Odds on Open Hedge Fund Manager Ernie Chan: Use GenAI to Manage Risk, Not Predict Return
Sep 2, 2025
Join Ernest P. Chan, a leading figure in quantitative finance and founder of PredictNow.ai, as he shares his insights on the intersection of AI and trading. He reveals when machine learning models thrive and the pitfalls of data sparsity in financial markets. Discover how generative AI enhances risk management and portfolio optimization. Ernie also discusses alternatives to find alpha, the evolution of his quant strategies, and the vital traits for success in this evolving field. A must-listen for aspiring quants!
AI Snips
Chapters
Books
Transcript
Episode notes
Pre-Training And Joint Models Reduce Data Needs
- Generative AI lets you pre-train on vast unrelated data to overcome financial data scarcity via transfer learning.
- Modeling the input distribution P(X) or joint P(X,Y) provides a stronger prior than discriminative-only approaches.
Use Semi‑Supervised Learning On Financial Text
- Use semi-supervised pre-training on unlabeled financial text (analyst reports, Fed speeches) then fine-tune with a few labels.
- This yields usable signals without buying expensive labeled sentiment feeds.
Execution Is The Real Competitive Moat
- Although models and papers are public, reproducing state-of-the-art AI systems requires rare engineering talent and months of effort.
- Top AI engineers are expensive and scarce, so execution ability remains a competitive moat.
