This week at CES, Jensen Huang (CEO Nvidia) introduced the idea of a world foundation model and Nvidia’s new Cosmos World Foundation Model development platform.

This debut is timely given one of the main focuses of my project is to link world models with LLMs in order to increase LLM intuition and knowledge based on social models. Given this, and the many examples of neural network connections to physics based models, I feel like my project needs a bit of an adjustment, or at least re-emphasis.

Rather than making a case for linking world models with LLMs, I will instead demonstrate that a single LLM can generate code for linking to a world model and then by interacting with this world model, the LLM can increase its understanding and intuition of the domain.