Robots don’t get drunk or drowsy, so why hold up driverless cars?

shutterstock_603501413

The following op-ed was co-authored by Adam Thierer, senior research fellow with the Technology Policy Program at the Mercatus Center at George Mason University.


The House Energy and Commerce Committee is considering important legislation to streamline the process of getting driverless cars on the road.

With more than 90 deaths and 6,500 injuries in car crashes each day in the United States—94 percent of which are attributable to human error, according to the National Highway Traffic Safety Administration (NHTSA)—you’d think that automobile “safety advocates” would be doing everything in their power to push for a rapid rollout of autonomous vehicle systems. This is sadly not the case.

The committee’s legislative package, which aims to get a vote on the House floor in early August, has two important features. First, it harmonizes the role of the states and federal government through clear federal pre-emption of vehicle design and safety standards to avoid a patchwork of 50 different state laws (imagine if Illinois required a steering wheel, while Michigan said it wasn’t necessary).

Second, it raises the exemption cap, expanding the number of driverless cars allowed to test on public roads and increasing the amount of real world data we have about their performance. All this serves to speed up the deployment process. Traffic deaths are going up, not down, so every additional day matters.

Smarter cars should help radically reverse these troubling trends because—as NHTSA’s statistics indicate—robots don’t get drunk, drowsy or distracted like humans do when pushing 2 tons of steel and glass down the road at high speeds.

While automakerstrade associationsdisability advocates and think tanks have all expressed strong support for these moves, some auto-safety groups like Consumer Watchdog and Advocates for Highway and Traffic Safety are pushing back.

They have issued reports arguing that self-driving vehicles create “unprecedented risks,” that the “potential for an 80,000 pound rig using untested and unregulated technology on public roads is a very real scenario,” and that “[p]roviding further broad statutory exemptions … is both unnecessary and unwise.”

While their rallying cry of “safety first” is admirable, if safety is really our first priority, we need to test and deploy these cars as quickly as possible. Put aside for a moment the ludicrous claim that this technology is in any way “unregulated” — NHTSA already has broad recall authority to take any unsafe vehicles off the road, and any manufacturer seeking exemptions from federal safety standards must prove that their vehicle provides at least the same level of safety.

These critics have adopted the static view that gains in roadway safety come primarily through carefully controlled testing and restrictive regulatory policies. We can certainly learn a lot from restricted testing, but the greatest safety improvements will only come about through real-world learning.

Given that there will be some baseline rollout of driverless cars, there are fundamentally two ways we can save more lives: either try to make driverless cars safer before they hit the road, or get them on the road more quickly.

While both are important, auto safety advocates are so focused on the first method that they ignore the second to their own detriment. We could theoretically pause all deployment of driverless cars and tweak them forever in our laboratories until we are 99.99 percent certain that they are really, really, really safe. Meanwhile, tens of thousands of people would die on our highways.

Furthermore, this idealized notion of testing ignores the nitty-gritty reality that the “only testing that really matters is testing in the real world,” as General Motors Corp. President Dan Ammann has noted. Driverless cars will ultimately get safer through discovering the problems we can’t anticipate in carefully controlled test environments.

Not only have many major studies forecast significant life savings from autonomous vehicles, but autonomous features are already proving themselves safer both in the lab and on the open road. NHTSA’s recent report on Tesla’s semi-autonomous Autopilot function revealed 40 percent fewer crashes when it was activated.

Just how many lives might we lose if we slow down deployment? In a recent filing to NHTSA, we estimated that even a 5 percent slowdown in the deployment of driverless cars would cost an additional 15,000 lives over the next 30 years. A 10 percent slowdown would bring an additional 34,000 fatalities.

Beyond saving lives, we can improve the lives of millions of Americans. A coalition of disabilities groups noted that “innovative designs — which have the potential to unlock transportation for the 57 million Americans with a disability — are still subject to onerous barriers to entry.”

Therefore, these groups argue, “the need for flexibility is urgent… to ensure that our disabled, elderly, and wounded veteran populations have access to the same user-friendly transportation options that many today take for granted.”

Driverless cars are coming, one way or another, and if we don’t clear the roadblocks, other countries will gladly take the lead. Safety is, of course, everyone’s ultimate priority. But a full account of safety also means weighing the costs of inaction.

The House Energy and Commerce framework would streamline the real-world deployment of autonomous vehicles, while making sure NHTSA still has the tools it needs to verify safety gains. Which, by all accounts, seems like a balanced way forward.


Image by Turboman

 

FacebookTwitterEmailPrint
Top



Email this page.
Print Friendly and PDF