Commuters ride an L train in New York CityLucas Jackson / Reuters

The New York City subway is a miracle, especially at 3 a.m. on a Friday night. But the system is also falling apart, and it’s going to cost billions to keep the old trains running: $19 billion, at least according to one estimate from city planners. The time has come to give up on the 19th-century idea of public transportation, and leap for the autonomous future.

Right now, fully autonomous cars are rolling around Pittsburgh, the San Francisco Bay area, and parts of Michigan, shuttling people from here to there with minimal manual intervention. Instead of fixing the old trains, let’s rip out the tracks and fill the tunnels with fleets of autonomous vehicles running on pavement. The result would be radical improvements in throughput while saving money and increasing the ability of the system to survive a fire, flood, or terrorist attack.

These subterranean highways would be dramatically simpler than public roadways for an autonomous artificially intelligent system because the tunnels could be limited to authorized vehicles only. No jaywalkers on cellphones. No babies in runaway carriages. Just a collection of competing fleets, centrally orchestrated and offering different levels of service to different groups at different prices.

Savings in time and energy would come from replacing extremely heavy trains that stop at every station with lightweight vehicles that depart immediately and go directly from A to B, stopping only at one’s destination. No more waiting or stopping every few blocks.

Prices could be flexible, adjusting to congestion and smoothing demand with a reservation system. Some cabs could be upholstered with the finest leather and plated with gold for those New Yorkers who want a truly luxurious experience. Others could have a work table for those who want to write on the way to the office. Trendy models might come sporting graffiti designs by name-brand artists, created as an homage to the 1970s-era trains. But most would probably be slim, simple, and utilitarian. People would pay to reserve a slice of the pavement at a particular time; and the tunnels would be maintained by these fees.The prices would move up and down, adapting to demand.

The new system could be dramatically faster. The autonomous vehicles would take passengers from their initial station to the final one without stopping once. The stations would have to be redesigned, outfitted with little entrance and exit ramps that carry that cabs which carry the riders up to where the turnstiles are now. This way, the traffic would never stop flowing. The vehicles would stop only when they’re out of the flow of traffic. (They could then be recharged and stored at each station until another passenger comes along.) The vehicles might not run as fast as the current trains—which can go 50 miles an hour or more at top speed—but the lack of stops would pay off. (As every rider of the express trains knows, every stop really slows things down.)

Renting out the tunnels by the minute would also allow for new uses. There’s no reason the system couldn’t carry packages, food, or other freight in little autonomous trucks during off-peak hours.

This federated system of companies would be a return to an earlier era, when the subway tunnels and elevated trains were built by competing companies working with the city. The first elevated train lines were started in the 1860s, and the first subway tunnel opened at the turn of the century. A confusing system of leases and private contracts knitted together private industry and the city government until they were integrated around World War II.

[Readers respond: The NYC Subway Is Not ‘Beyond Repair’]

In that era, the tech of choice was the train, and the trains were big. Autonomous vehicles, by contrast, can be incredibly lithe, especially if you skip over the car-shaped models and head for the super-lightweight transports called “hoverboards” or “scooters.” These clever devices with computer-driven balancing look a bit like skateboards but carry enough battery power to go a dozen miles or more. Some cities are already filled with scooters that can be rented by the minute and driven on the streets. Now we just need autonomous ones that run in the subway system.  

The cost? The market for these vehicles is incredibly dynamic and flooded with extreme competition, producing prices that range from a high of about $1,000 down to somewhere below $150. It’s easy to find models that cost just under $300—and these come with onboard computers that sync with smartphones so you can listen to music from Bluetooth speakers while you ride. These hoverboards on the market aren’t autonomous, of course, but that can be fixed with some of the technology already shaping the current generation of autonomous cars.

Putting millions of people on cabs and hoverboards would require cleaning up the tunnel walls, adding nice lighting, and replacing the tracks with a smooth, paved road. There will undoubtedly be obstacles. Cleaning up the walls and removing the old track will be difficult and costly. Some of the sections have double tracks that would cost twice as much to remove. Even so, a budget of $8 million for each of the 240 miles in the current system would add up to only $2 billion. Even if that cost were to double, that would still only amount to about one-fifth of the current estimate to repair the existing system. (This is merely an estimate, of course. Anyone who’s followed civic construction knows that it’s not uncommon for estimates like this—or the $19 billion renovation—to have doubled or tripled by the time the work is finished.)

There will be other costs, of course: 240 miles of LED lighting would cost millions of dollars, but installing such a system would have side effects that would make them worth every penny. Rats, after all, hate the light and would be pushed to find new underground homes.

Lighter vehicles would also use dramatically less energy. The current system uses 1.8 billion kilowatt hours to power the nearly 2 billion estimated rides each year. It’s a bit tricky to compare the power that’s running the trains with that of the lightweight boards, but many hoverboards routinely go 15 miles on a battery charge that’s about 15 to 20 percent of a kilowatt hour. In other words, a lightweight board might save 80 percent or more of the power depending on the length of the rides.

The biggest reason is physics and weight. The R-160a subway car, which entered service in 2006, is about 10 feet wide and 60 feet long, and weighs more than 85,000 pounds. The official capacity is about 200 people standing and 56 sitting and, even with the population’s expanding waistbands, more energy is put into accelerating and braking the R-160a car itself than is used to actually move the people. In comparison, it’s easy to find a basic model of hoverboard that weighs anywhere from 18 to 50 pounds, depending on tires and configuration. Think of the energy savings that could result from replacing all of that heavy subway car with 20 pounds of hoverboard—savings that are even more profound in the off-peak hours, when there may be only one person riding on a hefty subway car.

After the transition, we’ll save by repurposing and recycling. Storing vehicles would take space, but almost certainly not as much space as is set aside for train yards. This land could be sold or leased. We might recycle quite a bit by melting down the 2,500 miles of cables estimated to be in the current system. The biggest economic gift, however, would be the opportunity for advertising throughout the system. Forty billboards a mile adds up to 10,000 billboards throughout the system. Estimates for the rent for a billboard in New York range from under $1,000 a month in the outskirts to $200,000 a month in Times Square. If subway billboards are worth $5,000 a month on average—an estimate based on a variety of sources—that alone might yield more than half-a-billion dollars a year in ad revenue.

The biggest advantage of cleaning up the subway tunnels isn’t fiscal, but the fact that New York’s underground would become safer, making it possible for people to walk to safety in case of a fire, flood, or terrorist attack. Right now, people are pretty much stuck in a broken train like Spam in a can.

Creating an open marketplace for autonomous fleets would encourage innovation and evolution. Perhaps the public would like fat, overstuffed chairs on wheels in some years and tiny hoverboards in other years. Who knows? Competition would ensure that the fleets adjust to our tastes and the seasons.

Starting with a clean slate and imagining an autonomous subway lets us escape the 19th century, when cities needed tons of iron just to move a few people at a steady speed. The 21st century deserves new ideas and approaches that take advantage of everything we can do with computers and artificial intelligence and robotics. We don’t need to stick with our old tech when we can dream of something newer and smarter.