General Intuition targets physical AI as investors value the startup at $6 billion

Navigating through space is an order of magnitude more complex for artificial intelligence than normal conversational tasks. General Intuition, a startup building foundation models that teach agents to operate in space and time, is negotiating a new investment. The expected valuation hits $6 billion, and according to TechCrunch, the round is set to include prominent investment funds.

It is a significant leap. General Intuition recently raised $320 million at a $2.3 billion valuation. The market is clearly signaling that physical AI and the ability to transfer model reasoning into robotic embodiments is currently a highly valued bet in Silicon Valley.

The data foundation combines millions of hours of gaming records

General Intuition's approach is based on an unusual data foundation. Founder Pim de Witte utilized data from his own gaming clip-sharing platform Medal. Hundreds of millions of hours of gameplay paired with precise button presses form a unique layer of action labels. The model learns how specific interventions alter the environment.

This approach is intended to provide the model with the ability to generalize behavior to new, unfamiliar tasks without having to explicitly program them beforehand. Instead of synthetic datasets, it builds on a massive record of real human reactions in a simulated space.

Simulations must cope with hardware imperfections

The current investment round is reportedly oversubscribed, and the money is critical for the firm. General Intuition needs to pay for massive computing infrastructure with its partner cloud provider CoreWeave and hire hardware integration specialists. The physical world, however, does not forgive mistakes.

The race for the first truly universal model for physical deployment is entering a phase where demonstrating capabilities in games is no longer enough. Real robot sensors suffer from noise, motors have latency, and materials behave unpredictably.

Portability to the physical world determines success

The decisive indicator will be how easily the capabilities trained on video games can be transferred to real machines. A successful, smooth transfer without extensive tuning for a specific chassis will confirm that gaming data is enough to understand physics. If General Intuition's model masters this transition, the billion-dollar valuation will prove justified in hindsight.

Dominance in simulation does not guarantee field resilience

The startup promises a universal agent, but practice shows that models trained without real feedback often fail with minor deviations in lighting or material friction. We will see if the massive quantity of gaming data can compensate for the lack of real physical experience with the imperfections of the ordinary world.

Lilith's verdict

Whoever controls continuous spatial movement data will govern tomorrow's logistics. The half-billion-dollar bet that physics can be extracted from gaming streams reveals where the demand for quality data is heading next.