By David Galland, for The Rational Optimist Society
Table of Contents
The Story So Far
Waymo’s Approach
Tesla’s Approach
What About Safety?
The Great AV Disruption
“I don’t know if you have ever seen a man who has just arisen from a good night’s rest, braced and refreshed, ready for anything. If not, go out and find one. He is a magnificent spectacle.”
—Joy in the Morning, PG Wodehouse
After a 50-year period of stagnation, the US—always a world leader in technology and innovation—has reawakened its animal spirits and gotten back to work.
And so it is that autonomous vehicles—or self-driving cars—are headed for a city near you.
In the following Deep Dive, I’ll update you on the current state of the autonomous vehicle revolution so you can begin to prepare for its arrival.
As you’ll read, it’s coming far faster than you might imagine, and it may be one of the most disruptive technologies we’ll experience in our lifetimes.
David Galland
***
I confess to not being a “car guy.”
I know, I know… not very American, or macho, or whatever. But it is what it is.
Thus, when we lived in Chile many years ago, we had a full-time chauffeur by the name of Herman.
Not because we had a lavish amount of money, but because the cost of labor in Chile was so cheap back then. Given my firm dislike of driving, especially on the congested roads of Santiago, it was a no-brainer.
Regardless of your own relationship with automobiles, you may relate to the idea of being able to settle into the back seat of a car with your newspaper at hand and be safely delivered to your destination by the artificial intelligence (AI) equivalent of a Herman.
In my opinion, being chauffeured around is close to the textbook definition of luxury.
Which brings us to autonomous vehicles (AV)—or self-driving cars, as they are often referred to. We’ll use those terms, and Full Self-Driving (FSD) cars, interchangeably.
It’s a topic I’ve been keenly interested in for over a decade.
For reasons I’ll endeavor to explain, AVs—which are already operating on US roads—are on the verge of widespread adoption and will trigger a wave of disruption with huge implications for our modern lives.
The Story So Far
While self-driving cars have been “just over the horizon” for over a decade, the final mile—so to speak—is only now becoming a reality, thanks to rapid improvements in AI.
It could be said without fear of rebuttal that, ultimately, the story of self-driving cars is the story of AI.
That’s because the software for self-driving cars was originally based on rules-based coding. For example, “stop when you see a red light” or “slow down when approaching an intersection.” That sort of thing.
Over the last decade, the big leap forward for AVs has been the demotion of rules-based code to a supporting role in favor of AI trained on hundreds of millions—or, in the case of Tesla, billions—of miles of real-world driving experience.
As a result, today, it is AI—interacting with sensors and cameras—doing the driving. Simply put, the exponential advances in AI we regularly hear about are being directly leveraged to create better and safer self-driving cars.
The chart below illustrates Tesla’s remarkable 100X improvement in the Disengagement Ratio—the number of times a human had to intervene by taking control—while driving in FSD mode over the course of just 2024.
Source: Freda Duan on X
Naturally, the regulators pay close attention to those metrics, looking for the point in time when the safety record of AVs becomes irrefutably better than that of us indelibly flawed humans.
We’ll come back to safety momentarily, and in some depth. But to underscore the key point: Every advancement in the computational power of AI directly translates into better, safer self-driving cars. And AI is improving at a torrid pace.
Which brings us to the two leading contenders for first-mover status in self-driving technology, and how they differ.
Waymo’s Approach
Waymo, a division of Google, relies heavily on “Light Detection and Ranging” sensors, or LiDAR.
LiDAR works like a bat using echolocation but with light instead of sound. Waymo cars emit laser beams and measure how long it takes for them to bounce back. The car’s software, and now AI, then creates a 3D map of the car’s surroundings, identifying objects like other cars, people, and buildings.
Source: Waymo
Demonstrating a “belt and suspenders” attitude, in addition to LiDAR, Waymo also uses:
Cameras to see objects and colors, such as traffic lights and road signs.
Radar to measure how far away objects are, and how fast they are traveling.
Geolocated mapping, so the car always knows where it is.
Ultrasonic sensors—mainly for close-up work, such as parking. Your current car probably has these as a standard feature.
Fusing all this information together is AI, which uses sensor inputs and geolocated mapping to drive the car.
On the surface, Waymo seems to have the right idea. After all, when hurtling down the highway enjoying a nice cuppa and the daily news in the back seat of your driverless car, the more safety measures the better.
