All the world’s a robot-staging ground for tech entrepreneurs building ‘physical AI’
All the world’s a robot-staging ground for tech entrepreneurs building ‘physical AI’
Publish Date: 2026-06-24 00:07:00
Source Domain: halifax.citynews.ca
Here is a summary of the article with between 4 and 8 key points:
-
Shift from Language to World Models: Computer scientist Louis Castricato left traditional large language model (LLM) research and started a company, Overworld, which aims to develop AI that understands and navigates the physical world, moving beyond language.
-
World Models: Next Frontier: A growing number of AI entrepreneurs, including prominent scientists, are focusing on “world models” that teach AI systems how to react in a physical environment, seen as the next frontier of AI.
-
Beyond Text: World models aim to make AI truly intelligent by not just processing text but understanding spatial and temporal relationships, physical laws, and dynamics of environments.
-
Applications in Robotics and Games: Proponents argue that world models can enable smarter robots and more interactive video game worlds where virtual characters interact with detailed environments.
-
Types of World Models: World model researcher Fei-Fei Li categorizes world models into “renderers,” “simulators,” and “planners,” each differing in visual fidelity and physical simulation capability.
-
Investment Interest: Venture capitalists are increasingly showing interest in world model research, viewing it as a lucrative potential investment area despite near-term applications not being fully evident.
-
Physical and Embodied AI: Researchers like Carnegie Mellon’s Martin Hebert see world models as a path to developing physically aware AI, akin to the brain’s role in enabling human adaptability and movement.
-
Future Diversity: The future is expected to consist of various types of models with different philosophies, architectures, according to Steve Jang from Kindred Ventures, moving away from the idea of a single, universal model.
