​Why researching Physical Simulations?

​The reason why LLMs excel as coding agents is that code can be run and tested in a closed feedback loop.

​We are lacking this closed digital feedback loop in the realm of physical science and engineering.

​The training of multimodal-models with a true understanding of physical reality would be enhanced if we could let them trial and error in an accurate physical sandbox.

​We are discussing how current tools like Nvidia Omniverse need to be enhanced so that simulations across all domains of physics (mechanical, electrical, fluidic, optical, biological, quantum, nuclear) can be run.

​This breakthrough would allow for training a general physical engineering intelligence, creating superhuman hardware designs and discover novel science.

About Alexandria