Connect with us

News

Very little of Tesla FSD distance driven is actually useful for training: Musk

Credit: @WholeMars/YouTube

Published

on

Tesla’s Full Self-Driving (FSD) has been making headlines in recent weeks, as the company expands the software program in preparation for both its upcoming robotaxi launch and its launch outside of North America. This week, however, CEO Elon Musk has commented on how few of the miles driven with the system engaged are actually useful for training the neural net-based model, emphasizing the need for as many training miles as possible.

In response to an investor discussion about FSD shared on X on Tuesday, Musk wrote that only one out of every 10,000 miles of distance driven is actually useful for training the FSD neural network, which may come as surprisingly low to some. He also notes that Tesla and FSD are no longer AI training compute-constrained, but rather, that validation of driving circumstances is now the system’s most significant limiting factor.

You can see Musk’s post below, along with the original 30-minute discussion he was responding to.

With Tesla’s shift to end-to-end neural network vehicle control in FSD v12, the need for increased validation makes sense, given that the model is now being trained on millions of clips of video from real-world human driving scenarios—rather than on human-written code and “rules-based” processing.

Advertisement

As Musk highlights, interventions are also becoming rarer as the software improves from this camera- and video-based training, predicating the need for increased validation, especially to handle unique road scenarios. Tesla has been on an all-out offensive with FSD, recently dropping the beta moniker as v12 has been rolling out to the public, and instead replacing it with the term “Supervised.”

The company has also set a date for the reveal of its robotaxi platform, slated for August, and it has gained its first approval for the software outside of North America. In addition, the company has been rolling out one-month free trials and FSD demo drives for new vehicle buyers, substantially increasing FSD miles driven in recent months.

Just weeks after Tesla’s FSD surpassed 1 billion miles driven cumulatively, and after Tesla launched FSD demo drives and free trials:

Tesla FSD users pass 1.3 billion cumulative miles

Advertisement

What are your thoughts? Let me know at zach@teslarati.com, find me on X at @zacharyvisconti, or send us tips at tips@teslarati.com.

Zach is a renewable energy reporter who has been covering electric vehicles since 2020. He grew up in Fremont, California, and he currently lives in Colorado. His work has appeared in the Chicago Tribune, KRON4 San Francisco, FOX31 Denver, InsideEVs, CleanTechnica, and many other publications. When he isn't covering Tesla or other EV companies, you can find him writing and performing music, drinking a good cup of coffee, or hanging out with his cats, Banks and Freddie. Reach out at zach@teslarati.com, find him on X at @zacharyvisconti, or send us tips at tips@teslarati.com.

Advertisement
Comments

News

Tesla FSD (Supervised) fleet passes 8.4 billion cumulative miles

Tesla’s Full Self-Driving (Supervised) system has now surpassed 8.4 billion cumulative miles.

The figure appears on Tesla’s official safety page, which tracks performance data for FSD (Supervised) and other safety technologies.

Tesla has long emphasized that large-scale real-world data is central to improving its neural network-based approach to autonomy. Each mile driven with FSD (Supervised) engaged contributes additional edge cases and scenario training for the system.

The milestone also brings Tesla closer to a benchmark previously outlined by CEO Elon Musk. Musk has stated that roughly 10 billion miles of training data may be needed to achieve safe unsupervised self-driving at scale, citing the “long tail” of rare but complex driving situations that must be learned through experience.

The growth curve of FSD Supervised’s cumulative miles over the past five years has been notable. 

As noted in data shared by Tesla watcher Sawyer Merritt, annual FSD (Supervised) miles have increased from roughly 6 million in 2021 to 80 million in 2022, 670 million in 2023, 2.25 billion in 2024, and 4.25 billion in 2025. In just the first 50 days of 2026, Tesla owners logged another 1 billion miles.

At the current pace, the fleet is trending towards hitting about 10 billion FSD Supervised miles this year. The increase has been driven by Tesla’s growing vehicle fleet, periodic free trials, and expanding Robotaxi operations, among others.

With the fleet now past 8.4 billion cumulative miles, Tesla’s supervised system is approaching that threshold, even as regulatory approval for fully unsupervised deployment remains subject to further validation and oversight.

Published

on

Credit: Tesla

Tesla’s Full Self-Driving (Supervised) system has now surpassed 8.4 billion cumulative miles.

The figure appears on Tesla’s official safety page, which tracks performance data for FSD (Supervised) and other safety technologies.

Tesla has long emphasized that large-scale real-world data is central to improving its neural network-based approach to autonomy. Each mile driven with FSD (Supervised) engaged contributes additional edge cases and scenario training for the system.

Credit: Tesla

The milestone also brings Tesla closer to a benchmark previously outlined by CEO Elon Musk. Musk has stated that roughly 10 billion miles of training data may be needed to achieve safe unsupervised self-driving at scale, citing the “long tail” of rare but complex driving situations that must be learned through experience.

The growth curve of FSD Supervised’s cumulative miles over the past five years has been notable. 

Advertisement

As noted in data shared by Tesla watcher Sawyer Merritt, annual FSD (Supervised) miles have increased from roughly 6 million in 2021 to 80 million in 2022, 670 million in 2023, 2.25 billion in 2024, and 4.25 billion in 2025. In just the first 50 days of 2026, Tesla owners logged another 1 billion miles.

