General Intuition raises $134m to train AI agents using gameplay videos

• The AI startup spun out of the gaming clip platform Medal.
• Medal’s 2 billion annual videos fuel AI spatial reasoning research.
• Khosla Ventures and General Catalyst led the $134m seed round.
• The company’s early focus includes gaming and search-and-rescue drones.

AI platform General Intuition has raised $134 million in seed funding to train artificial intelligence agents in spatial reasoning using video game clips.

As reported by TechCrunch, the company will use the funds to expand its team and develop an AI agent for real-world use, starting with games and rescue drones.

General Intuition aims to build virtual worlds to train AI and help them navigate real environments on their own - a plan the startup said sets it apart from other AI companies. 

The AI research startup spun out of video-sharing platform Medal, leveraging its massive library to train foundation models and agents. General Intuition was built on Medal’s dataset of over 2 billion videos annually from 10m active users. 

Khosla Ventures and General Catalyst led the seed round with participation from Raine.

Learning from players 

The startup aims to develop smarter game bots and NPCs that outperform traditional preprogrammed ones with more dynamic, adaptive behaviour.

“When you play video games, you essentially transfer your perception, usually through a first-person view of the camera, to different environments,” said Medal and General Intuition CEO Pim de Witte.  

He explained that gamers usually upload highly positive or negative moments, creating ideal edge-case data for AI training.

Furthermore, De Witte said General Intuition’s model can interpret new environments and predict actions using only visual input, mimicking human gameplay. This method, he added, could easily extend to real-world systems like robots, drones, and autonomous vehicles.

“Our goal is not to produce models that compete with game developers,” said de Witte.