Spirit AI, a Chinese startup, is working on giving robots an industrial "brain" that could reach a major milestone by mid-2027. The company aims to make robots more reliable and economically useful by accumulating sufficient real-world data. According to co-founder and chief scientist Gao Yang, the goal is to achieve a milestone comparable to GPT-3 in robotics. This would enable users to give robots instructions in natural language and have them perform a series of physical actions.
Currently, Spirit AI's robots can complete simple tasks in structured environments with around 90% success rate. However, the company acknowledges that it still faces challenges with fine motor skills and unfamiliar situations. To overcome these limitations, Spirit AI is collecting high-quality data through around 1,000 operators equipped with data collection devices. These operators record how humans interact with their environment at home or in factories.
Spirit AI's approach prioritizes real-world interactions over simulation-based approaches. The company argues that certain objects, such as deformable cables, are difficult to model accurately in simulations. By accumulating real-world data, Spirit AI aims to improve the generalization and reliability of its models. This approach is crucial for robots to learn how to manipulate objects and react to complex physical situations.
The company has already begun deploying its robots in industrial settings. Dozens of its Moz1 robots on wheels have been deployed on production lines at CATL and JD.com. This strategy requires significant investment, with Spirit AI having raised over $670 million and being valued at around $2.9 billion since its founding in 2024.
Spirit AI's next technological leap is expected to occur in mid-2027, which will be a test of the robots' generalization and reliability. The company's strategy involves collecting data, training models, deploying robots, and learning from new situations. This iterative process is crucial for the development of more advanced robots.
The use of robots in industrial settings is becoming increasingly important. By deploying robots on production lines, Spirit AI can collect more data and improve its models. This approach enables the company to refine its robots' performance and increase their reliability.
Spirit AI's goal of giving robots an industrial "brain" has significant implications for various industries. With more advanced robots, companies can automate complex tasks, improve efficiency, and reduce costs. As Spirit AI continues to develop its technology, it is likely to have a major impact on the robotics industry.
Key points
- Spirit AI aims to make robots more reliable and economically useful by accumulating real-world data.
- The company's robots have achieved around 90% success rate in completing simple tasks in structured environments.
- Spirit AI has raised over $670 million and is valued at around $2.9 billion since its founding in 2024.