Yet, after General Motors spent upward of $10 billion on Cruise—its entry into the AV space—it pulled the plug due to a couple of high-profile accidents in San Francisco. In one notable incident, a Cruise vehicle mowed down a pedestrian and then carried on, dragging its victim along on an unwanted ride (they lived).
The now-defunct Cruise warrants mention because it used much of the same mix of sensors as Waymo—LiDAR, cameras, and radar—though we assume there was a bug in the programming that caused it to fail so spectacularly.
Even so, every new technology has its failures, each of which provides a learning opportunity on the way to perfection.
(Correction: On the urging from an ROS member who works at GM, I would like to clarify that the pedestrian hit by the Cruise car had been first hit by another car and thrown into its path, which made the accident all but unavoidable. An independent review of the accident did, however, chastise Cruise management for a lack of transparency in reporting on the accident, attributing it to: "poor leadership, mistakes in judgment, lack of coordination, an 'us versus them' mentality with regulators, and a fundamental misapprehension of Cruise’s obligations of accountability and transparency to the government and the public.")
Tesla’s Approach
Tesla has always relied heavily on its array of cameras, which provide a full view of the car’s environment. In the early days, it leaned heavily on rules-based programming, but AI has largely replaced that in recent years.
Source: Tesla
In the same way human eyes and brains sync up nearly seamlessly while driving, Tesla Vision uses AI to analyze the camera feeds in real time and to make instantaneous driving decisions.
Tesla has also begun reintroducing radar in some models to assist in navigating unfavorable conditions, such as poor visibility. But the camera array remains the company’s focus.
While both Waymo and Tesla collect data from their self-driving cars, Tesla’s not-so-secret sauce is the staggering amount of data it collects—exponentially more data than Waymo.
More specifically, while Waymo’s robotaxis have gathered data from an impressive 30 million miles driven across a handful of test cities, Tesla has been collecting real-time driving data from all 5 million vehicles it has sold since the company’s inception. This has enabled Tesla to accumulate data from over 9 billion miles driven, including more than 1 billion miles logged in Autopilot mode.
Tesla processes all this data using state-of-the-art technology, AI, and a team of skilled engineers to develop and test updates. These updates are then wirelessly deployed to all Tesla vehicles simultaneously, or, in some cases, tested as a beta with a select group of users.
While Waymo’s technology is reportedly more expensive than Tesla’s—adding an estimated $15,000 per car—the relative safety of the two approaches will ultimately determine which technology prevails.
What About Safety?
Overall, autonomous vehicles are already safer—by a wide margin—than human drivers.
However, based on available data, is Waymo’s sensor-laden approach objectively safer than Tesla’s? Or vice versa?
The following is an excerpt from the official Waymo Blog, referencing a recent study by Swiss Re—one of the world’s leading reinsurance companies—on the safety of Waymo’s autonomous taxis:
In real numbers, across 25.3 million miles, the Waymo Driver was involved in just nine property damage claims and two bodily injury claims.
…For the same distance, human drivers would be expected to have 78 property damage claims and 26 bodily injury claims.
Turning our attention to Tesla, there is this from the company’s regularly updated Vehicle Safety Report:
In the third quarter (2024), we recorded one crash for every 7.08 million miles driven in which drivers were using Autopilot technology. For drivers who were not using Autopilot technology, we recorded one crash for every 1.29 million miles driven.
If you assume the bodily injuries happened in conjunction with the property claims reported by Waymo, which seems logical, it works out to Waymo being involved in just one accident per 2.81 million miles driven.
By comparison, according to Tesla’s Q3 2024 report, it only recorded one accident in a total of 7.08 million miles driven in FSD mode.
While Tesla’s approach might initially appear far safer, there is an important nuance to consider: during all those miles driven in FSD mode, a human was in the driver’s seat, ready to intervene if something went wrong.
Waymo doesn’t have a driver to intervene in order to avoid a crash, and the safety data reflects that.
That said, as noted in the opening paragraph of this section, it’s clear that either approach is significantly safer than having a human driver behind the wheel. Moreover, you can be confident that AV safety will only continue to improve from here.
In time—and it won’t be long—the hard data will become both compelling and irrefutable, and the regulators will be forced to concede the point, ushering in the Great AV Disruption.
The Great AV Disruption
While a lot of people are keeping the proverbial “one eye” on advancements in self-driving cars, relatively few are actively preparing for the sweeping changes the coming Great AV Disruption will usher in.
