Opinion: It's Past Time for a Tesla Autopilot Recall

Corey Lewis
by Corey Lewis

The evidence keeps stacking up against Tesla. As the National Highway Traffic Safety Administration investigates crash after crash involving Tesla vehicles under the influence (or suspected influence) of Autopilot, when is enough too much?

As we reported most recently, the NHTSA has racked up 14 investigations into Tesla vehicles that collided with other vehicles, including three over the past month. Most often, it seems the Teslas mistake large emergency vehicles for empty space and plow right into them. Through a combination of the company’s seemingly flawed Autopilot system and driver inattention, the death count keeps rising. It’s time for a recall of the whole damn thing.

Human beings prove time and time again that they’re willing to over-trust any driver assistance provided to them. Whether said assistance comes in the form of ABS, an airbag, all-wheel drive, parking sensors, stability control, or automatic headlights, the public puts immediate and complete trust in whatever assistance systems their vehicles left the factory with.

It comes as no surprise that the public is treating Autopilot like the fully autonomous system it isn’t. The system allows drivers to make momentary and occasional eye contact with the road, rather than demanding it full-time. Surely, if significant driver attention were actually required, collisions with enormous parked fire trucks would not occur. And that’s why the NHTSA needs to enforce a recall in this situation. Until Autopilot works reliably around all other vehicles and demands driver attention, it’s still in the development phase. It’s not ready for prime time.

Of course, the most important improvement is required driver attention. Drivers will do whatever they can to get around actually driving, as their phones beckon them like sirens of the sea. The system as it stands is clearly too lenient: Infrequent and distracted attention to the road is not enough to provide the driver with awareness of a small obstacle like a parked fire truck. I’d recommend that after a certain number of attention violations, the Autopilot system shuts down entirely. A cool-down period punishment of eight hours should suffice. Perhaps after two or three such cool-downs, a dealer maintenance visit is required to reactive Autopilot. Just spitballing here.

But until such time as the system is fool proof for the fools behind the wheel, it’s going to continue killing people. Safety is just a system update away; the NHTSA should file that recall paperwork ASAP.

[Image: NTSB]

Corey Lewis
Corey Lewis

Interested in lots of cars and their various historical contexts. Started writing articles for TTAC in late 2016, when my first posts were QOTDs. From there I started a few new series like Rare Rides, Buy/Drive/Burn, Abandoned History, and most recently Rare Rides Icons. Operating from a home base in Cincinnati, Ohio, a relative auto journalist dead zone. Many of my articles are prompted by something I'll see on social media that sparks my interest and causes me to research. Finding articles and information from the early days of the internet and beyond that covers the little details lost to time: trim packages, color and wheel choices, interior fabrics. Beyond those, I'm fascinated by automotive industry experiments, both failures and successes. Lately I've taken an interest in AI, and generating "what if" type images for car models long dead. Reincarnating a modern Toyota Paseo, Lincoln Mark IX, or Isuzu Trooper through a text prompt is fun. Fun to post them on Twitter too, and watch people overreact. To that end, the social media I use most is Twitter, @CoreyLewis86. I also contribute pieces for Forbes Wheels and Forbes Home.

More by Corey Lewis

Join the conversation
7 of 155 comments
  • Z9 Z9 on Jan 11, 2020

    I've had access to Autopilot in a couple of different cars for about three years. There are two basic features -- one is cruise control (TACC) which in my limited experience works substantially better than any other similar system that I've tried. Just as an example, I've found the Kia / Hyundai system to be positively dangerous in its inability to deal with slow traffic, so if anyone's going to do a recall we can start there. The Tesla system has been rock solid and it if it isn't, it's very aware of its limitations and will not work in cases of limited visibility (and why would you be trying to use it in those situations in the first place?). However, one of the features of the Tesla TACC is the ability to set a following distance between 1 and 7 (where the higher number is farther away from the car in front of you). I believe it now defaults to 5, but it used to be more like 3, which scared the crap out of me. So one regulation that could be an immediate improvement would be a minimum following distance for these systems that would allow a little more time for a last-minute reaction. If the system does fail, Tesla has an automatic emergency braking system that has saved me from a minor crash once, but the greater the following distance the better. Mostly I use TACC in heavy traffic; it is far more pleasant than constantly speeding up and slowing down. But I keep my foot poised over the brake pedal in case there's a problem. Even with that, a car making decisions about how fast it should go is definitely soporific. I imagine people who use the autosteer experience sleepiness to an even greater degree. Research I've seen suggests that when a random person gets in a self-driving simulator suggests they almost invariably fall asleep after a while. This is to me incredibly concerning. As for Tesla's autosteer adnd all the other stuff, I am already a nervous passenger so why would I want my car steering itself? I wouldn't have a problem if the NHTSA just said automatic steering is not happening, now or ever. My new Model 3 is the best handling, most fun to drive car I have ever owned so I wouldn't miss self-steering for a second. I've met a couple of other Tesla owners who are unfortunately convinced Autopilot is safer than their own driving. Another thing I am sure is happening is that elderly drivers are seeing Autopilot as a way to get a few more years of independence before Junior takes the keys away. In its current incarnation it almost certainly would not have prevented the crash that precipitated taking my mom's car away from her. You might also be aware that Tesla is now offering their own car insurance. I don't know if they do this now, but you can imagine, given their inclination, that they might start offering discounts for greater use of Autopilot. Again this combined with everything else around this technology has an unsettling dystopian feel to it. Or imagine you could agree to pay a little more if you wanted to be a sick daredevil and keep your TACC following distance at 1. Ah, good old vertical integration...

