Why this CEO thinks video games make better training data than the internet

2026-07-08 · Source: AI News & Artificial Intelligence | TechCrunch · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Gaming & Interactive Media · Depth: Novice, quick

Summary

General Intuition, a New York-based startup backed by Jeff Bezos and recently valued at \$2.3 billion after closing a \$320 million funding round with investors including Coatue, Eric Schmidt, MIT, and Google DeepMind, posits that video game data is crucial for achieving Artificial General Intelligence (AGI). CEO Pim de Witte, on TechCrunch's Equity podcast, explained that while large language models excel at text, they lack the spatial and temporal understanding necessary for generalized intelligence. The company's bet is that world models trained on gaming data can bridge this gap, leading to advancements in physical AI and real-world AI agents, despite potential ethical considerations regarding defense applications.

Key takeaway

For AI/ML Directors evaluating future model development strategies, consider the potential of gaming data to overcome current large language model limitations in spatial and temporal understanding. This approach, championed by General Intuition, could be pivotal for advancing physical AI and creating more capable real-world agents. You should explore startups leveraging this novel data source, while also carefully assessing the ethical implications, particularly for sensitive applications like defense.

Key insights

Gaming data offers a superior training ground for AI to grasp physical world dynamics and achieve AGI.

Principles

• Large language models lack spatial and temporal understanding for AGI.
• Gaming data can enable physical AI and real-world agents.

Method

Training "world models" on gaming data is proposed as the next significant step for developing physical AI.

In practice

• Develop AI agents for real-world applications.
• Consider AI models for defense applications.

Topics

• Artificial General Intelligence
• Gaming Data
• World Models
• Physical AI
• General Intuition
• AI Ethics

Best for: Investor, Entrepreneur, Director of AI/ML

Related on AIssential

• BREAKING: Expensive new evidence that scaling is not all you need — Pure scaling of compute and data is proving insufficient for achieving AGI, as evidenced by recent high-profile setbacks.
• General Intuition’s $2.3B bet that video games can train AI agents for the real world — Video game action data provides a scalable shortcut for training generalizable AI agents in spatial-temporal reasoning.
• Yann LeCun Pushes Back on AGI Hype at India AI Summit: "Superintelligence Isn't Close" — True AI requires a "world model" for physical interaction, a capability current LLMs and systems lack.
• AGI: Francois Chollet + Sam Altman — AI's rapid advancement necessitates evolving social contracts and re-evaluating human agency and intelligence.
• Machine Learning Is Changing iGaming Software Development — Machine learning and AI are transforming iGaming development, enhancing security, personalization, and operational efficiency.
• General Intuition in talks to raise $300M at around $2B valuation — General Intuition leverages unique interactive video game data to train embodied AI agents for spatial-temporal reasoning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI News & Artificial Intelligence | TechCrunch.