Before OpenAI’s GPT-3 ushered in the era of basic models, companies were building specialized natural language processing models from scratch, training each with large amounts of task-specific data. Today, most organizations start with a general-purpose model like OpenAI’s GPT, Claude, or Llama series and then adjust or customize it to solve their specific needs.

Pim de Witte, CEO of General Intuition, believes embedded AI will follow a similar pattern. Instead of collecting huge real-world data sets to build specialized robot models, he argues that the industry should focus on better quality data sets that can produce basic models capable of transferring intuition about movement and interaction in many environments.

“Right now, a lot of companies are doing a lot of specialized work focused on individual incarnations, individual environments, and individual robots,” de Witte told TechCrunch on a recent episode of Equity.

Much of that work will soon become redundant, he maintains, with the emergence of general models like the one General Intuition has been developing and deploying.

“The generalization of the model itself is the product,” he said. “The fact that you have a basic level of reasoning about space and time will be the reason why people stop collecting hundreds of thousands or millions of hours of real-world data. Because the reality is, it only takes a few minutes.”

General Intuition built its own basic model after training on millions of hours of video game data, including information such as which buttons on a controller a human pressed and when. Both de Witte and General Intuition’s lead investor, Vinod Khosla, argue that action data is the key to developing human intuition for spatio-temporal reasoning.

Last month, the startup raised $320 million at a $2.3 billion valuation thanks to that thesis. The company has shown that its current model is capable of playing a video game for hours and powering a quadruped robot; the latter after fine-tuning it with just eight minutes of real-world robotic data.

“The fact that [the robot] In fact, being able to take zero shots with just the front camera, without other sensors, in the office with dynamic objects coming in and people walking was a big surprise for us,” says de Witte. “I think it’s a sign of things to come.”

General Intuition’s ultimate goal is not to build robots per se, but to become the base model for physical AI, a base model for other robotics companies to build their own machines. Or, as De Witte put it: “We’re not going to create an autonomous vehicle company. We’re going to make it 10 times easier for the next person to create an autonomous vehicle company.”