Waymo published a list of 10 artificial intelligence lessons it derived from operating its vehicles for more than 200 million miles in fully autonomous driving, but one of those lessons appeared to be a direct criticism of Tesla’s strategy, even without naming it. Waymo says that attempting to turn a driver-assistance system into a fully autonomous driving system represents a “false summit”; according to the company, Level 4 maturity can be reached safely only through a system designed for that purpose, tested on closed courses, and then improved through actual operation without a backup human inside the vehicle.
These points appeared in a post written by Srikanth Thirumalai, head of AI Foundations at Waymo, and the 10 lessons focus on safety and the experience accumulated from fully autonomous trips. The company argues that billions of simulated miles or millions of miles driven by the system with a human driver present are not equivalent to what the system learns when it operates without a person who can intervene and correct an error.
An engineering dispute over the path to Level 4
The essence of the disagreement is not only the amount of data, but also the nature of the data and the environment in which it is collected. Tesla relies on a fleet of customer vehicles to train its Full Self-Driving system, while drivers remain legally responsible for the vehicles. Waymo’s position, by contrast, assumes that rare and dangerous situations, or what is known as the long tail of edge cases, do not appear in their full difficulty until there is no human to contain a system failure.
Waymo also warns that cameras alone are insufficient to achieve safe autonomous driving at scale, pointing to its use of a combination of cameras, lidar, and radar to provide a degree of sensor redundancy. Tesla, by contrast, relies on the camera-based Tesla Vision system. Waymo also criticizes end-to-end neural networks that convert raw camera inputs directly into steering commands, arguing that a closed or “black box” system makes it more difficult to build trust in it.
The gap appears in real-world operations
Waymo’s confident tone is based on operational indicators cited in the article. The company provides more than 500,000 fully paid rides without a human driver each week in its U.S. markets, and is targeting one million rides per week before the end of the year.
By contrast, Tesla’s Robotaxi service continues to operate on a limited scale following its launch in Austin, with human safety monitors inside the cars. According to the article, Tesla also says it has driven only 380,000 miles of unsupervised driving, a figure Waymo compares with what it accomplishes in a single day. The article adds that Elon Musk has repeatedly predicted the spread of unsupervised driving before the end of the year, while moving this target more than once.
What does this mean for the market?
From certi.news’s perspective, the importance of this debate lies in how it shows that competition in autonomous driving is not only about improving an artificial intelligence model or increasing fleet size. There is a fundamental difference between an assistance system that depends on a human remaining ready to intervene and an autonomous transportation service that assumes responsibility for the trip in the absence of a driver.
However, this does not settle the technical dispute. The source presents Waymo’s position and the article author’s conclusions, but it does not provide an independent test proving that the company’s approach is the only path to Level 4. Nor does the text contain clear evidence that Tesla has actually come close to broadly deploying unsupervised driving. The open questions therefore remain tied to each approach’s ability to handle rare cases, scale the service, and maintain safety beyond the current operating scope.
Editorial opinion: The phrase “false summit” clearly describes the risk Waymo sees in the gradual transition from driver assistance to full autonomy. Its current operational figures support its position in terms of field experience, but they are not sufficient on their own to prove that the path Tesla is following is impossible. The decisive measure will be verifiable performance once human monitoring is removed and the service is scaled, not promises or software updates alone.