theory and development of computer systems capable of performing tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation. Platforms like "ChatGPT" and Google's "Bard" have granted everyday users the ability to seek opinions, solve mathematical problems, and even generate complete university essays with proper referencing. The sustained growth of AI suggests that we have only scratched the surface of its potential, despite the existence of critics.
So, how does AI impact the financial services sector? Recent developments, such as ChatGPT's demonstrated ability to comprehend financial news headlines and analyze their potential influence on stock prices, have sparked speculation that AI might even be capable of predicting stock market movements. With advanced language processing capabilities, ChatGPT can discern nuances and subtleties within headlines and stories, enabling it to make informed predictions about the stock market.
Alejandro Lopez-Lira, a professor at the University of Florida, and his assistant, Yueha Tan, recently published a paper outlining how Large Language Models (LLMs) can prove useful in predicting stock market prices. By employing ChatGPT for sentiment analysis of news headlines, Lopez-Lira and Tan discovered that the chatbot could predict the direction of next day's stock returns by assessing whether the headlines were "good or bad" for a particular stock.
In their experiment, they fed over 50,000 headlines about publicly listed shares on the Nasdaq, New York Stock Exchange, and the American Stock Exchange into ChatGPT. The bot then assessed the subsequent stock returns during the next trading day. Since the bot can only process data from September 2021 onwards, the sample period for the experiment spanned from October 2021 to December 2022.
The experiment's results led Lopez-Lira and Tan to conclude that "incorporating advanced language models into the investment decision-making process can yield more accurate predictions and enhance the performance of quantitative trading strategies."
The experiment yielded several notable findings that could have a profound impact on the world of finance going forward:
The study's results have the potential to benefit asset managers and institutional investors by providing evidence of LLMs' ability to predict stock market returns. This could serve as a basis for deciding whether to incorporate AI into their work or not.
The research can assist regulators and policymakers in understanding both the benefits and risks associated with increasing the presence of LLMs in financial markets. This understanding can pave the way for the establishment of regulatory frameworks and fair governance of AI in finance.
Lopez-Lira also predicts that as AI becomes more integrated into stock market prediction, the market itself will become more efficient. Keeping this in mind, it is highly likely that the predictability of ChatGPT's stock returns will diminish significantly in the next five years.
Nevertheless, it is essential to acknowledge the limitations of the professor's experiment. For instance, ChatGPT solely analyzed headlines and language, never considering target prices or conducting any calculations. Lopez-Lira himself noted that without a headline, ChatGPT has less than a 1% chance of making an accurate prediction. This demonstrates that the technology is far from perfect, and in the realm of stock forecasting, all results must be considered within the appropriate context.
Moreover, LLMs have struggled in the past to tackle complex mathematical questions. Google's "Bard" famously provided an incorrect answer to a math question during a keynote address, and recently, CNET's AI-written articles received criticism for making basic mistakes that a human advisor would easily catch but an AI bot might overlook. While these errors may seem inconsequential in isolation, ongoing oversights in this industry can have a significant impact on people's finances and livelihoods.
It is also important to question what consumers truly desire: financial advice from a real human who can comprehend the complexities of their concerns, or dialogue from a generated bot trained to regurgitate financial advice without any financial stake of its own. The exciting possibilities presented by AI are indeed captivating, but they do not yet constitute a new reality. While these tools can enhance the efficiency and output of financial services as a whole, they cannot replace the grounded advice provided by a human financial advisor who can understand and empathize with each individual's unique circumstances.