The authors also discuss the potential of LLMs in synthetic data creation, which could be beneficial given the scarcity of financial market data. They suggest that while LLMs may not currently be suitable for quantitative trading, they could enhance fundamental analysis by helping analysts refine investment theses or uncover latent business relationships. Despite the challenges, the authors advocate for an open-minded approach, given the unexpected successes in the AI field.
Key takeaways:
- The AI revolution has sparked interest in using Large Language Models (LLMs) for predicting price or trade sequences in financial markets, similar to how they predict sequences of words.
- Despite the potential, predicting financial data is more challenging than language due to the unpredictable nature of markets and the higher noise-to-signal ratio in financial data.
- Emerging areas of AI research like multimodal learning and residualization could have promising applications in finance, such as combining different types of data for predictions or creating synthetic financial data.
- While the current likelihood of AI models like GPT-4 taking over quantitative trading is low, the rapid advancements in AI suggest keeping an open mind for unexpected breakthroughs.