A Sendhil Mullainathan delivered a fantastic keynote on Economics in the Age of Algorithms, a talk lasting approximately one hour (at normal speed). In it, he compellingly argues that algorithms (machine learning, AI, etc.) are, at their core, economics. Among the many fascinating topics he covered, I particularly enjoyed how he framed the difference between Estimation and Prediction. Estimation represents the traditional approach to tackling problems, while Prediction is the newer approach enabled by supervised learning (ML). In essence, estimation focuses on optimizing the estimated Beta, while prediction aims to optimize the estimated Y. Sophisticated algorithms that optimize Y (the prediction) are remarkably valuable for uncovering new and strikingly original hypotheses. However, economists (humans) play a crucial role in designing recommendations for payoffs and loss functions, ensuring that ML outputs are meaningful and actionable. Did this triggered your interest? Sit back, relax, and enjoy an hour of insights from one of the great economic minds! (text revised by a LLM) https://videosolutions.mediasite.com/Mediasite/Play/cb9d64c0274d4aae98b61dd6779791b31d?playfrom=970000
- Pedro
Read on Substack
No comments:
Post a Comment