My latest Tesla FSD test revealed a glaring problem with Elon Musk's camera-only approach

During an extensive trial of Tesla's Full Self-Driving (FSD) system, I uncovered significant issues that could hinder the company's ambition for fully autonomous robotaxis. My experience, predominantly in a Model 3 Performance equipped with the latest hardware, was largely positive; however, it highlighted a critical flaw in Tesla's camera-only approach. Throughout December, I utilized FSD for around 90% of my driving, including complex trips from Silicon Valley to Lake Tahoe and multiple journeys into San Francisco—all without significant problems. The system proved to be a safer alternative to human driving, allowing me to focus more on the road and less on navigation. Interestingly, my spouse and I found ourselves in fewer arguments about directions, as FSD handled route decisions seamlessly, effectively justifying the $100 monthly subscription fee. As my trial concluded, I was left contemplating the future of Tesla's autonomous capabilities. Currently, FSD operates under supervised conditions, with human oversight required at all times. Unsupervised autonomy, where cars navigate without driver intervention, remains the next major hurdle. Competing firms like Waymo have successfully implemented this in select urban areas, employing a multi-sensor approach that combines cameras, Lidar, and other technologies for enhanced safety and reliability. Tesla's strategy, while cost-effective for mass production, raises questions about its viability in unsupervised environments. My testing revealed that the front camera often accumulated condensation and dirt, particularly when facing direct sunlight, which led to alarming warnings demanding immediate driver control. Although a service appointment resolved the camera cleaning issue, the potential for such problems raises concerns about Tesla's reliance solely on camera technology. The implications of this approach are significant, especially as the company aims to deploy millions of robotaxis. The need for regular cleaning and maintenance of the front camera could burden Tesla's service resources. In response to these challenges, Tesla has begun implementing solutions, including software updates to enhance the cleaning of the camera area and recent designs that may improve moisture sealing. Furthermore, Tesla has patented innovative glare shields intended to minimize interference from sunlight, which could enhance camera performance in various conditions. While it remains uncertain how effective these solutions will be, they indicate Tesla's commitment to overcoming the obstacles tied to its camera-centric system. As the company pushes forward, the question remains: can a camera-only approach truly achieve the level of autonomy needed for widespread, safe use? I reached out to Tesla and Elon Musk for comments regarding these findings but did not receive a response by the publication deadline.