What Happens When Humans Stop Driving... | Interesting Times with Ross Douthat

ELI5/TLDR

Ross Douthat talks to transportation analyst Andrew Miller about a near future where robotaxis are normal in most American cities by around 2035. Waymo is winning on safety, Tesla is betting on cheap cameras, and the real bottleneck is not technology but liability law and human resistance to weird-looking accidents. The good version of this future has fewer cars, less parking, safer roads, and people getting their attention back. The bad version has worse congestion, public transit in a death spiral, and a slow erosion of one of the last embodied rites of passage left in American life.

The Full Story

The pitch, and why now

Miller opens with the macro number. Roughly forty thousand Americans die on the roads each year, and the vast majority of those deaths trace back to driver error. None of it has to happen. Below the death toll, he points to the immense reservoir of human attention that driving consumes, hours that could be spent on anything else. He is careful to say he does not want to prescribe what people do with that time. They can read or scroll. It is theirs.

The technology is moving now for two reasons. AI has made some of the harder perception problems tractable, and the economics are unusually good. Teaching a car to drive is expensive once, then almost free to copy. A robotaxi is a piece of capital that can earn money as many hours a day as you can keep it clean and charged. Miller dates the inflection point to mid-2024, when Waymo crossed the threshold from a city or two to more than fifteen, and he anchors the moment when this feels normal at around 2035.

Two technology stacks

Waymo and Tesla are running the same race with opposite assumptions. Waymo loads its cars with cameras, radar, and lidar, the laser-based system that builds a high-fidelity map of the car’s surroundings to within tens of meters. More senses means more reliability in rain, snow, and confusion. It also means each car costs a lot, which slows scaling.

Tesla bets that cameras alone are enough. If that bet pays off, Tesla can build robotaxis very cheaply and saturate the market. If it does not, Tesla has a safety problem that Waymo does not. Today Waymo runs without a human in the vehicle. Tesla’s Austin fleet still has safety drivers, and they intervene often. Miller’s bottom line is dry: he wishes Tesla all the best in this contest, and he thinks they are going to need it.

Suburbs, not the countryside

Robotaxis will not save rural America. There is no scenario where a Waymo waits for a fare in a town of two thousand. The suburbs are different. The American suburb, designed in Levittown’s image, is almost impossible to retrofit for public transit. Robotaxis fit it well. Miller imagines suburbs paying a small stipend to the operators to make the unit economics work, which means rethinking what a public transit agency even is.

The remote humans behind the curtain

When a Waymo executive was asked by a senator how many of Waymo’s remote operators sit in the United States, she did not have a number. The senator noticed. Waymo’s framing is that these humans are not driving the cars by joystick. They are dropping digital breadcrumbs. The car gets confused by knocked-over traffic cones and pings a human, who looks at the situation and tells the car it is safe to proceed, or routes it through points A, B, and C until it stops being confused. People disagree about whether that counts as driving. Miller thinks it does not, but he concedes the car cannot always work it out alone.

Liability is the whole game

Miller’s biggest point. The single unresolved question that decides whether this future arrives is who pays when an autonomous car causes harm. His answer is blunt. If the manufacturer claims the car is safe, the manufacturer eats the liability when the automated driving system is in control and at fault. Put up or shut up.

Waymo has accepted that standard for its fleet. Tesla, to its discredit in Miller’s view, has been slippery about driver-assist responsibility. The reason carmakers are nervous is Cruise, the General Motors subsidiary that was killed by a single accident. A jaywalker was thrown into the path of a Cruise vehicle, which struck them and then, executing its safe-position protocol, pulled to the curb dragging the injured person under the car. No human driver would have done that. One incident plus some regulator-baiting and the whole company was gone.

This is why automated driving is held to a much higher safety bar than humans are. Engineers call it six nines. Some critics find that obnoxious, since on net you would save lives by letting a merely above-average system loose. But regulators do not think on net. They think about who can be blamed.

The weirdness tax

When a self-driving car fails, it fails in an alien way. A Tesla in autopilot once tried to take Miller down a private road sealed off by a chain; he braked first, but only just. Waymos have hit a bodega cat in San Francisco and a child in Santa Monica. Each incident gets noticed in a way that the hundred daily human-on-human collisions do not. Miller takes this seriously as a regulatory problem, not just a vibes problem. The slow, humble rollout strategy is precisely how you earn the right to scale.

