By Jonel Juste
When a weekend power outage knocked out traffic lights across San Francisco, human drivers adapted by easing through dark intersections while Waymos halted and blocked traffic —but should these autonomous vehicles be more human, learning to show similar flexibility instead of freezing?
Last weekend, a power outage hit San Francisco hard. A fire at a PG&E substation knocked out electricity for about 130,000 people and turned off traffic lights across large parts of the city. Human drivers kept moving—they slowed down, looked around, and often rolled through dark intersections when it felt safe. But Waymo’s self-driving cars did something different: many of them simply stopped and stayed stopped, blocking streets.
This wasn’t a bug. It was by design. Waymo programs its cars to treat a dark traffic light exactly like a four-way stop sign. That matches California law, which says every driver—human or robot—must come to a full stop and wait their turn. In normal situations, this rule works well. But when dozens of lights go out at once and drivers start acting unpredictably, the robots’ strict obedience turned into a problem. They waited too long for perfect clarity in a messy, real-world jam.
The incident raises a simple question: Do self-driving cars need to act a little more like people in emergencies?
Autonomous vehicles are built to be safer than humans. They don’t get tired, drunk, or distracted. Waymo has driven millions of miles and claims better safety numbers than the average driver. Most days, their careful rule-following is a strength. But when the power grid fails—or when an earthquake hits, or a big storm comes—the world stops being orderly. In those moments, humans adapt. We read the situation, make quick judgments, and sometimes bend small rules to avoid bigger problems. Waymo’s cars, stuck to their code, created new problems instead.
Could engineers give these cars more flexibility? Yes, in theory. They could train the AI on thousands of hours of chaotic traffic footage, teaching it to inch forward safely or pull over faster when things go wrong. Some competitors, like Tesla, claim their systems handled the same blackout without freezing up.
But there’s a catch. If we make robots too flexible, we risk bringing back the very mistakes we’re trying to eliminate. Humans cause crashes because we speed, cut people off, or misjudge gaps. If self-driving cars start “bending” rules the way we do, they could become less predictable—and predictability is one of their biggest advantages. Regulators, insurance companies, and the public all want to know exactly how these vehicles will behave. Too much human-like improvisation could erode that trust.
The blackout also reminds us that self-driving cars don’t operate in a bubble. They depend on working traffic lights, reliable cell networks, and drivers around them who follow basic rules. Cities will need backup power for signals, better emergency plans for robot fleets, and maybe even limits on how many driverless cars can operate in dense areas until the technology handles chaos better.
In the end, the San Francisco outage wasn’t a failure of ambition—it was a reality check. Perfect obedience sounds great on paper, but roads are messy. If Waymo and others want to replace human drivers someday, their cars will need to survive the same unpredictable world we do. That might mean adding a small dose of human common sense: not to break rules recklessly, but to navigate tough moments without grinding everything to a halt.
The future of driving may not belong to machines that are perfectly lawful or perfectly human. It may belong to ones that know when to choose safety over strictness—just enough to keep the city moving.
Waymo currently offers fully driverless ride-hailing services to the public in Phoenix, the San Francisco Bay Area, Los Angeles, Atlanta, and Austin. The company is coming to Miami next, with public rides launching there in 2026.
jonelblog 