    • See 3 previous
    • Vulpine Vulpine on Jan 14, 2020

      @dal20402: To you and me, the freeway is not a racetrack; to others, it most certainly is. Again, I don't know where you live but I-95 between DC and Boston is one of the more heavily traveled stretches of freeway in this country at 440 miles long with three-lanes or more going each way over most of its length. Other highways may have more dense traffic over shorter distances but few offer similar density over similar range. Personally, I use cruise control almost all the time on the open highway; I have a lead foot and driving manually would have me running right along with those racers if I wasn't careful. When I was a young driver having to drive hundreds of miles on a moderately-regular basis, I discovered that driving by foot would have me maintaining a specific sense of speed that had nothing to do with my actual velocity... I'd try to cruise at the speed limit and before I knew it, I'd be going 80 or 90 mph because it 'felt' comfortable. I got more than one speeding ticket that way. So at the ripe old age of roughly 25, I bought a car with cruise control in it and with few exceptions I have had cruise in every car since. Traffic-aware cruise would be an advantage simply because I don't like having to override cruise for a slower mover, especially when other traffic won't let me pull out to pass. TA cruise would let me stay behind the slower car without 'footing' it until I had the opportunity to pass, at which point I might 'foot' it long enough to complete the pass and return to my own lane. Then again, I follow the old rule of "one car length per 10mph speed" to a very rough extent. At 10mph I might be farther because I expect unexpected maneuvers like braking for no apparent reason or someone in the left turn lane deciding they want to make a right turn instead (or vice-versa.) Such maneuvering doesn't always come with a warning signal from the vehicle making the maneuver. And yes, I use cruise even on two-lane highways simply because it's too easy to 'foot' it over the limit, given the relative power of today's cars vs the cars I first drove. It used to be that 200 hp was more than enough for everyday driving but it seems today that if you don't have 300 horses or more (and the associated torque) then you're too slow. I was happy with 150-200 horses but with current plans to buy a travel trailer, I needed more horses and a bigger tow vehicle--which means I have to be even more aware of my speed AND my size.

  • ToolGuy ToolGuy on Jan 11, 2020

    On the general topic of automotive fatalities (and injuries, and crashes in general), NHTSA has a new query tool (Fatality and Injury Reporting System Tool, or FIRST) which is fairly easy to use: https://cdan.nhtsa.gov/query Refer to the "sample queries" on the right side of the page for a quick start (you may then modify the sample query on your own). Potential areas to explore: Seat belt use, blood alcohol content, speeding, time of day/light condition/atmospheric conditions, rollover, type of vehicle, age of driver, etc etc. A surprising level of granularity is available (see the "...by State" sample query as an example), and geographic mapping is available for some queries (see the "...Police Reported..." sample query).

    • ToolGuy ToolGuy on Jan 11, 2020

      16,212,576 people involved in motor vehicle crashes in the U.S. in 2018 (roughly one in twenty).

  • Redapple2 Another bad idea from the EVIL gm Vampire.
  • Daniel J Alabama is a right to work state so I'd be interested in how this plays out. If a plant in Alabama unionized, there are many workers who's still oppose joining and can work.
  • ToolGuy This guest was pretty interesting.
  • NJRide So this is an average age of car to be junked now and of course this is a lower end (and now semi-orphaned) product. But street examples seem to still be worth 2500? So are cars getting junked only coming in because of a traumatic repair? If not it seems a lot of cars being junked that would still possibly worth more than scrap.Also Murilee I remember your Taurus article way back what is the king of the junkyard in 2024?
  • AMcA I applaud Toyota for getting away from the TRD performance name. TuRD. This is another great example of "if they'd just thought to preview the name with a 13 year old boy."