Nvidia is questioning the leaders of autonomous driving systems—Tesla and Waymo.

New Approach to Autopilot from NVIDIA

Every six months the head of the company’s automotive division – Xinzhou Wu – invites CEO Jensen Huang for a test‑pilot, but only if the first fully trusts the system. In their latest trip they used a Mercedes CLA to drive from Woodside to downtown San Francisco. The car was operated by “MB.Drive Assist Pro,” a driver assistance system developed jointly with NVIDIA and similar to Tesla’s Full Self‑Driving. A 22‑minute video shows the vehicle navigating construction sites, streets with dense parking, and narrow sections marked with orange cones. During the test the system never disengaged.

Why NVIDIA is Investing in Autonomous Driving
* Partnership relationships – already working with Mercedes, Jaguar Land Rover, and Lucid.
* At CES 2024 the company unveiled “Alpamayo” – a set of AI models, simulators, and data for creating Level 4 autopilots (the human does not intervene under specified conditions).

NVIDIA uses end‑to‑end models that immediately make decisions based on external signals. At the same time the company maintains traditional verification schemes so that the decision‑making process can be observed and controlled. The result is systems that combine a “human” driving style (with smooth reactions) and verifiable robotics.

> *“End‑to‑end models react better to stationary police cars or lane changes, without creating the feeling of a robot behind the wheel,”* Wu notes.
> *“That’s why the moment for ChatGPT has arrived”* (implying that autopilot could make a real breakthrough).

Technology and Safety
* Sensor diversity – unlike Tesla, NVIDIA does not limit itself to cameras. In its Drive Hyperion system cameras and radars can be used, and in more expensive models (from $40 000 to $50 000) a full set of sensors is added.
* Training on virtual scenes – instead of relying solely on real drives, NVIDIA generates virtual scenes from real recordings. This allows testing the autopilot in extreme situations that rarely occur in reality.
* Partner support – the company receives video footage from radars and cameras from its partners to create more accurate models.

The idea is to build a unified architecture where visual perception, language understanding, and physical actions are combined under one roof. It is compared to teaching a person to drive: first you learn to see the road, then understand the rules, and finally operate the vehicle.

Conclusion

NVIDIA aims to become a key player in autonomous driving by offering flexible solutions that combine advanced AI with reliable verification systems. Their approach allows not only “smooth” reactions to road events but also ensures a high level of safety thanks to a wide range of sensors and virtual testing.