Elon Musk: Robot Cars Should Drive Like Humans Using Vision
Elon Musk stated that he had warned the relevant parties.
Humans drive using neural networks and optical sensors, and robot cars should do the same.
The event-driven pure vision autonomous driving path has been validated, with funding flowing towards neural network and camera solutions, putting pressure on lidar-dependent companies while benefiting the Tesla ecosystem.
Source: Public Information
ABAB AI Insight
Since 2016, Elon Musk has insisted that Tesla's FSD should use pure vision and end-to-end neural networks, repeatedly criticizing lidar as an expensive and unnecessary crutch, having demonstrated its feasibility in multiple presentations.
The capital path indicates that Tesla is concentrating computing resources on Dojo training and camera data loops, motivated by the goal of replicating human driving biological mechanisms and achieving scale cost advantages, specifically through iterating models with massive real-world road data.
The evolution path of biological vision systems is similar to other automakers' shift from lidar to multimodal fusion, with the current autonomous driving industry converging from sensor stacking to data-driven neural networks.
This is essentially a technological replacement, as the mechanism of optical sensors combined with neural networks can cover human driving scenarios, simplifying hardware and reducing costs per vehicle.
ABAB News · Cognitive Laws
The human way is often the optimal solution
Warnings, when unheeded, become validations
Neural networks will eventually replace redundant hardware