

Causal AI, Modularity & Learning || Andrew Lawrence || Causal Bandits Ep. 002 (2023)
10 snips Nov 7, 2023
Andrew Lawrence, Director of Research at causaLens, shares his insights on the fascinating world of causality and modularity in AI. He discusses his journey from academia to industry, the role of Bayesian non-parametrics in understanding causal relationships, and the importance of collaboration in causal discovery. Andrew highlights the challenges of applying generative AI in high-stakes scenarios and underscores how teamwork is vital in translating research into practical applications. He also reflects on his supportive childhood and the significance of mentorship in tech.
AI Snips
Chapters
Books
Transcript
Episode notes
Andrew's Causality Journey
- Andrew Lawrence's causality journey began after his PhD, where he studied Bayesian non-parametrics.
- He transitioned from industry to academia and back to industry, joining causaLens in 2019.
Bayesian Non-parametrics and Causality
- Bayesian non-parametrics, specifically focusing on conditional probabilities and latent variable models, is applicable to causal inference.
- This background helps understand how to factor the joint distribution and capture the true data-generating process.
Deep Learning's Limitations
- Deep learning's focus on minimizing in-sample error can lead to overfitting and poor generalization.
- Identifying the true drivers of the target variable is crucial for accurate predictions on unseen data.