RGI Labs (YC Summer 2021)

RGI's mission is to build the world's first robust & recursively self-improving general intelligence. We're building the environment for intelligent systems to be able to develop in such a way where they can: - Recursively self-improve across all modifiable planes including architecture, parameters, scaffolding, etc. - Continually learn from online experiences and update its parameters while maintaining robustness against catastrophic failures. - Use its robust general intelligence of the world to acquire new skills in few or zero shots. - Use its world model to predict future states and plan via multi-step rollout. - Operate on a learned latent space that is compressed and feature-rich, including a robust latent space for reasoning.