At the current pace, the fleet is trending towards hitting about 10 billion FSD Supervised miles this year. The increase has been driven by Tesla’s growing vehicle fleet, periodic free trials, and expanding Robotaxi operations, among others.

With the fleet now past 8.4 billion cumulative miles, Tesla’s supervised system is approaching that threshold, even as regulatory approval for fully unsupervised deployment remains subject to further validation and oversight.

Continue Reading

Elon Musk

Elon Musk fires back after Wikipedia co-founder claims neutrality and dubs Grokipedia “ridiculous”

Musk’s response to Wales’ comments, which were posted on social media platform X, was short and direct: “Famous last words.”

Published

on

UK Government, CC BY 2.0 , via Wikimedia Commons

Elon Musk fired back at Wikipedia co-founder Jimmy Wales after the longtime online encyclopedia leader dismissed xAI’s new AI-powered alternative, Grokipedia, as a “ridiculous” idea that is bound to fail.

Musk’s response to Wales’ comments, which were posted on social media platform X, was short and direct: “Famous last words.”

Wales made the comments while answering questions about Wikipedia’s neutrality. According to Wales, Wikipedia prides itself on neutrality. 

“One of our core values at Wikipedia is neutrality. A neutral point of view is non-negotiable. It’s in the community, unquestioned… The idea that we’ve become somehow ‘Wokepidea’ is just not true,” Wales said.

Advertisement

When asked about potential competition from Grokipedia, Wales downplayed the situation. “There is no competition. I don’t know if anyone uses Grokipedia. I think it is a ridiculous idea that will never work,” Wales wrote.

After Grokipedia went live, Larry Sanger, also a co-founder of Wikipedia, wrote on X that his initial impression of the AI-powered Wikipedia alternative was “very OK.”

“My initial impression, looking at my own article and poking around here and there, is that Grokipedia is very OK. The jury’s still out as to whether it’s actually better than Wikipedia. But at this point I would have to say ‘maybe!’” Sanger stated.

Musk responded to Sanger’s assessment by saying it was “accurate.” In a separate post, he added that even in its V0.1 form, Grokipedia was already better than Wikipedia.

Advertisement

During a past appearance on the Tucker Carlson Show, Sanger argued that Wikipedia has drifted from its original vision, citing concerns about how its “Reliable sources/Perennial sources” framework categorizes publications by perceived credibility. As per Sanger, Wikipedia’s “Reliable sources/Perennial sources” list leans heavily left, with conservative publications getting effectively blacklisted in favor of their more liberal counterparts.

As of writing, Grokipedia has reportedly surpassed 80% of English Wikipedia’s article count.

Advertisement
Continue Reading

News

Tesla Sweden appeals after grid company refuses to restore existing Supercharger due to union strike

The charging site was previously functioning before it was temporarily disconnected in April last year for electrical safety reasons.

Published

on

Credit: Tesla Charging

Tesla Sweden is seeking regulatory intervention after a Swedish power grid company refused to reconnect an already operational Supercharger station in Åre due to ongoing union sympathy actions.

The charging site was previously functioning before it was temporarily disconnected in April last year for electrical safety reasons. A temporary construction power cabinet supplying the station had fallen over, described by Tesla as occurring “under unclear circumstances.” The power was then cut at the request of Tesla’s installation contractor to allow safe repair work.

While the safety issue was resolved, the station has not been brought back online. Stefan Sedin, CEO of Jämtkraft elnät, told Dagens Arbete (DA) that power will not be restored to the existing Supercharger station as long as the electric vehicle maker’s union issues are ongoing. 

“One of our installers noticed that the construction power had been backed up and was on the ground. We asked Tesla to fix the system, and their installation company in turn asked us to cut the power so that they could do the work safely. 

Advertisement

“When everything was restored, the question arose: ‘Wait a minute, can we reconnect the station to the electricity grid? Or what does the notice actually say?’ We consulted with our employer organization, who were clear that as long as sympathy measures are in place, we cannot reconnect this facility,” Sedin said. 

The union’s sympathy actions, which began in March 2024, apply to work involving “planning, preparation, new connections, grid expansion, service, maintenance and repairs” of Tesla’s charging infrastructure in Sweden.

Tesla Sweden has argued that reconnecting an existing facility is not equivalent to establishing a new grid connection. In a filing to the Swedish Energy Market Inspectorate, the company stated that reconnecting the installation “is therefore not covered by the sympathy measures and cannot therefore constitute a reason for not reconnecting the facility to the electricity grid.”

Sedin, for his part, noted that Tesla’s issue with the Supercharger is quite unique. And while Jämtkraft elnät itself has no issue with Tesla, its actions are based on the unions’ sympathy measures against the electric vehicle maker. 

Advertisement

“This is absolutely the first time that I have been involved in matters relating to union conflicts or sympathy measures. That is why we have relied entirely on the assessment of our employer organization. This is not something that we have made any decisions about ourselves at all. 

“It is not that Jämtkraft elnät has a conflict with Tesla, but our actions are based on these sympathy measures. Should it turn out that we have made an incorrect assessment, we will correct ourselves. It is no more difficult than that for us,” the executive said. 

Continue Reading