He compares the cultural transition to elevator operators in Mad Men. The first time you stepped into an elevator with no human inside, it felt strange. The fifth time it was nothing. Waymo’s strategy of seeding many cities with small fleets is essentially mass inoculation against the strangeness.

The good world and the bad world

Miller’s optimistic 2035: Waymo, Zoox, and Tesla all reach scale. Robotaxi rides cost forty to fifty percent less than today’s ride-hail. Two-car households go to one, one-car households go to zero. Parking lots become other things. Fewer people die. Everyone is electric, so the air gets cleaner. People get their hours back.

The pessimistic 2035 is more interesting because it is more plausible. Robotaxis arrive in enough volume to make congestion worse but not enough to actually displace private cars. Public transit enters a death spiral, because anyone who can afford a robotaxi defects, fares go up, service degrades, more people defect. The people too poor for the cheaper robotaxi end up with worse mobility than they had before. Trip times grow. The Candy Crush passengers do not notice. Everyone else suffers.

The fork between these worlds turns on two things. First, regulators willing to let robotaxis scale rather than capping them. Second, public transit agencies willing to fold robotaxis into their own systems instead of treating them as the enemy, which is hard because transit unions are powerful and the threat to their jobs is real.

The end of driving as a rite of passage

The conversation’s most interesting turn is at the end, where Douthat presses on what is lost. Learning to drive used to be one of the few remaining rites of passage in liberal America. It is also a kind of embodied knowledge, the recognition that you are not just a brain, that you have to pay attention to a heavy machine and develop reflexes for it. Both are vanishing already. Teenagers are postponing licenses. The general drift of American life is away from working with your hands.

Miller does not really argue with this. He says if we want driving to bear the weight of giving us full and healthy relationships to the world, we are asking too much of driving. Save the lives, take the time back, and put that money into solving the embodiment problem somewhere else. He admits he is not the one to tell you where.

Douthat closes with a left-right inversion he finds delicious. Today the libertarian states, Texas and Tennessee, welcome self-driving cars while Massachusetts and New York throw up friction. So the libertarian states are, in his phrase, “building the gallows on which human agency and independence will eventually be hanged.” Miller does not disagree. History’s ironies run deep.

Key Takeaways

2035 is the date to hold in your head for robotaxis being normal in major American cities.
Waymo (lots of sensors, accepts liability) is ahead. Tesla (cameras only, dodges liability) is gambling.
Liability law, not technology, is the gating issue. The clean rule: the manufacturer eats it when the system is driving.
One Cruise incident killed an entire company. Carmakers are correctly afraid, and that fear is what makes them slow.
The bad future is not that robotaxis fail. It is that they succeed just enough to gut public transit without replacing it.
The libertarian states are the ones inadvertently dismantling the most American symbol of personal freedom.

Claude’s Take

This is a good conversation because Miller is honest about uncertainty without being squishy. The liability framing is the cleanest version of the argument I have heard. If your car is so safe, prove it by paying when you are wrong. Everything downstream of that, the rollout pace, the regulatory caps, the safety theater, becomes much easier to reason about once you fix that one variable.

The strongest part is the bad-future scenario. It is rare for a tech-optimist guest to seriously articulate the partial-adoption trap, where a new mode is good enough to siphon the middle class off public transit but not good enough to replace private cars. That is the world American cities should be most afraid of, and it is the world they are likeliest to drift into by accident.

Where the conversation underdelivers is on the embodiment question. Douthat lands a real punch and Miller takes it gracefully, but the answer is more or less “not my job.” Fair enough as a personal position. Less satisfying as a public one, because the same logic that retires driving also retires every other physical skill that contains risk. Saying we will find embodiment elsewhere is a promissory note nobody has cashed.

Score is a 7. Smart guest, clean structure, the liability point alone is worth the hour. Not a 9 because the technical material is fairly familiar to anyone who has been following Waymo coverage, and the cultural questions get raised but not really worked through.