The engineers who worked at FSD do not trust their own creation

Elon Musk has been ensuring for a decade that full autonomous driving is just around the corner. Although the company has advanced in its driving assistance systems, a Reuters investigation reveals something worrying. Several people who worked on this project have denounced that the technology continues to suffer from basic and dangerous errors, and they confess that they would not get into a Tesla autonomous car for the world.

what has happened. In this investigation, Reuters had the testimonies of nine “data taggers”—the people who train that Tesl AI system—as well as a software engineer who worked on the project. According to them, vehicles with these systems collide with animals, ignore the presence of school buses or accelerate in construction zones. One of the team’s veterans summed up everything in one sentence: “we have all seen the FSD fail.”

Beware of public demos. Tesla has already launched robotaxi pilot programs in cities like Austin (Texas). Musk claims that his software is a generalized system that can be adapted to any city without high-precision maps, but these interviewees indicate that the operational reality is different.

The trick. Tesla staff spent months recording videos and mapping the area of ​​Austin where the tests were to take place, and they spent hundreds of hours labeling curbs or road markings just to avoid problems during the demonstrations. According to these former employees, this level of intervention is unaffordable on a global scale.

Comparing pears with apples. To maintain that the FSD system is up to ten times safer than human driving, Tesla uses a methodology criticized by experts. For example, he compares his cars (4.1 years old on average, modern safety systems) with the average American car, which is almost 13 years old. Phil Koopman, a professor at Carnegie Mellon University, explained that “It’s like saying my jet plane is faster than a World War II bomber.” The data reveals that if only accidents with airbag deployment are compared, Tesla’s advantage would not be 10 to 1, but 3 to 1, and even that figure is questionable.

The controversial “Mad Max” mode. Internal videos have shown Tesla cars driving at speeds much higher than those allowed after the introduction of certain aggressive driving modes like the so-called “Mad Max”. Some of the employees who participated in the investigation reported cars traveling at almost 100 km/h in zones limited to 40 km/h. This aggressive driving is often treated as a low priority problem by its engineers, despite the risk it poses to road safety in these urban environments.

Investigations in progress. The National Highway Traffic Safety Administration (NHTSA) currently has four open investigations into FSD and Autopilot. These cases include situations in which Tesla vehicles ignored red traffic lights or they turned directly into oncoming traffic. Fatal accidents that occurred are also being investigated in low visibility conditions —fog, sun glare—, and where the Tesla sensors, which are focused entirely on the use of cameras, have turned out to be insufficient.

Where are the robotaxis? Almost a year after its launch in Austin, Tesla’s fleet of robotaxis it’s still tinyand consists of about 50 vehicles. It is also limited to very specific areas, and in cities like Dallas or Houston, users have complained that the cars do not drop them off at their exact destination. Besides, many of these vehicles They still have human drivers in the passenger seat who are there to avoid problems. It’s a reasonable practice, but it destroys the promise of full unattended autonomy that these vehicles offer.

In Xataka | Elon Musk has come up with two names for Tesla’s self-driving taxi. And legally you can’t put any on it

Leave your vote

Leave a Comment

GIPHY App Key not set. Please check settings

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.