Two University of Florida finance professors, Alejandro Lopez-Lira and Yuehua Tang, conducted a study exploring the potential of ChatGPT and other large-scale language models in predicting stock market returns. The professors used ChatGPT to parse news headlines to determine whether they were good or bad for stock prices, and found that his ChatGPT’s ability to predict the direction of next-day returns was far superior to random. discovered.
Lopez-Lira said he was surprised by the results, which seemed to suggest that sophisticated investors weren’t yet using ChatGPT-style machine learning in their trading strategies.
Researchers also found that other basic models such as GPT-1, GPT-2 and BERT were not as accurate in predicting returns. This shows that the ability to predict returns is the emerging power of complex models. These findings suggest that incorporating advanced language models like ChatGPT into the investment decision-making process may lead to more accurate forecasts and improved performance of quantitative trading strategies. .
As a result, Bloomberg has released a new GPT-based language model called BloombergGPT. It is trained on a dataset called FinPile and consists of financial documents, news, filings, press releases and social media in English. The company claims the new model improves existing natural language processing tasks such as sentiment analysis, news classification, headline generation, question answering, and other query-related tasks.
The model is not trained on multilingual data, but on Bloomberg’s vast repository of financial data spanning the last 40 years. It’s worth noting that Bloomberg already has a Bloomberg Terminal that leverages NLP and ML models to provide financial data. That raises the question of how much value BloombergGPT adds and how it compares to other his GPT models.
While the use of LLMs such as ChatGPT is gaining momentum in various areas, the potential of LLMs in predicting stock market returns remains relatively untapped in financial economics. There are arguments on both sides. On the other hand, LLMs are not explicitly trained for this purpose, so they may be of little value in predicting stock market movements. On the other hand, given its ability to understand natural language, it can be a valuable tool for processing textual information to predict stock returns.
Not bullish on ChatGPT
Business, transportation, science, law enforcement, healthcare, and more are embracing AI, but the financial industry is doing the opposite. For some time, Wall Street has used computer programs to handle tasks such as trading and risk management. However, investors have not made much progress in overcoming the main hurdle of using artificial intelligence to outperform the market. While some see ChatGPT as a vehicle to boost sales and research efforts, the use of AI in investing has not yielded notable results.
Jonathan Larkin, managing director of Columbia Investment Management, which manages and invests in Columbia University’s $13 billion endowment, said: various funds.
Decades ago, financial pioneers like Jim Simons of Renaissance Technologies used machine learning to develop algorithms that allowed computers to use historical data to make investment decisions with minimal human input. created. However, despite their success in building trading models that can identify patterns and generate profitable trades, these companies continue to rely on automated trading using state-of-the-art AI techniques such as self-learning and reinforcement learning. Not fully transitioned to operation. Instead, they continue to rely on advanced statistics and a “theory first” approach to establish hypotheses and build models on them, according to industry insiders.
“Most quants still take a ‘theory-first’ approach of first establishing a hypothesis as to why a particular anomaly exists and then forming a model on that basis,” Larkin said. wall street journal.
So while there may be potential use cases for ChatGPT across stock markets and financial structures, the market is concerned given the potential risks involved in not providing the expected accuracy and assistance. There is. So the market as a whole seems to have chosen to stick with the old ways until something truly groundbreaking comes along.
