Accelerated Understanding, a company founded by former NVIDIA scientists Anima Anandkumar and Benedikt Jenik, recently released an AI model. This model focuses on predicting physical phenomena rather than traditional language processing. The company stated that the model processed 5 trillion data points in a single prompt, far exceeding the contextual processing capabilities of typical AI. The technology utilizes a neural operator rather than a Transformer architecture, and is primarily aimed at enterprise applications such as chip design, robotics, extreme weather, and energy.