Welcome to Off-Ramps! Today I’ll highlight three interesting pieces that I think you will enjoy reading, all on the theme of road vehicles. Please enjoy these on your morning commute, or save them for your weekend.
I had meant, at long last, to post my interview with freight specialist Sandra Rothbard, but a nasty flu that I picked over the weekend kept me from putting the finishing touches on it. So that will happen this weekend, and you may look forward to reading it on Tuesday.
Tesla Fails the Wile E. Coyote Test
Mark Rober is a prominent YouTuber, with 65 million subscribers (more than Taylor Swift, though not by much). I’ve enjoyed his work in the past; my wife and I particularly like his Squirrel Ninja Warrior series. In his latest video, he addresses a long-standing argument in the automated-driving space: is it safe for an automated vehicle to rely only on cameras, or should it have lidar as well?
My interview with Steven Waslander (embedded above) goes into some of the stakes in this debate, and how Waymo has placed its bet on lidar being a necessary part of the tech stack, while Tesla disdains it as a “crutch”, and thinks cameras alone are sufficient.
Rober, as befits a YouTube sensation, goes a different route. He takes two vehicles, one an unnamed car fitted with lidar, and his personal Tesla equipped with Autopilot and only cameras for sensors, and puts them through their paces.
He begins with a sidequest, where he uses lidar to secretly map Walt Disney World’s Space Mountain rollercoaster, and then offers a decent explainer of how lidar works. At timestamp 9.06, Rober begins his trials. He finds that both vehicles do well at spotting a humanoid figure in the roadway, and one that darts out into the vehicle’s path without warning. And they both do well in situations of extreme brightness, like when one drives into the setting sun or an opposing vehicle’s high beams. But the lidar succeeds, and the Tesla fails, in situations of fog and rain, much as our interview with Prof. Waslander would have led us to expect.
The highlight, of course, is driving into a wall painted to resemble a roadway, much as Wile E. Coyote tried to trick the Roadrunner into smashing into. The lidar vehicle was able to tell it was driving toward a wall, because it wasn’t looking at the surfaces. The Tesla, conversely, sees only surfaces, was fooled, and smashed through the wall at high speed (here is a link with the appropriate timestamp).
Artist’s conception
Fun-ruiners are quick to tell us that this experiment doesn’t really prove anything, because Tesla Autopilot is less sophisticated than the Unsupervised Full Self Driving that the company expects to roll out with the Cybercab. They also point out that Rober clearly smashed through once, rebuilt the wall, and smashed through a second time so as to leave an amusing cartoon-style cutout in it (clearly visible, pre-collision, at timestamp 15.39).
Still, I salute Rober for finding a clever, memorable way to establish the basic point that there are some safety scenarios in which lidar is clearly superior, and that we should therefore hold anyone not using lidar in their automated vehicles to a high level of scrutiny.
Speaking of which…
Waymo Talks Smack About the Tesla Cybercab
Well, not Waymo per se. It’s John Krafcik, Waymo’s CEO from 2015 to 2021, who is deprecating Tesla’s chances of introducing a viable robotaxi service.
In a recent interview with Germany’s Manager Magazine (and hat tip to Electrek for highlighting it), Krafcik pours disdain all over the Cybercab design. As per Krafcik, the design flaws include:
Failure to place sensors to maximize coverage: so not just on the sides and corners of the vehicle, but also on the roof (and, I presume, the undercarriage, to prevent incidents like the one that drove Cruise out of business)
Failure to include sensor housing that aims to keep the sensors functional and un-obscured, e.g., wipers, compressed-air nozzles, and so forth
Failure to account for accessibility issues: the Cybercab has a low-slung coupe design, which will pose challenges for the elderly, people with mobility issues, etc.
He notes tartly that
If a company were serious about building a safe robotaxi business, the robotaxi wouldn’t look anything like this prototype.
And when asked about Tesla’s announced launch of robotaxi service in Austin this June, Krafcik says
There are many ways to fake a robotaxi service.
Certainly, Krafcik may be talking his book in this interview, but to my mind his observations make up yet another data point in Cybercab skepticism. As per the embedded article above, I am already on record as being dubious that Tesla will be able to successfully launch a robotaxi business at all, let alone do so successfully in 2025. Nothing I have learned yet has encouraged me to change my mind.
Monster Trucks versus Monster Tariffs
Americans love their pickup trucks.
As a reminder: the Ford F-series truck is consistently the most popular vehicle in the United States. In 2022, on average, an F-series truck was sold every 49 seconds.
And what competes with the F-series? Why, other pickups: the second- and third-most popular vehicles were the Chevy Silverado and the Ram Pickup respectively.
At Changing Lanes, we don’t judge. People like what they like, and some people like driving trucks, even though they get relatively-terrible gas mileage, and are increasingly unhelpful at carrying things: as Axios notes, the first generation of F-150, back in the 1970s, was 36% cab and 64% bed. Today, those numbers are flipped to 63% cab and 37% bed.
Image courtesy of Axios.com.
We don’t judge, but we do condemn the safety risk these vehicles cause. The ever-increasing weight of the truck, and the height of its cab, means the Ford F-150 and its peer vehicles simultaneously make it harder for the driver to see other road users and more likely to injure them. A pickup that collides with another car is 2.5 times more likely to kill the driver than if a conventional four-door had been responsible.
These are the sorts of collective-action problems that policy is supposed to solve. Given the increased danger to others that pickups pose, it’s reasonable for regulation to impose limits on height, weight, or maximum speed of such vehicles, or even limits on their sale. To date, though, there has been no successful such regulation: Congress considered one such bill in late 2024, but it died in committee.
But what the legislature can’t achieve, the executive might—inadvertently—through tariffs. Pickup frames are mostly aluminum, to the tune of 500 pounds per vehicle on average; the auto industry is the single largest consumer of aluminum in the United States. And as of last week, it will be paying a global tariff of 25% on all aluminum imports, a commodity notorious for being unavailable in the USA: in 2024, out of nearly 5 million metric tons of aluminum consumed in the USA, only 0.68 million tons were produced domestically. This means that American-made pickup trucks are poised to become much more expensive relative to smaller, lighter vehicles.
To which I say:
Much as I would love to see safer roads from more people switching to lighter vehicles, I'm not sure if aluminum tariffs will be enough to move the needle.
At $1.20/pound, 500 pounds of aluminum cost $600. A 25% tariff only adds an extra $150, maybe 0.3% of the total cost.
While having LiDAR in a vehicle would be nice, humans drive with just two cameras (eyes). Yes, we have accidents in the fog and at night more frequently. Maybe don't make perfect the enemy of good enough?