To ensure that you and other valued members of The Rational Optimist Society are better prepared than most, here is a partial and necessarily brief list of some of the ways autonomous vehicles could impact key business sectors and our daily lives…
Car Manufacturers
Global revenues ~ $3 trillion
There are currently about 1.4 billion cars on the road, with 300 million in the US alone. Eventually, all of those cars are going to be replaced by self-driving vehicles.
The stakes for car manufacturers are huge, with the losers facing an existential threat. Those who are late to the game, or who fail to impress, will dry up and blow away.
Since every Tesla sold now comes equipped for fully autonomous driving—complete with regular over-the-air updates to keep them at the cutting edge—the company may seem to have an insurmountable lead. However, one could have said the same about AOL at the dawn of the internet age, and we all know how that turned out.
And one must not overlook China, as BYD—its largest EV maker—is heavily investing in autonomous driving technologies, including partnering with Nvidia.
This from The Daily Upside:
“China has been aggressively pushing AV technology, testing more driverless cars than any other country, according to a June report from The New York Times.”
Another major disruption for car manufacturers is that widespread adoption of AVs will likely lead to a significant decline in the number of cars sold for personal use, potentially a dramatic one.
In fact, Morgan Stanley estimates that autonomous ridesharing could reduce the total number of vehicles needed in the United States by 50%.
That’s because the average car sits idle about 95% of the time. Once traveling from Point A to Point B in a driverless car becomes safer, easier, and less expensive than ordering an Uber—less expensive because there’s no driver cost—we can expect a significant percentage of people to view personal car ownership as a thing of the past.
After all, once the regulators green light self-driving cars, who will want to buy an obsolete manual-driving one? It would be like choosing a phone with a cord and rotary dial versus a smartphone. Who would do that?
Additionally, given the clear safety advantages of autonomous vehicles, it is certain that driving your own car will result in significantly higher insurance premiums and increased personal liability in the event of an accident.
“So, Mr. Jones, can you explain to the court why you were driving when the accident occurred?”
“I like driving,” mumbles the plaintiff in response.
“Ah, so simply because you like driving, you decided to drive home late at night and, missing the stop sign, ran into poor Mrs. Flaherty who is now facing a lifetime of disability?”
“Ubbada, ubbada, ubbada…”
You get the point.
Insurance Companies
Global revenues ~ $1.3 trillion
Based on the data, the introduction of self-driving cars will reduce the number of traffic accidents by upward of 90%. That will require redefining the whole idea of personal car insurance.
For starters, the amount an insurance company can charge to insure self-driving cars will be greatly reduced, as higher premiums will be impossible to defend.
In a sign of things to come, insurance companies are already beginning to offer discounted insurance policies for self-driving cars.
Moreover, there will be far fewer cars to insure, and many of those will be owned by large corporations or investor groups with substantial negotiating power. Since approximately 25% of insurance companies’ total premiums come from vehicle insurance, a significant portion of that revenue will disappear. Half? More?
Trucking
Global revenues ~ $2.7 trillion
On the positive side, trucking companies could see significant benefits from reduced costs due to a smaller workforce, fewer accidents, and lower insurance and legal expenses. On the negative side, however, approximately 3.5 million US truck drivers may face unemployment as a result.
The impact on railroads is also worth considering. Autonomous trucks, capable of operating 24/7, could become much more competitive with trains in terms of cost and efficiency for many types of goods. Rail transport requires adhering to specific schedules and incurring additional costs for trucking goods to and from rail yards. Autonomous trucks, by comparison, offer a simpler and more compelling alternative.
For large bulk shipments—iron ore and other commodities—trains will always maintain an advantage. But that leaves a lot of other “stuff,” for lack of a better word, up for grabs.
We’ll pause here for now, with Part 2 of Driving Toward Disruption set to be released in February.
In it, we’ll continue exploring the numerous sectors of the economy that will be impacted by the widespread adoption of autonomous vehicles, discuss the remaining challenges, and provide an estimate of how long it might take for this adoption to occur. We’ll conclude by examining the investment opportunities and more.
Until next time…
David Galland
For The Rational Optimist Society
Would you ride in an autonomous vehicle? Will you stop driving yourself once they are widely available? All comments and thoughts are welcome… you can share them directly with me at Galland@rationaloptimistsociety.com.
Sources and Resources: