source: techcrunch ai: this startup thinks robotics is about to have its chatgpt moment
level: business
general intuition, led by ceo pim de witte, argues robotics is moving toward general-purpose foundation models, similar to how gpt-3 changed natural language processing. instead of collecting massive real-world datasets for each robot, the startup trained its model on millions of hours of video game data, including controller inputs. this approach teaches spatial-temporal reasoning that transfers across different robots and environments.
the company recently raised $320 million at a $2.3 billion valuation. in a demo, its model controlled a quadrupedal robot after fine-tuning on just eight minutes of real-world data. the robot navigated an office with people and moving objects using only a front camera, surprising even the team. de witte says this shows that general intuition about movement can come from simulated data, reducing the need for expensive real-world collection.
general intuition does not plan to build robots itself. instead, it wants to be the base model for physical ai, making it easier for others to create robots for tasks like self-driving cars. de witte compares this to how companies now start with large language models and fine-tune them, rather than building from scratch. the goal is to lower the barrier for robotics startups by providing a model that already understands space and time.
why it matters: this could reduce the cost and time to develop robots by replacing task-specific data collection with a reusable foundation model, accelerating deployment in logistics, manufacturing, and autonomous vehicles.
source: techcrunch ai: this startup thinks robotics is about to have its chatgpt moment