Tesla AI chief highlights FSD v12.5.6.3 with end-to-end highway driving capabilities

Ashok Elluswamy, Vice President of AI Software at Tesla has shared some insights on the new Full Self Driving (FSD) v12.5.6.3 and its end-to-end highway capability.

Tesla has been developing a combined software stack for city and highway to provide a smooth driving experience. Other than driving, this FSD version aims to improve decision-making capability for all on-road maneuvers.

Last month, it made a debut for AI4 cars. Now, the expansion is underway and the company is working on expansion. Ashok said the new v12.5.6.3 is using end-to-end neural networks for driving across highways, city streets, and parking lots.

FSD 12 is a shift to end-to-end neural networks and this new combined software stack reflects the advancement in self-driving using this new technology. End-to-end networks can learn to perform complex tasks from raw input data, without human intervention.

The company has trained a new FSD system with millions of video footage including city and highway driving. This new combination tops the traditional approach to self-driving, which involves rules and algorithms that are manually programmed by Tesla engineers. Furthermore, the previously controlled stack is also rigid and less adaptable compared to an end-to-end neural network.

Additionally, Tesla is sending new driving styles to set speed and lane change preferences with this new FSD version. You can apply speed profiles to roads and highways with 50mph (80kph or higher speed limit).

FSD v12.5.6.3 is rolling out in batches and its availability is limited to AI4 cars but AI3 (HW3) car owners might also be waiting for end-to-end on the highway. However, Ashok has not shared any update for HW3 owners.

Meanwhile, people upgrading to this new FSD version will get vision-based attention monitoring with sunglasses, natural lane change decisions, and a new Max Speed setting.

(source)