Tesla rolls back FSD v10.3, Elon Musk says "rolling back to 10.2 temporarily"

Tesla's latest self-driving software was rolled out on October 23 and 24, called FSD v10.3. While the release notes mentioned a long list of improvements like vehicle semantic detections, static obstacle control and speed profile, the new software does not seem to be performing well. The new update was rolled back just a few hours after the launch and Elon Musk himself tweeted about the same. 

In his tweet, Tesla CEO says "Seeing some issues with 10.3, so rolling back to 10.2 temporarily." The new version was supposed to improve several aspects of the self-driving mechanism of Tesla, including detection of brake lights, signals, improved offsetting for pedestrians and hazard lights from other vehicles on the road. However, issues reported include the disappearing Autosteer option, traffic-aware cruise control issues and AutoPilot panic. Owners have also criticised Tesla's software issues over social media platforms. 

Tesla rolls back FSD v10.3 on October 25, 2021

A report by The Verge states that it is not clear which issue caused the rollback and how widespread the issues were. The Full Self Driving v10.3 was released as a beta version for testers and due to the bugs in the software, their safety score is fallen so low that they might not be able to access the beta anymore. The report also mentions that FSD v10.3 rollback has entirely removed the full self-driving capabilities of their cars. It is important to note here that Tesla says even with the FSD functioning properly, Tesla's vehicles will require supervision, hence they would not be entirely autonomous. 

Tesla FSD v10.3 release notes 

• Added FSD Profiles that allow drivers to control behaviours like rolling stops, exiting passing lanes, speed-based lane changes, following distance and yellow light headway. 
• Added planning capability to drive along oncoming lanes to manoeuvre around patch blockage. 
• Improved creeping speed by linking speed to visibility network estimation and distance to encroachment point of crossing lanes. 
• Improved crossing object velocity estimation by 20% and yaw estimation by 25% by improving the surround video vehicle network with more data. 
• Improved vehicles semantic detections 
• Improved static obstacle control 
• Reduced false slowdowns 
• Improved turning profile for unprotected turns 
• Improved speed profile