Autonomous Autos – Fatal flaw

Reply to this post

First fatality underlines how long this is really going to take. 

• A fatal accident in Florida involving a Tesla Model S that had Tesla’s Autopilot feature activated, serves to highlight that it is going to be many years before the market is ready to adopt fully autonomous driving.
• The accident happened in May and involved a tractor trailer which appears to have failed to give way when crossing the highway causing the Tesla to pass under the trailer with tragic results.
• The trailer was painted white and was highlighted against a brightly light sky and neither the Mobileye sensor nor the driver saw the obstacle before the collision as the brakes were not applied.
• Both Tesla and Mobileye have been quick to highlight that this incident was both a very rare occurrence and that Autopilot can only be activated by the driver after accepting all liability.
• Mobileye also pointed out that its sensor system is not designed to deal with a laterally crossing vehicle but that this is on its roadmap for 2018.
• I think that although both Tesla and Mobileye are blameless in this incident, the automotive safety regulators will come down hard on what it is likely to perceive as untested features being made available in production units.
• Fortunately for Tesla and Mobileye, this incident did not involve the fatality or serious injury of a third party, but the National Highway Traffic Safety Administration (NHTSA) is likely to consider this as a real risk.
• Every component that goes into an automobile needs to be rigorously tested to ensure that it is reliable enough not to cause fatal accidents and I can see this being extended to software.
• Many automakers have set a deadline of 2020 (Baidu 2018) by when they expect to have fully autonomous functionality in commercial vehicles but I can’t see the regulator letting them leave the factory.
• The end result is that while the technology might be ready, it is clear that the market is almost certainly not going to be ready resulting in this technology sitting on the shelf for a while.
• The issue of liability alone will keep the regulator and the insurance companies busy for years (see here).
• The default position today is that the driver of the vehicle is liable for any accidents that were the fault of the software.
• This is an easy solution but it completely obviates the point of an autonomous system as no one wants to go to prison due to software bugs.
• Consequently, I continue to believe that the auto makers much more time than the current hype would suggest to get their autonomous systems to work properly.
• This gives them the luxury of not being forced to go with Google (see here) which I continue to believe remains a threat to any automaker hoping to differentiate based on digital services or data.
• The net result is that the market will take much longer than expected to develop and those that are now rushing to get to market are likely to arrive way too early.
• I still do not expect to hire an autonomous vehicle in Silicon Valley much before my retirement.