Well, some are, anyway, and there is a lesson in that. It’s been a while since I wrote on autonomous vehicles, but there has been quite a lot of action lately in this space, and I thought it was a good time to bring everyone up to speed.

Waymo’s Level 4 system continually advances

Once upon a time, Tesla was thought to be the industry leader in autonomous vehicles. Their 2016 “Paint It Black” demo convinced most of the world that Tesla was on the verge of having the technology to get people autonomously to and from any destination that could be mapped.

Mind Matters News, however, was skeptical. The problem is simple—Tesla was betting on the wrong horse. There are multiple levels of autonomous vehicles, and Tesla was aiming for the top one—Level 5 autonomy—the ability for a car to drive anywhere a human can. The problem, however, is that Level 5 autonomy is not an engineering project, it’s a philosophy project.

We have noted the problems with Level 5 autonomy for a number of years. In 2020, we pointed out that Waymo, owned by Alphabet which also owns Google, despite not being as flashy, was picking the right goal. Instead of reaching for Level 5, Waymo chose to target Level 4 autonomy. In Level 4 autonomy, the vendor specifies the area and conditions under which the vehicle operates, and operation is only allowed within this area. This approach allows the vendor to create high-precision maps and generate the necessary infrastructure and tooling required to navigate the specified areas. It converts self-driving from a philosophy project to an engineering project.

This choice of target has paid off in huge dividends for Waymo. While the Tesla autonomous vehicles division has possibly gotten worse, Waymo is steaming ahead, offering driverless rides in San Francisco, Phoenix, Austin, and, soon, Los Angeles. These are fully autonomous vehicles with no safety driver.

How does Waymo do this? The firm’s extensive mapping technology gives its Waymo Driver technology “a deep understanding of its environment, from road types and the distance and dimensions of the road itself, to other features like lane merges, stop signs, crosswalks, and so much more…not only can it sense a car that might cut across its path, but because of our custom maps, it also knows that vehicle has a stop sign.”

Additionally, the Waymo vehicles are not completely on their own. They can report back to the command center about changes in road condition. Additionally, if they encounter unexpected situations, they can request remote operators for help. The remote operators don’t necessarily joystick the car, but instead provide additional hints to the car about its environment so it can take the correct actions in a given situation.

Waymo’s next target is the highway. Many users have complained that its Phoenix service is too slow because it takes only city streets. While initially the highway will be available only to employees, Waymo has started rolling out its technology for including the highway system as part of its vehicles’ operational area.

GM Cruise Gets Kicked Out of California

While Waymo jumped ahead of Tesla by picking the right autonomy target, it jumped ahead of Cruise by prioritizing safety. After a string of safety incidents involving GM Cruise vehicles, last year California altogether revoked GM’s permit to operate autonomous vehicles. The reason wasn’t solely safety concerns, but also that Cruise was found to have hidden video footage from a safety investigation in which a pedestrian was critically injured.

While this episode has been a black eye for the industry as a whole, most people recognize that it was triggered more by Cruise management’s misbehavior than anything else. On the whole, Waymo’s commitment to safety has shielded it from a lot of the pushback that results from these incidents.

Waymo has published data showing that its systems are safer than the average driver. While I don’t take that data very seriously (there is a huge selection bias in the fact that Waymo can choose where the cars drive), it has done a lot to show the public both that the company is concerned about safety and that it is taking effective actions to enable safe vehicles.

Some Californians Are Tired of Being Experimented On

There has been some general pushback, however. In August, the San Francisco city attorney is attempting to reverse recent rulings that have allowed autonomous vehicles to operate in the state. More recently, the city of San Francisco filed a lawsuit to get its public utilities commission to reconsider whether to allow Waymo to operate its driverless services there.

Adding to the threat of litigation is the fact that autonomous vehicles, at least in California, seem to be above the law. It turns out that there is a vicious double standard when it comes to rules for vehicles. The San Francisco police chief had instructed officers to not issue citations for driverless cars. As has happened many times in the past, big tech gets away with violations which would cost an ordinary person dearly.

Irina Raicu, a program director at the Markkula Center for Applied Ethics, says that the emergence of autonomous vehicles in San Francisco “has the feel of a human subject mass experiment… without the kind of consent that we usually want to see as part of that.” This captures a lot of the angst that many people feel about autonomous vehicles.

Are autonomous cars profitable?

One question that hasn’t been answered, and probably won’t be for a long time, is whether or not autonomous taxi services can be profitable. Cruise lost over 2 billion dollars last year, and is now essentially out of the game. Alphabet is also betting big on Waymo. While Alphabet doesn’t break out the numbers specifically, it is estimated to be losing over a billion dollars every year.

The problem for getting to profitability isn’t just market share; it is that not every part of the business operates at an economy of scale. Every new market requires additional continual investment in keeping the maps and routes up-to-date and keeping remote drivers around to help the cars make decisions.

What I personally believe is necessary, to cause a shift in the profitability calculus, is municipal establishment of advanced infrastructure. If the roads become smart roads, then vehicles can be programmed to better accommodate the changing infrastructures within a city.

Conclusion

Other current technology ventures could learn a lot from looking at what has gone on in the autonomous vehicle space. What works well is (a) focusing on reachable engineering targets rather than trying to make an “everything” system, (b) a focus on consumer safety and well-being, and (c) not taking the public for granted—they are not only the customers; their political pressure, if applied, helps make the rules of the road as well.