
Product Thinking for ML Under Uncertainty

Building Effective and Thriving Machine Learning Teams - David Tan & Dave Colls
Topic 6 of 9
Published Jan 27, 2025 · Clip 31:41–39:23 (7:42)
The guests discuss how to build ML products that deliver business value despite uncertainty. David Colls outlines essential characteristics of ML solutions: they're superhuman in speed and scale but not necessarily better than best experts, they make mistakes by design, and products must be designed to handle failure. They advocate starting with simple baselines, understanding cost sensitivity of errors, and designing products that mitigate failure modes (e.g., showing top-three predictions or asking users to correct). David Colls introduces Gen AI as a 'stone soup for innovation,' noting composed solutions need differentiated data, and that data plumbing for production always takes longer than expected.