Key Takeaways

  • General Intuition is developing foundation models for robotics.
  • The startup raised $320 million, valuing it at $2.3 billion.
  • It aims to simplify the robotics development process for others.
  • Current models can operate effectively with minimal real-world data.

Foundation Models in Robotics

Before the rise of models like OpenAI’s GPT-3, companies created specialized natural language processing models from scratch, relying on extensive task-specific data. Today, many organizations utilize general-purpose models and adapt them for their unique requirements. Pim de Witte, CEO of General Intuition, believes that embodied AI will follow a similar trajectory.

Shifting Focus to Quality Data

De Witte argues that instead of gathering vast amounts of real-world data to create specialized robotic models, the industry should prioritize high-quality datasets. This approach could lead to foundation models that effectively transfer knowledge about movement and interaction across various environments. He noted that many companies are currently engaged in specialized work focused on individual robots and environments, which may soon become obsolete.

General Intuition’s Approach

General Intuition has developed its own foundation model, trained on millions of hours of video game data, including detailed action data from controllers. Both de Witte and lead investor Vinod Khosla emphasize that this action data is crucial for cultivating a human-like understanding of spatial-temporal reasoning.

Recent Funding and Model Capabilities

Last month, the startup secured $320 million, bringing its valuation to $2.3 billion. The company has showcased its model’s ability to play video games for extended periods and control a quadrupedal robot, requiring only eight minutes of real-world robotics data for fine-tuning. De Witte expressed surprise at the robot’s performance, which was achieved using just a front camera in a dynamic office environment.

Aiming for a Broader Impact

General Intuition’s goal is not to manufacture robots but to serve as a foundational model for physical AI, enabling other robotics companies to build upon its technology. De Witte stated, “We’re not gonna build a self-driving car company. We’re gonna make it 10 times easier for the next person to build a self-driving car company.” This vision reflects the startup’s ambition to streamline the robotics development landscape.