Nobody should be surprised about the automotive news of the week: General Motors shutting down its Cruise robotaxi division. The die was cast when a Cruise cab dragged a pedestrian in October 2023.
Asked about the Cruise news, Phil Koopman, professor at Carnegie Mellon University, described it “a rip-the-band-aid moment.”
There is no need to sustain the fallacy that robotaxi companies, currently motoring through a city-by-city grind, will somehow nail the technology before investors get impatient.
The Ojo-Yoshida Report invited to our podcast “Chat with Junko and Bola” Koopman and Bryan Reimer, research scientist at MIT. The two men are leading experts on automated vehicles. We discussed the far-reaching impact of GM’s Cruise surrender for the future of ADAS, robotaxis and personal autonomous vehicles (PAV) for the people and companies involved.
Q1: Why wasn’t this shocking news?
Robotaxi is a problematic business model, in part because the technology is so hard, and in part because you must have a ton of money and a ton of patience.
Phil Koopman
Koopman:
Well, I think the writing has been on the wall for a long time. Robotaxi is a problematic business model, in part because the technology is so hard, and in part because you must have a ton of money and a ton of patience.
Reimer:
Driverless isn’t humanless. There is human capital all over the system, from tele-operation to sensor maintenance to software design … So, you’re displacing low cost, if not subsidized, ride share drivers with high cost, high labor pool technology. Furthermore, the life cycle cost of this technology is probably even higher than the initial development costs.
Both agreed that shutting down Cruise wasn’t a tough decision. If the OEM had to choose between “ten years of making money selling cars and ten years of losing money in robotaxi,” as Koopman said, “that’s a pretty straightforward choice.”
Q2: What does this mean to the future of robotaxi?
Reimer:
Sometime in the next century, well past my lifetime, I think automation wins. We will find ways to solve this. The big question is how many decades before that happens.
Koopman:
The kind of commitments it takes to make [robotaxi] happen is, [you have investors saying] “here’s our ten billion war chests. Catch you in a few years when it’s done.” It’s going to be many billions of dollars, five or ten, and it’s going to be city by city, grind, grind, grind.
Q3: Just last month, Waymo raised $5.6 billion. Isn’t that an affirmation of Waymo’s first commercial success?
Koopman:
$5 billion only gets Waymo a year and a half down the road, but the money is not even in the bank. The investors could change their minds. The jury is still out in Waymo right now.
I don’t see self-driving vehicles as a technology race anymore. I see a funding runway race.
Bryan Reimer
Reimer:
I go a step further. I think yesterday’s events create a bigger issue for Alphabet in raising the funding. Call it $10 billion plus to carry Waymo for the next five years. If you’re an outside investor, you’re getting even more skittish, based on what we’re talking about.
Reimer summed up: “I don’t see self-driving vehicles as a technology race anymore. I see a funding runway race. Who can build a runway to Mars?”
Q4: What would one bad mishap do to robotaxi?
Koopman noted that Waymo hasn’t had its bad mishap yet. “But if you’re out on the public road, eventually there will be a bad crash.” Cruise had their moment, TuSimple had a bad crash, although nobody got hurt. Uber ATG had a fatality. “Everyone who’s had the bad crash hasn’t made it,” Koopman noted.
Both agreed that with Waymo under operational pressure to expand fast, mistakes are more likely. Waymo really needs to pay attention and double down on safety. One bad crash could finish Waymo.
Q5: How will Tesla come into the future of robotaxi?
Reimer:
Tesla, to date, has blamed the operator. But as soon as Tesla pulls the [licensed human] driver out of the cab, I think a lot of questions are going to be asked. Can we allow these cars to hit things on the road if Tesla’s Cybercab behaves like Tesla’s full self-driving model today? The number of incidents will be so high that governmental involvement, even with Musk involved in the administration, will function.
Companies will continue to level up (i.e. Level 2, Level 2+, Level 2++, Level 3, Level 4) just like a video game. But they’re going to make the drivers responsible for safety.
