General Intuition Raises $320 Million for Embodied AI Model at $2.3 Billion Valuation
Updated
Updated · TechCrunch · Jul 8
General Intuition Raises $320 Million for Embodied AI Model at $2.3 Billion Valuation
2 articles · Updated · TechCrunch · Jul 8
Summary
$320 million in fresh funding values General Intuition at $2.3 billion as it pushes a foundation model for embodied AI rather than building robots itself.
Millions of hours of video-game data trained the model, with CEO Pim de Witte arguing action-rich datasets can replace the vast task-specific real-world data robotics teams now collect.
Eight minutes of robot data were enough to fine-tune the system for a quadruped, which the company said then operated with a front camera alone in an office with moving people and objects.
That pitch mirrors how general-purpose language models displaced bespoke NLP systems: General Intuition wants its base model to become infrastructure other robotics and physical-AI companies build on.
In the race for physical AGI, can a startup's gaming data truly outmaneuver tech giants' vast resources?
Is paying gamers to train AI a viable new career path or the blueprint for our own obsolescence?
Can AI trained on video games truly master our physical world, or will reality prove too complex?
General Intuition Raises $320M Series A to Build AI Agents from Billions of Gameplay Videos
Overview
General Intuition has raised $320 million in Series A funding at a $2.3 billion valuation to advance its unique approach to AI training. The company stands out by using billions of action-labeled gameplay video clips as a proprietary data asset, enabling the development of AI models with advanced spatial-temporal reasoning. This strategy, seen as a 'generational bet' by investor Vinod Khosla, positions General Intuition as a potential backbone for future generalized agents and world models. The new funding will help expand compute capacity and further strengthen their innovative use of video game data for robotics and AI applications.