Alphabet Inc.’s (NASDAQ:GOOG) (NASDAQ:GOOGL) Waymo co-CEO Dmitri Dolgov warns physical-AI developers against applying Silicon Valley’s break-things culture to autonomous cars and robots, where errors can cause irreversible harm.
Dolgov Recasts Silicon Valley’s AI Mantra
Speaking at a Y Combinator interview released on Monday, Dolgov said the technology industry’s familiar approach becomes unacceptable once artificial intelligence moves from digital products into machines operating around people. "Move fast and ship safely," Dolgov suggested instead.
Dolgov said developers must make safety part of their models, training methods and system architecture from the beginning rather than add protections after deployment. Chatbot mistakes typically require another attempt, he said, but failures involving vehicles can cost "human lives, not tokens."
"There’s simply not an undo and a retry button," Dolgov said.
Waymo’s Expansion Strengthens Its Safety Case
Waymo’s rapid expansion gives that warning commercial weight. The Alphabet subsidiary has logged more than 220 million rider-only miles and now delivers more than 500,000 fully autonomous trips weekly. Research comparing 56.7 million rider-only miles with human benchmarks found statistically lower rates of injury-reported and airbag-deployment crashes.
Recalls Expose Limits of Physical AI
That said, the record is not spotless. Waymo recalled nearly 3,800 robotaxis in May because software could send vehicles into flooded roads. It later recalled almost 3,900 vehicles after more than a dozen incidents in which robotaxis entered closed freeway construction zones, temporarily suspending freeway service while installing updates.
Waymo has also faced federal scrutiny over vehicles passing stopped school buses and behaving unexpectedly around traffic controls. NHTSA warned autonomous-vehicle developers in July about a "clear pattern" of driverless vehicles entering emergency scenes, blocking ambulances and firefighters, or failing to respond correctly to lights, smoke and traffic cones.
NHTSA Administrator Jonathan Morrison in a press release in early July said any automated vehicle unable to interact safely with first responders presents a danger to the public and urged companies to prioritize fixes.
Dolgov’s argument cuts both ways. Waymo’s safety studies support cautious commercial deployment, but its recalls show why physical AI requires continued testing after launch.
Image via Shutterstock
Login to comment