This startup believes that bots are about to have their ChatGPT moment

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

Pim de Wit, CEO General intuitionbelieves that embodied AI will follow a similar pattern. Instead of collecting huge real-world datasets to build specialized robotic models, he argues that the industry should focus on better quality datasets that can produce basic models capable of conveying intuition about movement and interaction across many environments.

“A lot of companies are now doing a lot of specialized work that focuses on individual avatars, individual environments, individual robots,” De Wit told TechCrunch. Last episode of stock.

He believes that much of this work will soon become redundant, with the emergence of generic models such as the one General Intuition is developing and deploying.

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

General Intuition built its own base model after training on millions of hours of video game data, including information like which buttons on a controller a human pushed and when. Both De Wit and General Intuition’s lead investor, Vinod Khosla, see action data as key to developing human-like intuition for spatial and temporal reasoning.

Starting last month It raised $320 million at a valuation of $2.3 billion Evaluation against the background of that thesis. The company has demonstrated that its current model is capable of playing a video game for hours and operating a four-legged robot – the latter after being tuned to just eight minutes of real-world robotics data.

“The fact that (the robot) was actually able to shoot only the front camera, with no other sensors, in the office with the input of dynamic objects and people walking by, was a very big surprise for us,” says de Wit. “I think it’s a sign of what’s to come.”

General Intuition’s ultimate goal is not to build the robots themselves, but rather for them to become the primary model for physical AI, and a basic model for other robotics companies to build upon in their own devices. Or as De Wit put it: “We’re not going to build a self-driving car company. We’re going to make it 10 times easier for the next person to build a self-driving car company.”