Like most rational people, I’m thrilled about the potential for self-driving cars to take the wheel from the terrible drivers that clog our world’s roads. Driving a car is fun, but saving 1.24 million lives a year is a lot better.

This shift will happen eventually, but our path there is unclear. There’s the smart-car strategy, which envisions vehicles equipped with sensors, onboard computers and GPS data to navigate the world. Google leads this pack.

Then there’s the smart-roads approach, which calls for sensor-rich streets that help guide our vehicles. A recent example of this is a Volvo research project that placed magnets almost eight inches below a roadway to help cars determine their position.

Volvo (admirably protecting yuppies since 1927) has its heart in the right place. One of its goals is for no one to be killed or injured in a Volvo by 2020. And while these magnets have the advantage of still working during rain and snow — something Google’s car struggles with — there are plenty of drawbacks.

According to the American Society of Civil Engineers, 32 percent of our country’s major roads are in poor or mediocre condition. When a nation struggles to fill its potholes, why expect that we can succeed in installing and maintaining a network of sensors in our roads?

The cost of magnetic sensors would be significant. According to Wired, upgrading a major highway would cost $39,362.90 a mile. With roughly 4 million miles of roads in the United States, making every one of the country’s roads smart could cost $157.5 billion. Then there are maintenance costs.

In fairness, the price tag on a self-driving car isn’t cheap either. The equipment added to Google’s cars has been estimated at $250,000. The costs will have to come down, and history suggests they will. For some perspective, one gigabyte of hard drive space cost about $100,000 in 1980. Now the price per gigabyte for a hard drive is less than 10 cents.

“Cost is the least of my worries,” Google co-founder Sergey Brin told the New Yorker.

Let’s say road sensors get a lot cheaper. Does installing magnets along every road in America made sense? Not so fast.

Because roads are generally publicly owned and operated, governments would be in charge of installing these networks. Recent history has shown that in the quickly-evolving tech sphere, we’re probably better off letting the nimble private sector fight it out to deliver the best solution. Governments move slower and struggle in these spaces, as examples such as Philadelphia’s attempt at a municipal WiFi network or the rollout of illustrate. 

In the private sector, the best idea will emerge — consumers will gravitate to it — and the failures will fade away, costing only a handful of already wealthy venture capitalists money, or the shareholders of certain companies. A government-led approach would have all its bets on one solution, making failure more likely. And failure would be costly for all taxpayers.

So the solution won’t be state-owned smart roads, but smart cars built by private companies. Additionally, the software behind our self-driving cars will probably need updates to protect against security vulnerabilities, hackers and terrorists. Tesla has shown how easily this can be done with its cars. The “oil change” of the 21st century may be sitting in your garage and downloading a new version of self-driving software via your home WiFi. On the other hand, imagine replacing every out-of-date sensor on every road. It’s a headache to consider.