While Large Language Models (LLMs) like ChatGPT and Claude have demonstrated remarkable proficiency in processing text, some experts believe they lack a fundamental component of true intelligence: a physical understanding of the world. General Intuition, a new AI startup, is betting that the path to Artificial General Intelligence (AGI) lies not in the open internet, but within the structured environments of video games.
The Spatial Reasoning Gap
Current AI models primarily learn from static data and text, leaving them with a significant deficit in understanding how objects move through space and time. This limitation prevents models from developing the 'general' intelligence required to interact effectively with the physical world. According to General Intuition's leadership, gaming data provides a high-fidelity simulation of physics and cause-and-effect that the internet's text-heavy archives simply cannot match.
A Synthetic Sandbox for Intelligence
By training on video game data, AI can observe complex interactions, spatial navigation, and temporal sequences. This approach allows models to learn the underlying rules of reality in a controlled, data-rich environment. Proponents of this method argue that this 'embodied' data is the missing link needed to transition AI from a sophisticated chatbot into a system capable of generalized reasoning and physical interaction.
