As part of his 2017 budget proposal, Secretary of Transportation Anthony Foxx proposes to spend $4 billion on self-driving vehicle technology. This proposal comes late to the game, as private companies and university researchers have already developed that technology without government help. Moreover, the technology Foxx proposes is both unnecessary and intrusive of people’s privacy.

In 2009, President Obama said he wanted to be remembered for promoting a new transportation network the way President Eisenhower was remembered for the Interstate Highway System. Unfortunately, Obama chose high-speed rail, a 50-year-old technology that has only been successful in places where most travel was by low-speed trains. In contrast with interstate highways, which cost taxpayers nothing (because they were paid for out of gas taxes and other user fees) and carry 20 percent of all passenger and freight travel in the country, high-speed rail would have cost taxpayers close to a trillion dollars and carry no more than 1 percent of passengers and virtually no freight.

The Obama adminstration has also promoted a 120-year-old technology, streetcars, as some sort of panacea for urban transportation. When first developed in the 1880s, streetcars averaged 8 miles per hour. Between 1910 and 1966, all but six American cities replaced streetcars with buses that were faster, cost half as much to operate, and cost almost nothing to start up on new routes. Streetcars funded by the Obama administration average 7.3 miles an hour (see p. 40), cost twice as much to operate as buses, and typically cost $50 million per mile to start up.

The point is that this administration, if not government in general, has been very poor at choosing transportation technologies for the twenty-first century. While I’ve been a proponent of self-driving cars since 2010, I believe the administration is making as big a mistake with its latest $4 billion proposal as it made with high-speed rail and streetcars.

The problem is that the technology the government wants is very different from the technology being developed by Google, Volkswagen, Ford, and other companies. The cars designed by these private companies rely on GPS, on-board sensors, and extremely precise maps of existing roadways and other infrastructure. A company called HERE, which was started by Nokia but recently purchased by BMW, Daimler, and Volkswagen, has already mapped about two-thirds of the paved roads in the United States and makes millions of updates to its maps every day.

Foxx proposes to spend most of the $4 billion on a very different technology called “connected vehicle” or vehicle-to-infrastructure communications. In this system, the government would have to install new electronic infrastructure in all streets and highway that would help guide self-driving cars. But states and cities today can’t fill potholes or keep traffic lights coordinated, so they are unlikely to be able to install an entirely new infrastructure system in any reasonable amount of time.

Moreover, the fixed infrastructure used for connected corridors will quickly become obsolete. Your self-driving car will be able to download software upgrades while sitting in your garage overnight–Teslas already do so. However, upgrading the hardware for a connected vehicle system could take years and might never happen due to the expense of converting from one technology to another. Thus, Foxx’s plan would lock us into a system that will be obsolete long before it is fully implemented.

Privacy advocates should also worry that connected roads would also connect cars to government command centers. The government will be able to monitor everyone’s travel and even, if you drive more than some planner thinks is the appropriate amount, remotely turn your car off to “save the planet.” Of course, Foxx will deny that this is his goal. Yet the Washington legislature has passed a law mandating a 50 percent reduction in per capita driving by 2050, and California and Oregon have similar if not quite-so-draconian rules, and it is easy to imagine that the states, if not the feds, will take advantage of Foxx’s technology to enforce their targets. No such monitoring or control is possible in the Google-like self-driving cars.

Foxx’s infrastructure is entirely unnecessary for self-driving cars, as Google, Audi, Delphi, and other companies have all proven that their cars can work without it. Not to worry: Foxx also promises that his department will write national rules that all self-driving cars must follow. No doubt these rules will mandate that the cars work on connected streets, whether they need to or not.

Some press reports suggest that Foxx’s plan will make Google happy, but it is more likely to disappoint. Google is already disappointed with self-driving car rules written by the California Department of Motor Vehicles. But what are the chances that federal rules will be any better–especially if the federal government is dead-set on its own technology that is very different from Google’s? If the states come up with 50 different sets of rules, some of them are likely to be better than the others, and the others can follow the best examples.

If Congress approves Foxx’s program, the best we can hope for is that Google and other private companies are able to ignore the new technology. The worst case is that the department’s new rules not only mandate that cars be able to use connected streets, but that they work in self-driving mode only on roads that have connected-streets technology. In that case, the benefits of self-driving cars will be delayed for the decades that it takes to install that technology–and may never happen at all if people don’t the extra cost for cars that can drive themselves only on a few selected roads and streets.

All government needs to do for the next transportation revolution to happen is keep the potholes filled, the stripes painted, and otherwise get out of the road. In contrast, Foxx’s is a costly way of doing more harm than good.