While Tesla’s Full Self-Driving system promises eventual autonomous driving, it currently operates at Level 2 automation and requires drivers to keep their hands on the wheel and eyes on the road at all times. The system doesn’t actually drive itself despite its name. Tesla’s aiming for Level 5 autonomy, which would mean cars could drive anywhere without human help.
Tesla’s approach differs from most competitors. The company relies only on cameras to see the road, while others use expensive lidar sensors that bounce laser beams off objects. Elon Musk called lidar “stupid, expensive and unnecessary.” Instead, Tesla’s neural network learns from over six million Tesla drivers on real roads. By April 2020, Tesla had collected three billion miles of driving data from customers.
Tesla ditches expensive lidar sensors, betting on cameras and billions of miles of real-world driving data instead.
The company built its own computer chip for self-driving, installing it in cars since March 2019. Tesla also created Dojo, a supercomputer that trains the neural network. Musk claims the hardware can already handle full self-driving, but the software isn’t ready yet. The latest Hardware 5, announced in June 2024 and scheduled for January 2026 release, promises to be ten times more powerful than the current Hardware 4.
Safety ratings haven’t been kind to Tesla’s system. In March 2024, the Insurance Institute for Highway Safety gave Tesla’s Autopilot and FSD a “poor” rating for safety features. Nine out of twelve tested systems got this lowest rating. The institute found Tesla’s driver monitoring and safeguards weren’t good enough. Many Tesla owners believe basic features like auto parking should be standard rather than bundled with the expensive full self-driving package.
Tesla’s strategy contrasts sharply with companies like Waymo and Cruise. Those competitors use highly trained test drivers and detailed 3D maps. They’ve run tens of billions of miles in computer simulations. Tesla instead uses regular drivers’ habits and basic 2D wayfinding maps. However, data shows that human drivers crash approximately every 702,000 miles, highlighting why automated systems could potentially improve road safety.
The big question remains whether Tesla’s unique approach will work. The company’s betting that millions of cars collecting real-world data will beat competitors’ careful testing methods. Tesla’s neural network keeps learning from every mile its customers drive.
But until the software catches up to the hardware’s capabilities, drivers must stay alert and ready to take control. The technology’s future depends on whether Tesla can turn its massive data advantage into truly autonomous vehicles.