Spirit AI sees a ‘robot brain’ breakthrough arriving as soon as 2027
The Chinese startup is betting that general-purpose control models can do for physical machines what large language models did for software interfaces.
The story
The founder of Chinese startup Spirit AI expects control systems for robots to approach a ChatGPT-style breakthrough as soon as next year, according to Reuters.
The ambition is to replace narrow, task-by-task programming with models that can interpret environments and transfer learning across machines. That would make robots easier to deploy in settings that are too variable for traditional automation.
The challenge remains physical reliability. A compelling demonstration must become safe, repeatable performance across different hardware, workplaces and edge cases before general-purpose robotics can scale.
INNOVOX analysis
Robotics does not receive the forgiving feedback available to a chatbot. A control mistake can damage equipment or harm people, and performance can change with lighting, surfaces and hardware. General models may reduce the cost of programming every task separately, but commercial value will depend on consistent behaviour in ordinary, messy environments.
What to watch
Pay attention to demonstrations outside controlled settings, the amount of human teleoperation required and performance over long periods. Safety certification, hardware compatibility and the cost of collecting physical-world data will shape the pace of adoption.