The era of universal artificial intelligence models in robotics is beginning
While artificial intelligence has made significant progress in processing text and images, robotics has long remained limited to models trained for specific, narrow tasks. This was reported by Zamin.uz.
However, startup General Intuition has announced that the era of universal foundational models is now entering robotics as well. Instead of training robots for each individual movement, the company aims to give them a general intelligence that understands spatial and temporal logic.
This development has been covered by influential publications. Previously, tech companies had to build new models from scratch for every precise task and train them on millions of hours of real-world data.
According to Pim de Vitte, leader of General Intuition, this approach is no longer viable. Just as modern language models have revolutionized natural language processing, the new generation of technologies can equip robots with the ability and instinct to operate in diverse conditions.
General Intuition leveraged millions of hours of video game data to develop its foundational model. This data includes not only visuals but also when and how human players pressed control buttons.
As a result, the system successfully learned the connection between actions and their outcomes—essentially acquiring a logic akin to human intuition. This enables robots to move through the physical world in a far more natural way.
One of the company’s major achievements is that its model learned to control a quadruped robot for eight minutes of real-world data—without any additional sensors, relying solely on a front-facing camera. The robot successfully navigated various obstacles and moving people in a warehouse environment.
This result came as a surprise and astonishment to experts in the field. General Intuition sees itself not as a robot manufacturer, but as a provider of technological foundations for other companies.
As Pim de Vitte puts it, they don’t build self-driving cars themselves—they deliver the software that makes creating such vehicles ten times easier for the next developer. This approach could fundamentally accelerate the pace of innovation across the entire industry.
This promising idea has also been highly valued by investors. Recently, the startup raised $320 million in funding, bringing its total valuation to $2.3 billion.
The company’s main investors also affirm their belief that action data is the key to enabling artificial intelligence to function perfectly in the physical world. In summary, General Intuition’s approach is expected to completely transform the robotics industry.
Whereas training a robot to walk across a room once required weeks of effort, it can now be accomplished in just a few minutes. This will significantly accelerate the integration of intelligent devices into daily life, manufacturing, and service sectors.
