Tesla abandones 300,000 lines of code in C++ and launches self-driving system controled by neural network

Tesla has started the rollout of the FSD Beta v12 update, which, according to CEO Elon Musk, introduces "end-to-end neural nets." The significant difference from previous versions of the FSD autopilot system is that the vehicle's control will now be carried out by neural networks rather than being coded by programmers.

"FSD Beta v12 updates the city street driving control stack to a single end-to-end neural network trained on millions of video clips, replacing over 300,000 lines of explicit C++ code," stated Tesla.

This means that both the car's behavior and its computer vision system will now operate under the control of an artificial intelligence system, rather than being predetermined by code. When Tesla first introduced Full Self-Driving, the automaker claimed that this approach would be used, but it took the company years to transition to neural network-based control.

Many Tesla enthusiasts believe that v12 is what will finally enable Tesla to fulfill its promises regarding autonomous driving. Previously, this version was only used in Tesla's internal testing fleet, but now it has finally reached customers.