How Self-Driving Cars will Destroy Cities (and what to do about it)

Self-driving cars are already operating in parts of the United States, but the most consequential claim here is that their rollout—driven by profit incentives, weak regulation, and a car-first design philosophy—could worsen city life rather than fix it. The argument isn’t that autonomy is impossible; it’s that deploying it on today’s road network and under today’s political economy will likely “destroy the fabric” of cities by increasing vehicle miles traveled, locking in car-centric land use, and reshaping streets for private vehicles instead of people.

The case begins with a reality check: autonomous vehicles are not fully hands-off everywhere. In San Francisco, “robo taxis” have required human assistance at intervals measured in miles, and remote operators have had to intervene when vehicles get stuck behind construction equipment or fail to execute maneuvers like turning right. Even when cars appear cautious and obey speed limits, the system still makes basic mistakes—such as driving in circles or honking—sometimes for days. Reports also describe driverless vehicles interfering with emergency response and blocking streets or transit.

A sharper warning comes from safety incidents. A widely cited example involves a pedestrian in Tempe, Arizona, struck by a modified Volvo SUV operating with an automated driving system developed by Uber. The crash report describes the vehicle switching its perception of the person between categories (vehicle, bicycle, and “other”), only recognizing the crossing risk about 1.2 seconds before impact—too late to stop safely. Another San Francisco case describes a woman hit by a human-driven car and then dragged under a Cruise robo taxi; the vehicle continued because it treated her as no longer present, prompting removal from service. The broader point: sensors and software can fail in ways that are not obvious until rare edge cases occur.

From there, the argument pivots to incentives and infrastructure. The rollout is portrayed as a bet that autonomy will be marketed as “safety,” while companies lobby for self-certification and reduced reporting of safety data. Even if autonomous systems reduce some crash types, the speaker argues they will introduce new ones—especially if they normalize higher speeds or treat pedestrians and cyclists as obstacles rather than participants in street life.

The transcript also targets American road design and induced demand. It argues that U.S. streets—especially high-speed “stroads” designed for fast travel—already endanger people, and that replacing human drivers with autonomous ones won’t fix the underlying geometry. Instead, cheaper, always-available robo taxi service could increase trips and longer-distance travel, intensifying congestion and sprawl. The same logic is extended to parking: if cars can drop off passengers and then circulate to find cheaper spaces, parking lots and garages may not disappear—they may be repurposed into charging and staging areas.

Finally, the transcript frames a political risk: autonomous vehicles could reshape cities around private subscriptions and lobbying power, undermining public transit. The proposed alternative is not “stop autonomous vehicles,” but “prepare cities so autonomy serves public goals”: limit where cars can go (including AVs), lower speeds, remove parking, expand high-capacity transit, and price driving to prevent empty vehicles from constantly circling. A comparison to Utrecht, Netherlands, is used to argue that car-light urban design can deliver safety and mobility without waiting for advanced automation—suggesting the real lever is car dependency, not driverless technology.