Tesla’s Full Self-Driving 10.5 beta release notes suggest phantom slowdowns and turning confusion seen in the 10.4 beta might be improved.
Tesla‘s 10.5 beta release notes for its Full Self-Driving software suggest a huge improvement might be seen over the previous 10.4 update. Since this is still considered a beta, access is limited and Tesla is taking extra precautions at this early stage in testing by requiring participants to achieve a certain degree of safety, as measured by the company’s automated insurance calculator. The latest version has started rolling out and it looks like it might solve many earlier issues.
Telsa’s FSD 10.4 beta was expected to be a big step forward, yet disappointed many testers that felt it was a step backward. Problems included phantom slowdowns for no apparent reason and hesitancy when making turns at what appeared to be easy intersections with little traffic. Identification and predictive action based upon pedestrians and bicycles were the focus of the 10.4 update and pedestrian interactions were much improved, even erring on the side of being cautious and courteous. Moving carefully around vulnerable pedestrians and two-wheeled traffic is very important, but so are turns.
An early look at Telsa’s Full Self-Driving (FSD) 10.5 beta release notes, via Dirty Tesla, show that this update is aiming to solve several troubling problems that plagued the 10.4 and earlier beta releases. Tesla’s FSD 10.5 update should be rolling out soon to drivers with a safety score of 98 or higher, according to Elon Musk. Tesla employees are getting to try it before other customers and the first YouTube video has already been posted by AI Addict, showing the car navigating a relatively quiet residential neighborhood with ease. The driver notes that the FSD 10.5 beta hesitates a bit longer than necessary at intersections and has a couple of phantom slowdowns for no apparent reason but otherwise performed well. The greater challenges will come from channels that have routes with construction and unusual lane markings that have been known to cause confusion with previous FSD betas.
According to the release notes, 165,000 more auto-labeled videos, which are used for training the artificial intelligence were added to the Tesla computer to better identify road lines, edges, and lane connections. Videos of previous beta versions showed the system getting confused by unusual lane markings, so this update might bring a big improvement. 15,000 videos were used to enhance traffic cone and sign detection, which was surprisingly inaccurate as noted in videos revealing an FSD-driven Tesla heading toward an obvious road closed sign without any slowdown, causing the human driver to take over and brake hard to avoid a collision. False slowdowns were reported in the FSD 10.4 beta and that should be improved as well, although a few were noted in the first 10.5 beta video posted.
Merging at the last minute also caused several problems in earlier versions and Tesla’s improved FSD 10.5 beta is expected to increase smoothness. Finally, high-curvature roads led to the car’s FSD system asking for help, but the new AI modeling takes sharper turns into consideration and should help with navigating the most crooked road in San Francisco. It will be interesting to see how well Tesla’s FSD 10.5 beta holds up in more rigorous tests which should begin to appear soon.