How Waymo Builds Self-Driving Cars

Summary

Srikanth Thirumalai, VP of onboard software at Waymo, explains the company's three-part autonomous vehicle tech stack: driver, simulator, and critique. He contrasts Waymo's approach with competitors' end-to-end machine learning methods, emphasizing safety validation as a key differentiator.

• Waymo uses a triadic tech stack: driver, simulator, and critique.
• The critique component detects suboptimal performance in both simulation and real-world driving.
• Waymo prioritizes on-board safety checks to ensure the driver's plans are safe.
• Competitors pursue an end-to-end machine learning approach with less hardware.
• Thirumalai argues Waymo's approach is safer due to an independent validation layer.
• The discussion centers on engineering philosophy rather than financial or market data.