Self-driving progress needs better AI, not more sensors, says Tesla AI chief
Self-driving progress needs better AI, not more sensors, says Tesla AI chief
Publish Date: 2026-02-04 00:47:00
Source Domain: www.cnbctv18.com
- Ashok Elluswamy from Tesla emphasizes that the main difficulty in achieving autonomous driving lies in artificial intelligence (AI) rather than hardware sensors.
- He argues that modern camera systems already collect sufficient data for autonomous vehicles to operate safely.
- According to Elluswamy, the real challenge is interpreting and extracting information from the data effectively, which is fundamentally an AI problem.
- Tesla’s self-driving system uses cameras and neural networks instead of multiple specialized sensors to handle perception, prediction, and planning cohesively.
- Elluswamy said real-world driving includes complex, interconnected situations that necessitate advanced AI inference for decision making, rather than basic sensor detection.
- Tesla trains its models by focusing on rare and risky driving scenarios observed in its global fleet to enhance safety performance.
- The company employs neural networks to simulate the environment and predict future road conditions, aiding both training and safety evaluations.
- Elluswamy concludes that because AI has advanced significantly, modern autonomous systems can solve the core perception problem primarily through vision, eliminating the need for additional sensors.
