In addition to launching the Rubin platform, NVIDIA also announced the NVIDIA Alpamayo family of artificial intelligence (AI) models that can be used to develop Level 4 self-driving systems for vehicles. This is to solve the biggest challenge of self-driving systems that need to deal with complex and rare situations that have not been trained in previous self-driving models.
With Alpamayo, self-driving systems can think like humans when faced with unusual situations by using the latest AI models and sensors in the vehicle. Alpamayo’s ability to think step-by-step like humans is like ChatGPT’s introduction to chatbots, according to NVIDIA CEO Jensen Huang.
Alpamayo consists of the open source Alpamayo 1 model, the AlpaSim simulation framework, and the Physical AI Open Datasets. Alpamayo 1 is trained using 10 billion parameters with the ability to solve problems step-by-step based on observed video input.
AlpaSim is an open source simulation framework that supports sensor modeling, dynamic traffic control, and scalable closed-loop testing. Finally, there is the Physical AI Open Datasets, a dataset for training self-driving vehicles with 1,700+ hours of video of vehicles driving in a variety of weather and geographies.
Lucid, Jaguar Land Rover, Uber and Berkeley DeepDrive are among those who have shown interest in using NVIDIA Alpamayo to develop their own Level 4 self-driving systems.
