Waymo Fails 1 in 20-30 SF Rides, Signaling Physical Job Automation Is Further Off
herbiebradley · x · 2026-09-30
Based on firsthand Waymo experience, the author argues AI automating physical jobs is further away than commonly assumed:
- Waymo cars run reasoning VLMs to plan out of sticky situations and operate in a fully mapped area, yet still fail roughly 1 in 20-30 rides in San Francisco, needing remote operators to rescue them.
- Self-driving at SL5 appears to be an AGI-complete problem; even a drop-in remote worker AI with vision would struggle with the long tail.
- He predicts applying GPT-8-class models to real-world robotics control will hit the same long-tail wall, implying near-term physical automation is overhyped.
More from AGI Musings
- Edge Agent Village V2: hundreds of humans to live with personal AI agents for 3 weeks — sydneymlevine · 2026-09-30
- Peter Diamandis: We're Still Living in the Singularity — PeterDiamandis · 2026-09-30
- Judea Pearl points to Eric Horvitz's remarks on probabilistic reasoning's 42-year-old rise in AI — yudapearl · 2026-09-30
- TasteLabs founder: AI floods the world with slop, so raise everyone's creative floor — korymath · 2026-09-30
- Bottleneck is diffusion, not price: why AI won't supercharge GDP by 2029 — herbiebradley · 2026-09-30
- Former anti-AI folks who raged at 'it'll only get better' are now all in on AI — justalexoki · 2026-09-30