General Intuition is a New York–based startup backed by Jeff Bezos that aims to move toward artificial general intelligence (AGI) by training world models on video game data. The company recently closed a $320 million funding round and is valued at $2.3 billion.
Investors and context
Participants in the $320 million round include Coatue, Eric Schmidt, and researchers from MIT and Google DeepMind. Pim de Witte, General Intuition’s CEO, discussed the company’s approach and plans on the TechCrunch Equity podcast with Rebecca Bellan.
Why game data?
General Intuition argues that large language models — such as ChatGPT and Claude — excel at text but lack the spatial and temporal understanding necessary to reason about how things move and interact in the physical world. The company believes that video game environments provide large, structured synthetic datasets that are well suited for training “world models” capable of capturing motion, dynamics, and cause-and-effect, which are important for more general intelligence.
On the podcast they cited a demonstration in which eight minutes of real-world data was sufficient to enable a robot to navigate an office environment, and said mixing such targeted real-world examples with extensive game-derived data can improve models’ ability to generalize.
Origins and independence
General Intuition was spun out of the gaming platform Medal TV. According to the company, it declined a reported acquisition offer—widely rumored in the press to be from OpenAI—in order to remain independent. Pim de Witte emphasized that retaining independence and having investors aligned with a long-term mission are critical to building a generational company.
Nerve: connecting gamers to work
To address potential job displacement caused by AI, General Intuition is building Nerve, a marketplace that connects gamers with data-labeling and teleoperation tasks. The platform is intended to create earning opportunities for players by having them annotate data and perform remote operations that support model training and system operation.
Ethical boundaries and risks
The company acknowledged ethical concerns on the podcast, noting that models that understand the physical world could be applied in defense and military contexts. General Intuition faces decisions about moral limits, product policies, and acceptable use cases as it develops technology with dual-use potential.
Implications for AGI development
General Intuition’s approach implies that language models alone may not be sufficient for AGI: understanding dynamics, motion, and physical interactions likely requires large, structured datasets from simulated environments supplemented by selective real-world data. The new funding and investor profile should allow the company to further develop its world models, expand its Nerve marketplace, and refine ethical guidelines as it moves to demonstrate practical capabilities while managing social and security risks.
