From The Economist, November 24:
It’s Waymo complicated than it looks
IN 1995 RESEARCHERS from Carnegie Mellon University drove 3,000 miles from Pittsburgh to San Diego without their hands on the steering wheel. The “No Hands Across America” tour marked the start of a long road trip towards autonomous driving in America. Thirty years later you can, at last, see the fruits of that journey in the proliferation of self-driving cabs in several cities across America. Soon they will be braving the less predictable weather—and the left side of the roads—in London and Tokyo.The number of American cities on the robotaxi roadmap is rising fast. Alphabet-owned Waymo, which has the world’s biggest fleet of 2,500 autonomous cars, provides paid rides in five places—Atlanta, Austin, Los Angeles, Phoenix and the San Francisco Bay Area—and plans to more than double that number next year. Elon Musk’s Tesla has expanded its rudimentary robotaxi service (which still has a human “safety monitor” in the car) from Austin to San Francisco. Zoox, an Amazon-owned company that has built a quirky-looking cab without a steering wheel or pedals, offers autonomous ride-hailing in Las Vegas and has just launched in parts of San Francisco. Like Doctor Dolittle’s pushmi-pullyu, it can go in both directions.
Many people still have concerns about the technology. In a recent YouGov survey, three-quarters of Americans said they had little or no trust in self-driving taxis. Yet their misgivings may be at least in part the result of unfamiliarity. Another poll by J.D. Power, a consumer-research firm, found that confidence was 56 percentage points higher among those who had ridden in a robotaxi than those who had not. What is more, the number of people who have given one a try is growing speedily. In April Waymo surpassed 1m monthly active users (MAUs), according to Sensor Tower, a data provider, more than a ten-fold increase in two years (see chart).
Waymo and its competitors are not the only companies lured in by the promise of the technology. Others looking to cash in include traditional ride-hailers such as Uber, carmakers such as Mercedes-Benz and Volkswagen, and technology suppliers such as Nvidia—not to mention the vibrant robotaxi ecosystem emerging in China.
They are betting that the market for self-driving technology will prove to be vast. Americans currently spend around $50bn a year on ride-hailing. Then there is the possibility of selling self-driving software to carmakers and manufacturers of other vehicles, such as long-haul trucks and delivery vans. Dara Khosrowshahi, Uber’s chief executive, has said that the market for self-driving technology could be worth $1trn or more in America alone. That raises two questions: when will the autonomous revolution hit full speed; and who will come out ahead?
To answer these questions, start with the technology. It is easy to forget the miracle of a self-driving car when sitting in a Waymo or a Zoox, because it quickly feels so familiar that you forget your life is in the hands of a computer on wheels. Unlike other robots that operate in clean, controlled environments such as factories, robotaxis have to deal with the messiness of everyday life—road rage, clueless pedestrians and pets, and weather conditions like snow that can be as soft as slush or hard as a brick wall.
Several innovations have combined to make that possible. Robotaxis rely on sensors such as cameras, laser-based lidars, microphones and radars to assess road conditions, judge distances and manage speed. They then use artificial intelligence both in the car and in the cloud to mimic the way human drivers process such information and react to it. And as robotaxi services have expanded, they have hoovered up ever more data, allowing them to refine their algorithms. The rise of multi-modal generative-AI models, which weave together text, images and sounds, has also accelerated progress, including by making it easier to train autonomous systems using simulations and teach them to react to unusual situations.
For the industry, building confidence that self-driving vehicles are safe is vital. Research conducted last year by Waymo with Swiss Re, an insurance giant, showed that its robotaxis generated 88% fewer property-damage claims and 92% fewer bodily-injury claims than the average human during 25m miles of driving, a performance which has improved further since. But public trust is easily shattered. A grisly accident in 2023 involving a robotaxi operated by Cruise, a competitor to Waymo, became existential after the company failed to provide full transparency during a federal investigation. General Motors, Cruise’s owner, subsequently shut its robotaxi service down.
