How many trucking jobs will self-driving trucks eliminate? All of them, if Uber subsidiary Otto has its way. What about Embark, last week’s alleged “Otto-killer”? Hard to tell from the vague press release regurgitations. But one company has just emerged from stealth mode with a genuinely fresh take on self-driving trucks—the first one to make truckers allies instead of enemies. It’s called Starsky Robotics.
And how is it doing that, exactly? By inverting the traditional “disruptor” role Silicon Valley loves to crow about. Starsky hopes to use AI to augment and positively transform the truck driver’s traditional role—and to do so with the cooperation of the trucking companies and regulators their competitors have so far taunted or ignored.
If Starsky succeeds, they will provide an example of how evolution can sometimes be better than revolution. Theirs is a genuine effort to adapt technology to political and cultural realities, a strategy others would do well to emulate, as Uber is finding out in country after country.
What Sets Starsky Apart
One peculiarity of self-driving technology is that, while automakers proudly brag about investing billions to catch up with Silicon Valley’s efforts, truck manufacturers have been fairly quiet, if not downright reluctant. That’s because the politics of the trucking industry, one of the largest employers in the United States, is a minefield even larger than that of the taxi industry. And you don’t need to be Travis Kalanick to know how badly Uber’s handled the latter.
Another reason? No one know when self-driving technology will be ready for primetime. Even once it works—and even developers have yet to define what “works” actually means—a universe of local and state regulations wait to be navigated or rewritten, unless or until the Department of Transportation announces a national policy the states will honor.
Enter Starsky Robotics, whose solution of adding autonomy without endangering trucking jobs seems so obvious, it seems insane that Otto, with Uber’s apparently unlimited resources behind them, didn’t launch with Starsky’s model out of the gate.
Whereas Otto’s business plan is to lease/sell/rent brand-new self-driving trucks the day after it’s legal (at some unknown point when full Level 5 autonomy actually works) Starsky’s is to keep humans in the loop from Day One.
They just won’t be in the truck.
Starsky’s first product is a robot that retrofits to existing trucks, comprised of a series of actuators to control the gas, brakes and steering. The robot uses cameras, radar and ultrasonic sensors to see, and is connected to a remote control facility from which truck drivers will take control during the first and last mile, similar to the U.S. Air Force’s facilities from which operators control drones all over the world.
Here’s a fascinating video of one of their recent tests:
As far as why Starsky doesn’t use LiDAR like the majority of automakers, Starsky founder and CEO Stefan Seltz-Axmacher says they “don’t want to be building a system reliant on technology that doesn’t exist yet,” in an apparent jab at companies like Quanergy, whose long-awaited, low-cost solid-state LiDAR units have yet to hit the market. “We’re just trying to build this based on technology that’s readily available.”
Comma.ai’s George Hotz agrees, and so does Elon Musk, who has promised “full self-driving” Teslas without LIDAR as soon as next year.
Assuming that trio is correct, the heart of Starsky’s model isn’t their self-driving tech, but a system that uses autonomy to improve safety, efficiency and the quality of life for existing truck drivers. While Otto and Embark’s business models marinate behind the narrative of a shortage of 100,000 drivers, their plans either replace them with machines, or keep them in the truck. Starsky, by moving drivers to a drone-control facility, turns them into supervisors, enabling one “driver” to monitor many trucks, taking control only as necessary, or when a problem occurs. If Seltz-Axmacher can execute, the Starsky system will resolve the driver shortfall in an industry-wide win-win.
I’m sick and tired of startups claiming to care about safety. Using AI to improve safety makes perfect sense, but it will never happen if the end result is masses of unemployed drivers who will vote in candidates opposed to AI. This is where Seltz-Axmacher gets it.
“AI is quickly becoming ubiquitous,” he says. “Everyone’s worried about where they fit into a post-AI economy, but human beings play a really important role. Humans with AI can achieve much more than the world’s best algorithm. We’re focused on empowering drivers.”
Starsky Robotics Isn’t Really a Tech Company
A lot of things have to fall into place for Starsky to begin executing their blue-sky plan, but Level 4 autonomy isn’t one of them. Rather than accumulate test miles at investor’s expense, Starsky began hauling freight for its first customer earlier this month, in one of the only states where self-driving trucks are legal: Florida.
There’s another reason to test and haul in Florida: the weather is great. Business-friendly states with good weather will likely embrace autonomy as Florida has, and do so before the DOT develops a federal policy. Self-driving trucks will function in Florida winters long before they do in snowy Michigan, allowing a company already in the trucking industry’s good graces to generate revenue market-by-market as the technological hurdles to Level 5 autonomy continue to fall.
Starsky claims the drive you see above was done 85% autonomously, but didn’t want to discuss details of the teleoperation component, and claimed multiple disengagements, which I don’t consider relevant because definitions of disengagements are so vague as to make apples-to-apples comparisons practically meaningless. What is relevant is Starsky’s plan to generate revenue from shipments on which they’re both testing teleoperation and gathering data for their self-driving tech—two systems which must seamlessly work in concert if Starsky’s model is to succeed. In other words, Starsky isn’t really a tech company as much as a trucking company developing self-driving tech.
Seltz-Axmacher, who stuffs George Hotz’s confidence and Jack Black’s humor into a Seth Rogan-esque package, put it another way.
“We’re a staffing agency,” he said. “We’ll lease you a robot on a per-mile basis. Driver-as-a-service.”
This is why I’m so fascinated by Seltz-Axmacher and Starsky. He’s clearly not revealing everything they’ve done, nor everything they intend to do, but you can draw a straight line from their first principles to where they intend to go. By using AI to augment a sector as it exists rather than obliterate it into a new form envisioned by “disruptors,” Starsky may prove to be one of the first public-facing examples of how and why humans needn’t be sacrificed on the altar of autonomy, and can in fact benefit from it.
If only someone would build a VR-based system so every Uber driver with a gaming wheel on their desk could remotely drive cabs, we’d be onto something. I’m talking to you, Travis Kalanick.
The future isn’t binary. We face a multi-decade plateau, a mixed environment of semi-autonomous and human-augmented driving, which are not necessarily the same thing. If both can save lives but only one sacrifices people’s livelihoods, I know which is more likely to be embraced by a vast industry with a powerful union behind it.
I also know who is most likely to profit from it, if only Seltz-Axmacher can execute, but we’ll cover that in the next chapter of my visit to Starsky’s secret HQ in San Francisco.
(UPDATED to reflect Starsky claim of 85% autonomy on video test run, rather than 65%.)
Alex Roy is Editor-at-Large for The Drive, author of The Driver, and set the 2007 Transcontinental “Cannonball Run” Record in 31 hours & 4 minutes. You can follow him on Facebook, Twitter and Instagram.