Having Perfected Self-Driving Cars, Tesla Is Working on a Crash Mode

We may earn a commission from links on this page.
Image for article titled Having Perfected Self-Driving Cars, Tesla Is Working on a Crash Mode
Screenshot: Tesla/YouTube

At Tesla’s Autonomy Day presentation today, an event orchestrated by the company to show off its self-driving capabilities, Elon Musk bragged that the company is close to being so good at self-driving technology that it will roll out even more driving modes for Tesla owners to use, depending on their mood and circumstance, including one that will, according to Musk, be cool with risking highway fender benders.


If at this point you are concerned how that’s actually better than human driving—read on!

During the Q&A session, someone in the audience asked about how Autopilot can manage Los Angeles traffic better, because his Tesla isn’t aggressive enough and other drivers take advantage of its timidity. To which Musk replied that Tesla will go beyond what’s called Mad Max Mode—currently available in Autopilot that seeks to change lanes to pass vehicles going even a few miles per hour below the assigned speed.


He then added:

We’re just being more conservative right now, and then as we gain higher and higher confidence, we’ll allow users to select a more aggressive mode. That will be up to the user.

But in the more aggressive modes and trying to merge in traffic, there is a slight, no matter how minute, there is a slight chance of a fender bender. Not a serious accident. But basically you’ll have a choice of: do you want to have a nonzero chance of a fender bender on freeway traffic, which unfortunately is the only way to navigate LA traffic.


So wait, let me get this straight: Self-driving cars—the whole point of which is to eliminate needless human errors that lead to preventable crashes, injuries and deaths—will only work well if we accept a risk on par with that of a human driver that it will crash sometimes?

Will Tesla drivers also get to decide what kind of car they would prefer to crash into? Will Autopilot take evasive maneuvers to ensure it doesn’t crash into another Tesla? Will Tesla drivers be able to determine how closely to pass cyclists, giving them the option to choose a non-zero chance, no matter how minute, of smacking the cyclist with your mirror? What about edging into pedestrian intersections? Will there be an option to ease your way in and crunch any toes that may get in your way?


We’ve reached out to Tesla for comment on this forthcoming feature and how it tracks with the company’s previous statements that “the safety of our customers is our top priority.” We’ll update the post if we hear back.

True to form, that was far from the only head-scratching comment Musk made during Autonomy Day. He also claimed that Tesla’s camera-based self-driving system is superior to ones that use LiDAR, saying: “LiDAR is a fool’s errand. And anyone relying on LiDAR is doomed. Doomed. Expensive sensors that are unnecessary. It’s like having a whole bunch of expensive appendixes.”


He then paused, before adding, “you’ll see.”

“It’s financially insane to buy anything other than a Tesla,” Musk added later in the presentation, after making a case for Teslas working as robo-taxis by next year earning its owners 30 percent of the car’s value back based on numbers they completely made up. “It’s like owning a horse in three years.”


Just prior to those specific Crashes-Are-OK Mode remarks, Tesla’s director of artificial intelligence Andrej Karpathy explained that dense urban traffic scenarios are some of the trickier circumstances for computers to learn. They require a very human level of intuiting what other humans are intending based on subtle cues that, shockingly, humans are simply better at detecting than computers.

So, begrudgingly, I have to admit this is an incredibly clever strategy to get around the trickiest problems with self-driving cars. Computers aren’t that great at learning these human communication methods. So what if they just... crashed sometimes? Problem solved.