Such incidents risk setting back the industry’s efforts to win over policymakers. Under President Donald Trump, the Department of Transportation has said it will develop federal regulations for autonomous vehicles, including robotaxis. But there are still myriad state-by-state rules (California has two agencies that regulate robotaxis). Some cities such as Seattle put up roadblocks as well. “Everything we do today is to build trust and safety, and we can’t scale the technology without that,” says Chris Mooney, a Waymo executive.
Still, the speed at which Waymo is accelerating its city-by-city rollout suggests it is becoming increasingly confident in the safety of its service. Turning it into a profitable business, however, is another matter. Human drivers are estimated to account for 50-70% of the cost of a traditional ride-hailing service such as Uber, so stripping them out sounds like a quick way to undercut the incumbents. Yet all robotaxi services currently lose money, including Waymo. (In addition to receiving fistfuls of cash from Alphabet, the company raised $5.6bn in a share offering to outside investors last year.)
That is partly because their vehicles are expensive; they are kitted out with costly safety features and trained on pricey AI chips. And unlike ride-hailing firms such as Uber, which rely on owner-drivers, robotaxi operators have to cover the cost both of acquiring and managing their fleets, including cleaning, fuel, maintenance and parking. They also need human supervisors to monitor their vehicles in case things go wrong.
As a result, Augustin Wegsheider of BCG, a consultancy, estimates that self-driving vehicles cost about $7-9 a mile to operate, compared with $2-3 a mile for traditional ride-hailers and $1 a mile for personal cars. Cutting costs “has not been a priority of the robotaxi companies so far”, says Philipp Kampshoff of McKinsey, another consultancy. “They haven’t even scratched the surface, because mission number one was making the operation safe.” McKinsey estimates that it will take a decade to bring costs below $2 a mile.
Cheaper hardware will contribute. Estimates for the cost of Waymo’s current generation of robotaxis range from $130,000 to $200,000 each. Using the sporty Jaguar I-Pace as the starting point hasn’t helped to keep costs down. This month the company started testing the more modest Hyundai IONIQ 5, equipped with Waymo’s newer generation of self-driving technology which requires fewer sensors. The cost of those sensors is coming down, too. American-made lidars that used to cost around $100,000 are now little more than $1,000.
The economics of robotaxis can be improved in other ways. To keep utilisation of its vehicles high, Waymo’s approach to expansion has been to deploy a relatively small number of cabs in several markets, rather than trying to saturate one in particular, though the tactic has obvious limits. Instead of managing all its vehicles, it is also starting to forge alliances with fleet operators, such as Avis, a car-rental firm, to help handle day-to-day operations.
Off to the races....
....MUCH MORE
Speaking of CMU:
May 2015 - Big Money: Uber Guts Carnegie Mellon Robotics Lab To Hire Autonomous Car Developers
June 2015 - "Uber Is Stealing Scientists, But Only So It Can Lay Off Drivers"
November 2016 - the introduction to "Interview: Manuela Veloso Head of Machine Learning, Carnegie Mellon University":
Our readers probably know Carnegie Mellon more for the top-ranked financial engineering program (Master of Science in Computational Finance) but artificial intelligence was pretty much invented at CMU by Herbert Simon and Allen Newell. Simon received the Nobel in Economics but it actually could have been for any of four or five subjects, he was quite the polymath.
Newell had to settle for the Turing award (along with Simon) from the Association for Computing Machinery, probably the root'in-tootin high-falootinest tchotchke in the computer biz.
The Association for the Advancement of Artificial Intelligence along with the ACM subsequently named an award in Newell's honor. Ditto for CMU.
The University's machine learning department was the first in the world to offer a doctorate and as far as I know is still the largest.
A department, for one branch of AI.
Carnegie-Mellon used to have a world class robotics Institute but Uber gutted it with a combination of cash and stock options leaving a Dean and a couple robots to rebuild.
One of the robots is said to be in advanced negotiations with the Ube-sters....
