tesla self driving car debate

While most self-driving car companies rely on expensive laser sensors and detailed 3D maps, Tesla‘s taking a different path. The electric car maker uses cameras and basic 2D maps instead. Elon Musk calls lidar technology “stupid, expensive and unnecessary.” This approach sets Tesla apart from competitors like Waymo and Cruise.

Tesla’s system learns from over 6 million drivers. The company’s neural networks train on three billion miles of real-world driving data from April 2020. Meanwhile, competitors use computer simulations for tens of billions of miles of training. Tesla built custom computer chips for its cars starting in March 2019. They also created a supercomputer called “Dojo” to train their neural networks.

Tesla’s neural networks train on three billion miles of real-world driving data from six million drivers.

The basic Autopilot feature includes flexible cruise control that keeps safe distances from other cars. It can slow down to complete stops in traffic. The system also centers the car in its lane and follows highway curves. Tesla’s Full Self-Driving Supervised adds more features. It can find parking spots and park automatically when drivers reach their destination. The technology uses GPS data and road information along with camera feeds.

Despite these capabilities, Tesla’s system isn’t truly autonomous. It’s classified as Level 2 automation, meaning drivers must always pay attention. While Mercedes-Benz offers a Level 3 system in limited areas, Tesla remains at Level 2 across all markets. The Insurance Institute for Highway Safety rated Tesla’s Autopilot and Full Self-Driving as “poor” in March 2024. Nine out of twelve tested systems received the same rating in the industry’s first partial automation safety report. In February 2023, NHTSA recalled 362,758 vehicles equipped with FSD Beta due to safety concerns from crashes and fatalities.

The technology works best with clear lane markings on both sides of the road. Drivers find it helpful for long trips and heavy traffic. It reduces fatigue during commutes. Users say they quickly learn to trust the system, though they must stay alert. The 2025 hardware shows improvements over earlier versions. Tesla reports that crashes occur every 7.44 million miles with Autopilot installed, compared to human drivers who crash approximately every 702,000 miles.

Tesla’s approach differs from the industry standard. The company avoids expensive sensors and detailed maps. Instead, it relies on cameras and learning from millions of drivers. As more people use Full Self-Driving, the system collects data for Tesla’s planned robotaxi service. Whether this strategy will achieve true self-driving remains uncertain.