General Intuition, a New York-based startup backed by Jeff Bezos, has raised $320 million in a funding round led by Coatue, Eric Schmidt, and researchers from MIT and Google DeepMind. The company, valued at $2.3 billion, is focused on training world models using video game data to advance physical AI. According to CEO Pim de Witte, large language models like ChatGPT and Claude are limited in their ability to understand spatial and temporal dynamics, which are crucial for achieving artificial general intelligence (AGI). This gap, de Witte argues, could be addressed by leveraging the structured and dynamic environments found in video games. The startup, which spun out of gaming platform Medal TV, aims to create models that can better simulate and predict real-world physical interactions. The company’s approach has sparked discussions about the potential of gaming data as a more effective training source than traditional internet data for AI development. The ethical implications of using such models for defense applications are also under consideration.
The funding round highlights the growing interest in physical AI and the role of gaming data in advancing this field. General Intuition’s CEO emphasized that video games provide a controlled environment where agents can learn to navigate and interact with objects in ways that are not easily replicated by the internet’s unstructured data. This approach could lead to more accurate and versatile AI systems capable of handling complex physical tasks. The company’s focus on world models suggests a broader effort to bridge the gap between abstract language understanding and concrete physical reasoning. By using gaming data, General Intuition aims to create AI systems that can generalize across different environments and scenarios, which is a key challenge in achieving AGI. The startup’s strategy reflects a shift in the AI industry toward more specialized and context-aware models that can handle real-world complexities.
General Intuition spun out of Medal TV, a gaming platform, and has since evolved to focus on physical AI through world models. The company’s CEO, Pim de Witte, discussed the potential of gaming data to provide a more structured and dynamic training environment for AI. The ethical considerations of using such models for defense applications are also part of the company’s ongoing discussions. The startup’s approach is part of a broader trend in AI research that seeks to move beyond text-based models and explore more realistic and interactive training environments.
Source: techcrunch
