Tesla might be focusing resources dedicated to working on its Full Self-Driving (FSD) software disproportionately on data from CEO Elon Musk and prominent FSD influencers, potentially skewing the development and performance of its FSD software.
According to a report from Business Insider, citing several sources, Tesla’s data annotators have been instructed to prioritize data from Musk’s vehicles and a select group of “VIP” drivers. This special treatment involves scrutiny of video clips and images from these vehicles, ensuring that any issues encountered are swiftly addressed.
According to over a dozen current and former Tesla employees that spoke with the publication, this practice has led to the FSD software being optimized for routes frequently traveled by Musk and these high-profile users.
One former employee stated, “It seemed like we were purposely making his car better to make Autopilot look different than it was. It felt dishonest.”
Data annotators at Tesla review camera footage from Tesla vehicles and label various elements to train the AI on how to respond in different driving scenarios. This includes identifying stop signs, lane markings, and other crucial aspects of driving.
However, the focus on data from Musk means that routes around Tesla’s key locations, such as its factories and Musk’s residences, receive more attention. Workers have recounted instances of labeling data from routes around Musk’s properties and frequently visited locations, ensuring that the FSD software performs optimally in these areas.
Tesla’s prioritization extends beyond Musk to include influencers who regularly share their experiences with the FSD software online. These influencers, who often push the system to its limits, have their data scrutinized more heavily to address any issues they encounter. A good example of this is Chuck Cook’s infamous unprotected left turn, which draws a number of test vehicles to drive the same route ahead of a release.
This practice, while beneficial in refining the software for challenging scenarios, means that these influencers experience a more refined version of the FSD software compared to the average Tesla driver.
John Bernal, a former Autopilot analyst and test driver, shared that Tesla sent multiple test drivers to routes popularized by YouTubers to address specific issues. For instance, after Tesla Raj posted a video of FSD struggling on San Francisco’s curvy Lombard Street, Tesla coded invisible barriers into the system to navigate the route more effectively.
Tesla’s approach to self-driving technology has not escaped regulatory scrutiny. The National Highway Traffic Safety Administration (NHTSA) is investigating Tesla’s Autopilot and FSD systems in connection with numerous crashes and fatalities. Additionally, the US Justice Department is probing whether Tesla misled investors and consumers about its vehicles’ self-driving capabilities.
Despite these challenges, Musk remains optimistic about the future of Tesla’s self-driving technology. He has frequently stated that achieving full autonomy is vital for Tesla’s success, with plans to launch a robotaxi service in the near future.