Phil Koopman
Maybe it floats to the states, localities or the people putting cones on the front of the cars. Some kind of government or quasi government involvement is going to say, “Not in my backyard, darn it.”
Koopman:
Companies will continue to level up (i.e. Level 2, Level 2+, Level 2++, Level 3, Level 4) just like a video game. But they’re going to make the drivers responsible for safety [regardless of the autonomy level]. That way, their vehicles don’t have to be perfect. They can just blame the driver for not paying attention.
It’s been Tesla’s playbook. Tesla might be able to pull that off for even a Cybercab. OEM trying to do the right thing may be reluctant, but all the other companies are clearly going down that path.
Q6: Are we stuck with the world of Level 2 autonomy?
Koopman:
Just as predicted at the end of 2022, we are back at Level 2+++. That is what’s happening.
Companies are going to tell everyone that their cars are Level 2, which means the driver is responsible. When a crash happens, that’s your problem. That’s Tesla’s playbook. And we’re going to see that for many, many years.
The holy grail of automation may be 30, 40 years out.
Bryan Reimer
Q7: The holy grail of automation?
Reimer:
Now, I do think of the holy grail of automation, maybe in my lifetime, 30, 40 years out.
[But when I think of autonomy] I’m thinking of a vehicle that lets me turn on the automation on the highway under certain conditions. What I want is to get on the highway, switch on a button and take a nap. Now, I do think we will figure this out over the long haul. The question here is, how many decades?
Q8: When economic trends start to pull back from the mobility space, what must happen?
Blowing $10 billion on a robotaxi is what you have when you have free money. That’s not where we are anymore.
Phil Koopman
Reimer:
The automakers and tech companies need to get much more strategic in their deployment of capital to get over what’s going to be a rough tumble of the next 1-5 years. [They need to see] how some of the political changes in Washington exacerbate expenditures or accelerate savings.
I think that the landscape looks a lot different, and the players are going to get much more strategic in how they allocate the capital.
Koopman:
Well, the simple version is free money. We’re past free money. Blowing $10 billion on a robotaxi is what you have when you have free money. That’s not where we are anymore.
Q9: Who’s the potential winner?
I think the Zoox model of staying smaller, focused and managing deployment within Amazon’s budget is likely more successful.
Bryan Reimer
Reimer:
Waymo just announced the other day and they’re going to Miami. You need to ask if spending more money for more trips is going to get you to the point of a positive revenue stream faster, or is staying smaller, nimbler, and focusing on operational effectiveness—to get the cost down—is going to be an easier way to get there.
I think the Zoox model of staying smaller, focused and managing deployment within Amazon’s budget is likely more successful than the bigger, the better approach. If I had to pick winners here, it’s Amazon. It can afford a half a billion dollars a year for the next ten years to deliver packages more effectively… maybe people as well.
But can Alphabet afford $3-4 billion a year for the next ten years, given all the other pressures?
Q10: Closing thoughts
Reimer:
The future of ADAS has never been brighter, because when the capital is not being derailed, and with safety issues on our roads, let’s hope some of the capital, and some of this talent begin to develop the driver support features and the active safety features that can improve AEB, Reverse AEB, pedestrian detection and level two features, with ancillary safety benefits as well.
Koopman:
What I want those Cruise machine-learning experts [presumably moving over to GM] to work on is the following: When a vehicle is moving up from L2, L2+ and L3, what it needs most is not a lane-keeping algorithm. It needs a vehicle to realize this machine learning is out of its depth and it’s time to hand back to the user, because [otherwise] it’s about to make a randomly wrong decision. That’s always been the Achilles heel for safety.
Related podcasts:
Junko Yoshida is the editor in chief of The Ojo-Yoshida Report. She can be reached at [email protected].
Copyright permission/reprint service of a full Ojo-Yoshida Report story is available for promotional use on your website, marketing materials and social media promotions. Please send us an email at [email protected] for details.