Can gaming data unlock AGI? Bezos-backed startup raises $320M to find out

A startup backed by Jeff Bezos is betting that the key to artificial general intelligence — the long-sought goal of machines that can think and learn like humans — lies not in more text data, but in the way gamers move through virtual worlds. General Intuition, a New York-based company valued at $2.3 billion, has just closed a $320 million funding round that includes Coatue, former Google CEO Eric Schmidt, and researchers from MIT and Google DeepMind.

The limits of large language models

Large language models like ChatGPT and Claude have demonstrated impressive capabilities in processing and generating text. But according to General Intuition CEO Pim de Witte, these models lack a fundamental understanding of how objects move through space and time. That gap, he argues, is a critical missing piece for achieving AGI. The company’s approach focuses on training what it calls “world models” — AI systems that learn the physics and dynamics of environments by analyzing vast amounts of gameplay data from platforms like Medal TV, from which General Intuition spun out.

From gaming data to physical AI

The company’s thesis is that gaming data offers a rich, structured, and scalable source of information about cause and effect, spatial reasoning, and sequential decision-making. In one demonstration, de Witte noted that just eight minutes of real-world data was enough for a robot trained on gaming-derived world models to navigate an office environment it had never seen before. This suggests that the virtual training generalizes surprisingly well to physical tasks.

Independence and ethical considerations

General Intuition reportedly turned down an acquisition offer from OpenAI to remain independent, with de Witte emphasizing the importance of investors who support a long-term mission. The company is also working to address potential job displacement from AI by creating Nerve, a marketplace that connects gamers with work in data labeling and teleoperation. This move signals an awareness of the societal impact of the technology it is building, even as it pushes toward AGI.

Why this matters

The pursuit of AGI has attracted massive investment, but most approaches rely on scaling up existing architectures. General Intuition’s bet on gaming data represents a distinct alternative path. If successful, it could reshape how the industry thinks about training data and the nature of intelligence itself. The involvement of high-profile investors and researchers lends credibility to the thesis, but the company still faces significant technical and ethical hurdles.

Conclusion

General Intuition’s $320 million raise and $2.3 billion valuation reflect strong investor confidence in an unconventional approach to AGI. By leveraging gaming data to build world models, the company is attempting to solve a fundamental limitation of current AI systems. Whether this bet pays off will be one of the more closely watched stories in the AI industry over the next several years.

FAQs

Q1: What is General Intuition?
A: General Intuition is a New York-based AI startup focused on building world models trained on gaming data, with the goal of achieving artificial general intelligence. It is backed by Jeff Bezos, Eric Schmidt, and others.

Q2: Why use gaming data for AI training?
A: Gaming data provides rich information about spatial reasoning, cause and effect, and physical dynamics that text-based models lack. This could help AI systems understand how objects move and interact in the real world.

Q3: How much funding has General Intuition raised?
A: The company just closed a $320 million funding round, bringing its valuation to $2.3 billion. Investors include Coatue, Eric Schmidt, and researchers from MIT and Google DeepMind.