General Intuition says video games can train AI for the real world

In short: A startup called General Intuition says video games can provide better training data for AI systems that need to understand movement and cause and effect.

What's going on

General Intuition is betting that video games can help train AI in ways that regular internet text cannot. The company’s CEO, Pim de Witte, made the case on an episode of TechCrunch’s Equity podcast.

The idea is that today’s popular AI chatbots are strong at working with words, but they often struggle with how things behave in space and time. For example, text can describe a ball rolling down a hill, but a game engine can simulate it step by step. Think of it like the difference between reading about driving and practicing in a driving simulator (a safe, controlled practice world).

General Intuition focuses on “world models,” which are AI systems that try to predict what will happen next in an environment. In simple terms, they are like an internal map and a set of rules that help an AI plan actions, not just write answers.

TechCrunch also notes that General Intuition is a New York based startup backed by Jeff Bezos. It was recently valued at $2.3 billion and it closed a $320 million funding round, with investors including Coatue and Eric Schmidt, plus researchers connected to MIT and Google DeepMind.

What to watch

De Witte also discussed ethical limits, especially if this kind of AI could be used for defense. A key question will be how companies share, restrict, or monitor these systems when they move from game like simulations into real world tools.

Source: TechCrunch AI