Tesla Chooses Cost Savings over Safety, Again

As if ditching radar wasn't good enough, Tesla's now choosing to eliminate yet another safety measure in the name of cost cutting.

For a company that loves to talk up the safety of its vehicles, Tesla has now established a pattern of making safety compromises in exchange for cutting cost and manufacturing complexity. This week we got another example, the company saying it will no longer include ultrasonic sensors on Model 3 and Model Y, while both S and X will follow next year. 

Now, this won't impact the crash-worthiness of these vehicles at all, and indeed they do test incredibly well on all the world's most stringent of crash tests, as I learned when I got an exclusive tour of their crash test facility back in 2019. 

There’s no doubt that such results are indeed quite important, but these days the focus of vehicle safety has swung massively towards the Active side of the equation, most manufacturers investing huge sums to develop systems that prevent collisions from happening in the first place. In my interview with Volvo CEO Jim Rowan that ran on TechCrunch last week, he painted a picture of a future where active safety systems get the Minority Report treatment, looking so far down the road and being so aware of both the vehicle's surroundings and the driver's level of awareness that they'll not only avoid crashes, but avoid situations that might have led to crashes.

After all, regardless of how safe your car is, the best kind of crash is the one that never happens. 

Avoidance requires an increasingly advanced view of the world around the vehicle, plus powerful hardware and software within the vehicle to make sense of the gigs and gigs of data those sensors generate. With this latest move, Tesla is removing yet another view into the world around their vehicles, leaving only one thing: cameras.

Briefly, ultrasonic sensors are the little pucks you see embedded into the bumpers of many modern cars. They emit subtle clicks and noises that bounce off of obstacles, just like sonar in a sub but without the fear of torpedo reprisals. That's how a Tesla can alert the driver it’s 12 inches away from hitting the car ahead while parking – indeed, that’s how a modern Tesla can park itself.

“Without these sensors, how can a modern Tesla park itself? Well, it can't, actually.”

Without these sensors, how can a modern Tesla park itself? Well, it can't, actually. If you buy a new Model Y today you won't have access to self-park or summon functionality. Tesla says it'll be a "short period of time" before those features are re-enabled, but if the past is anything to go by it could be months before this stuff is working as it should, which is bad news for anyone buying a 3 or a Y today on a short-term lease. 

Worse, ultrasonic sensors aren't just used for parking in most cars. They're also used for blind-spot monitoring and rear cross-traffic alerts, even active lane changing. 

So, how will these new Teslas enable that functionality? Tesla says it'll rely on Tesla Vision, which again just means cameras. Eight cameras to be specific, looking forward and backward and off to the sides. Camera-based adaptive safety systems can do some impressive stuff, but they have some major limitations. Here are just three.

Seeing in the sun

I've spent a lot of time in a Model Y, one of the newer models without radar, and I can tell you that, should you be driving during sunrise or sunset on a stretch of road where the sun is glaring in your eyes, that sun is also glaring in your Tesla Vision cameras. I've lost track of the number of times I've had messages from Autopilot telling me that it's limiting its active cruise or other capabilities until conditions improve. Plenty of others have reported questionable behavior in bright sunlight over the years. 

I'm skeptical that these cameras will maintain consistent levels of performance in all lighting conditions. 

Seeing over the bumper

In an age of F-150s and RAM 1500s with noses so high you could parade a marching band before the front bumper and the driver would be none the wiser, the schnoz on a Tesla Model Y or even X is positively petite. But, still, there's a risk here. 

Tesla's primary forward-looking cameras are mounted in the top of the windscreen, three of them in fact each with a different focal lengths: wide, medium, and tight. These are the cameras that drive the majority of what Autopilot does, looking for traffic and obstacles and lanes. 

These cameras cannot see directly in front of the nose, which means they won't be able to warn you about your proximity to curbing or other low obstacles. Remember, that’s pretty much exactly what ultrasonic sensors were designed for. It's certainly possible to envision a solution where the camera identifies those obstacles while you're entering a parking space, the car effectively remembering their locations as the car maneuvers into position. 

However, it's also possible to envision multiple situations where that wouldn't work, like making a sharp turn into a narrow parking space blocked on both sides by other cars, or an animal scurrying out at the last minute.

Unlikely situations? Absolutely, corner cases without doubt, but as cars get safer we should expect their capabilities to protect us from ever-more unlikely situations, not ignore them.

Verification

As was very clearly demonstrated when Tesla ditched radar sensors in its cars last year, vision-based systems are not infallible. Suddenly the cars without radar slammed on their brakes with wild abandon, unable to tell whether oncoming cars were safe in their lane or about to cause a head-on collision. 

This so-called "phantom braking" effect was so disruptive and so unsafe that in my review of the Model Y for CNET I outright said to not buy the car -- despite it being class-leading in so many respects. 

Why did phantom braking occur? Because Tesla Vision was formerly able to vet its findings with the radar. The results of the two inputs could be compared. When the radar sensors were removed -- and ultimately disabled even on the vehicles that have them -- suddenly the vision system was on its own. Clearly, that system wasn't ready. 

Ultrasonic sensors aren't as vital to keeping you alive and avoiding major wrecks, but they are used for verifying nothing is lurking in your blind spot when making automated lane changes. And even if we ignore safety, if this latest specification of cars can't keep from scraping into the next car in a parking lot, that'll be more than a major inconvenience.

Questioning the motives

Just like when Tesla ditched radar, it's impossible to see this as anything but a move to both cut costs and simplify supply chain issues. That Tesla is already discontinuing ultrasonic installation before the company’s software is updated to work with the new configuration is all the proof you need.

I can’t help wondering why Tesla keeps doing it this way. Is it because the company’s supply chain is stretched so thin that the only other option is delivery delays? Or, is it because Musk believes that his software engineers need a little extra pressure to shut up and ship a solution that will save the company money? I genuinely don't know, but if you're not cynical about these moves you're not looking at the full picture.

Regardless, this is now a pattern: Tesla removing safety features from its cars to save cost and simplify manufacturing. Sure, the company's crash test results won't change, but again, that's largely table-stakes these days. 

Going forward, "safety" in a vehicle won't mean airbags and crumple zones, it'll mean cars automatically avoiding collisions and telling their drivers to put down the smartphone and pay attention. More and more brands are offering fast, beautiful, long-range, fun-to-drive EVs with over-the-air updates and service stations just around the corner, vehicles with advanced sensors like Lidar. If Tesla keeps moving in this direction, a once enviably progressive brand will only become increasingly obsolete.