NVIDIA's Artificial Intelligence moves.

Moves to the physical world

Nvidia announced a massive expansion of its ecosystem now with a clear objective, to remove artificial intelligence from the screens and make it perceive, reason and act in physical environments and that is a big step, because now we are not talking about an AI that only responds, but rather about systems that need to understand the environment, make decisions and execute actions in real time.

For autonomous vehicles, the company presented Alpamayo 1, a model capable of interpreting what is around it and explaining its own decisions, something essential for systems that must operate safely in the real world. Along with this came Elpasim, a simulation platform that allows these systems to be trained in extreme scenarios and situations that would hardly occur in traditional tests.

The focus here is on the data.

NVidia unleashed massive sets, including more than 1,700 hours of real-world driving and tens of terabytes of sensor data, creating a foundation that enables technology to learn with diversity and complexity. This is part of a larger movement, with models like Cosmos, Isaac Groot and Nemotron, the company is building a unique base of intelligence that can be applied in cars, theft and physical systems in general. In practice, this represents the advancement of so-called physics, systems that not only process information, but must deal with the real world, with its unpredictability and continuous decisions.

But isolated intelligence still has a limit, it needs scope, it needs to be present, connected, distributed and that is exactly where a new layer begins to emerge.