You might have missed it from the way Tesla Motors just announced that all its new cars will have self-driving hardware, but the company Thursday put a brake on its autonomous-vehicle technology.
Naturally, it's installing sensing technology in all its vehicles now, because that's what the entire automotive industry is doing. Look at sales of Mobileye's EyeQ chip, the visual-recognition hardware that powered Tesla's autonomous systems before a falling-out between the companies this year. The number of cars being produced with this technology by other car manufacturers vastly outstrips Tesla's own volumes:
The crucial bit of the announcement is buried in the last paragraph:
Teslas with new hardware will temporarily lack certain features currently available on Teslas with first-generation Autopilot hardware, including some standard safety features such as automatic emergency braking, collision warning, lane holding and active cruise control.
In other words, the new cars will probably be able to operate in some sort of self-driving mode at some future date , but won't be doing it at once -- unlike the vehicles that have been involved in crashes in locations including Florida, California, Beijing, Germany and the Netherlands in recent months.
That's a welcome retreat. As Gadfly has argued before, driverless cars are going to end up killing a lot more people before they start getting safer, and Tesla needs to recognize that for both its own sake, and that of the broader industry.
Minimizing that risk is going to involve collecting petabyte after petabyte of data about how driverless cars perform on the road.
Google, whose driverless-car development has fallen behind rivals, provides a good analogy. By taking out about three-quarters of the market for online search and almost 90 percent of the share in mobile operating systems, it has built up a database of the world's online behavior that no other competitor can match.
The numerous automakers competing to get self-driving cars on the road want to achieve the same thing: Acquire the best possible database of autonomous behavior to ensure their cars are safer than the competition's. That's one reason why the industry is wary of a push by U.S. regulators to get them to share more data: The proposal essentially asks them to surrender their competitive advantage.
By encouraging drivers to give up more control to their vehicle, Tesla's Autopilot mode may have given it an edge in data-collection over the more passive lane-assist and sensing systems employed by the likes of Volvo, BMW, GM and Toyota. But the costs of that high-stakes strategy have been outweighing the benefits: Aside from the rising toll of accidents, Mobileye blamed Musk's over-enthusiasm about the technology for ending their relationship, and Germany last week asked Tesla to drop the term Autopilot in case it was leading drivers to reduce their attention.
Giving up on autonomous mode in new cars will reduce Tesla's ability to overcome its numerical disadvantage in data collection against the major volume automakers. But it will ensure the entire industry is able to develop more prudently, and give regulators faith that engineers aren't putting inter-company competition ahead of road safety. That can only be a good thing.
This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
(Adds reference to timing in fourth paragraph.)
It's still in existing Tesla cars but won't be used in future models.
Tesla says elsewhere on its website, but not in its statement, that an enhanced version of Autopilot with autonomous freeway and parking capabilities "is expected to complete validation and be rolled out to your car via an over-the-air update in December 2016, subject to regulatory approval." That won't be fully self-driving, but Musk said on a conference call after the announcement that a Los Angeles-New York test drive of driverless technology will happen before the end of 2017.
To contact the author of this story:
David Fickling in Sydney at email@example.com
To contact the editor responsible for this story:
Matthew Brooker at firstname.lastname@example.org