In this episode of the Financial Thought Exchange, Lotta Moberg, CFA, PhD, speaks with Stephen J. Brown, PhD, Emeritus Professor of Finance at Monash University in Australia and at the Stern School of Business at New York University, and winner of the CFA Institute Research Foundation 2025 James R. Vertin Research Award.

Brown discusses the origins of hedge funds, their role as liquidity providers, and why their performance often disappoints relative to public markets. He explains how hedge fund risk differs from traditional market risk, the limits of diversification, and why rigorous due diligence is essential. The conversation also explores his research on sensation‑seeking behavior among hedge fund managers and its implications for risk‑adjusted returns.

Related research by Stephen J. Brown:

• Why Hedge Funds

https://www.tandfonline.com/doi/full/10.2469/faj.v72.n6.6

• Sensation Seeking and Hedge Funds

https://www.jstor.org/stable/26656034

In Part 2, Francesco Fabozzi, PhD—Managing Editor of the Journal of Financial Data Science—joins host Lotta Moberg, CFA, PhD, to explore how modern NLP and large language models are reshaping investment management. Building on the technical foundations from Part 1, this episode turns to real-world applications: when to fine‑tune models versus rely on prompt engineering, how retrieval‑augmented generation (RAG) keeps models current with fast‑changing financial information, and why agentic systems are emerging as powerful tools for research automation. Fabozzi explains practical use cases ranging from sentiment‑driven return prediction to efficient knowledge‑distillation workflows, research assistants that read earnings reports, and coding agents that help back‑test investment ideas. The discussion closes with a look at where innovation is headed, including the potential of "general price transformers" for market forecasting.

This episode is essential for anyone applying AI within investment processes.

Francesco Fabozzi, PhD, Managing Editor of the Journal of Financial Data Science, joins Lotta Moberg, CFA, PhD to unpack how natural language processing matured into the powerful tool it is today. The discussion traces early finance‑focused techniques—dictionary counts, sentiment word lists, and sparse document‑term matrices, along with their limits around context and negation. Fabozzi then explains how neural networks introduced embeddings and contextual meaning, paving the way for recurrent models and eventually transformer architectures. He breaks down how self‑attention, encoder–decoder designs, and decoder‑only LLMs transformed language understanding and made large‑scale modeling feasible.

This episode lays the groundwork for understanding how modern NLP models interpret financial text. Look for Part 2, where the conversation turns to practical applications in investment management.