Capital Markets participants can’t trust black-box machine learning that overfits to past data.
Why do 87% of AI projects fail to make it beyond the experimental phase?
Darko, Andre and Max discuss the limitations of current approaches and share some cutting-edge success stories from causaLens clients. Andre and Max present a technical demo discussing execution strategy, portfolio optimization, and other exciting Causal AI use cases in Capital Markets.
Limitations of current approaches:
- “Black-box” models can’t be scrutinized or trusted
- Spurious correlations lead to severe overfitting
- AI pipeline fails to capture human insight
- Correlation-based models can’t optimize trading actions
- 00:00 – 09:36 Introduction to causaLens
- 09:37 – 17:25 Introduction to Causal AI (Treatment Effect Estimation, Simpson’s Paradox, Causal Graphs, Causal Discovery, Confounders, Counterfactuals and Spurious Correlations)
- 17:26 – 29:57 Causal AI in Trading & Execution (Linear vs Non-linear & ML vs Causal Models)
- 29:58 – 44:35 Causal Portfolio Optimization
- 44:36 – 51:06 Example Causal AI use-cases in Capital Markets: causaLake, Causal Model Risk Assessment and Stress Testing
- 51:07 – 1:01:29 Q&A