Can SGD Select Good Fishermen? Local Convergence under Self-Selection Biases

Name: Can SGD Select Good Fishermen? Local Convergence under Self-Selection Biases
Uploaded: 2026-09-14T13:25:35.713856+00:00
Duration: 593 s
Description: The episode discusses 'Can SGD Select Good Fishermen? Local Convergence under Self-Selection Biases,' arguing that assuming unbiased data is dangerously naive for AI optimization. Experts conclude that building trustworthy systems requires modeling data generation processes and quantifying selection biases, moving beyond simply minimizing observed loss.

The episode discusses 'Can SGD Select Good Fishermen? Local Convergence under Self-Selection Biases,' arguing that assuming unbiased data is dangerously naive for AI optimization. Experts conclude that building trustworthy systems requires modeling data generation processes and quantifying selection biases, moving beyond simply minimizing observed loss.

Open the full episode