Nutonomy President says "let's not compete on safety" since any crash hurts the whole industry

If you have fatalities, will people ride with you?

In a talk earlier this year Karl Iagnemma, from Nutonomy/Aptiv discusses public perception of robocars. In particular, he puts a focus on the difference between perception and reality, and the risks that presents. He then suggests that because the failure of any one company at safety (like Uber's fatality) affects all other companies, and thus companies should work together to improve each other's safety.

There are many different errors of perception, in both directions.

  1. Many polls released say that a large fraction of people are afraid of self-driving cars, though they are answering questions about a technology they have not experienced or which even exists.
  2. People who get in the cars trust them very quickly, even more quickly than they should.
  3. People attribute errors of one project (like Uber) to the others, or confuse a driver assist car (like the Tesla) with a robocar.

No matter what the misperception, though, Iagnemma is right when he suggests the problems of one player reflect on all the players. This is particularly true with the Uber fatality, which caused increases in the fraction of people who (prejudging) declare they would be afraid of riding in or even having robocars.

He went further to say that companies will not compete on safety, and as such they should all help one another.

In the past I have proposed that it would be very good if there were a standard simulator format for describing simulated scenarios, and that a large library of scenarios become publicly available -- free or for sale -- so that all teams can test as much as possible.

The US federal government, in version 1 of its regulations, even proposed requiring that all companies disclose logs of any incident, which would allow any incident to quickly be added to the publicly available simulation roster. They have since removed that proposed requirement.

This philosophy suggests there is a motive to share even without such a requirement. This is particularly true for the 2nd tier players. Waymo has less motive to share because right now they have more scenarios and real world data than everybody put together. They are unlikely to want to share, even to stop Uber from killing people and hurting their project.

Compete on safety?

The claim that players will not compete on safety is a bold one that defies simple analysis. Right now, it's obviously false, in that safety is the #1 top goal of every team for a valid commercial reason -- the first to safety is the first to market. You can't deploy your car on a road if you can't certify it "safe enough" on that road. For the first several years, who can deploy will be governed by who can promise safety and mean it. Having safety means a temporary monopoly until somebody else both reaches the safety goals and also has the resources to deploy in head to head competition with you.

The idea of not competing on safety arises only in an error when there are multiple players who have all reached "safe enough" in a large geographic region. For example, Lyft and Uber don't compete on safety. Most transportation companies (bus, train and airline) don't compete on safety. There is some competition on safety between airlines of different countries, almost entirely because of the lower safety requirements in some countries.

At the same time, car makers have competed on safety. Volvo built their whole brand around it, and owned the concept until Mercedes look that crown away. Even so, today these brands are considered safer.

Curiously, car buyers claim that safety is their strongest weighting factor in selecting a car. In addition to brand reputations for safety, buyers can look at crash test ratings and numbers of airbags and safety systems such as collision warning and avoidance systems. Other studies indicate that while buyers claim that safety is #1, it is in reality far lower, perhaps #7.

So if one robotaxi company pitches itself as safer, and has a record to back it up, will customers prefer that company? How strongly? If it's not a very strong differentiator, then it can make sense for everybody to work together on safety.

Safety, however, can be related for a long time to come to capability. A car that is safer can more easily go to new places. It might have more ability in bad weather or on complex roads. It might be able to go faster.

Famous injury accidents

A counter to the idea that companies will not compete on safety is the public's intolerance for famous injury accidents. While we accept the 100 people who die every day on U.S. roads and happily get into cars, will people get into a Waymo or Uber robocar the day after somebody died in one? Since perfection is not possible, this will be happening.

Some notable examples:

  • It didn't take too many (possibly false) reports of unintended acceleration in Toyotas for Toyota to take a serious hit
  • People still fly in Boeings after they crash, though in part we have faith that if a mechanical fault caused the crash, that those planes will be grounded until it is fixed
  • Some airlines lose passengers after crashes, but they usually recover.
  • Some forms of automated transportation have been shut down forever after an incident, like the Toronto Zoo domain ride
  • The early days of cars and planes and several other forms were rife with fatalities, but the forms eventually thrived

If some company can maintain a perfect record for a long period, it might give them a serious advantage. The public is fickle. Will the public become used to the idea of regular (but fewer) robocar crashes, and continue to use them after reading scary reports in the news? Or will ridership plummet?


Sensors help with safety. Almost every serious team believes in LIDAR as a key sensor, and the best and fastest way to safety. A few companies have made longshot bets that they can get there without it. The day -- whose date is unknown -- when cameras and radar can alone reach "safe enough" is a day when the no-LIDAR teams can finally get on the road, and also a day where the plans of the early teams suddenly become much more expensive. That's not a day early teams will relish, but they will be planning for it to make that transition more smoothly. However, this also makes it harder for the big-LIDAR, small-LIDAR and no-LIDAR camps to cooperate. This may also be true for other special sensors, like thermal cameras or high resolution radars.

Give up crown jewels

The elephant in the room is Waymo. Waymo has more miles than everybody else combined, and then some. They spent a large amount of money building up their scenarios and experience, and they also used their 5 year head start to built it. It is a large part of their advantage. Could they be convinced to reveal all they have learned about safety in those 10 years, in order to make lesser teams like Uber safer so the whole industry wins? It's a hard question that I will ask them.

If they don't, however, it is quite possible that all the smaller players, including the #2 and #3, might find virtue in joining together both to improve safety for the industry and to get a leg up on catching Waymo and getting to that world were safety is not a big competitive advantage.

Compete on other than safety

In my earlier article on robotaxi competition outlines the many other ways services might compete. Indeed, I only listed safety as a minor factor in comparison.


"Many polls released say that a large fraction of people are afraid of self-driving cars, though they are answering questions about a technology they have not experienced or which even exists."

Fearing a technology that doesn't exist is not necessarily irrational. Planet buster bombs that anyone can build in their basements don't exist, but that is no reason not to fear them in a hypothetical way.

The difference with driverless cars is that some form of them will exist soon, but what form? If one is skeptical of their actually being safe, then fearing them is perfectly rational. Is this skepticism warranted? We are talking about large corporations working to push product out the door, in a lag regulatory climate. It's not as if either car companies or tech companies have a spotless record about this sort of thing.

To put it another way, it is irrational to fear the hypothetical safe driverless car, but it is rational to fear the hypothetical unsafe driverless car. It would be unduly optimistic to dismiss out of hand the possibility of the second kind.

What I mean is the polls ask questions like whether you would feel comfortable riding in one. Since they have not seen one, they can't answer that question, but that does not stop the polls.

It is more akin to automated elevators, which for many years included an operator who pushed the buttons for you to make people feel better. The reality was that in time, people did not fear getting in automated elevators, though they thought they would.

Add new comment