A Chinese robotics startup is racing toward what it calls the "ChatGPT moment" for robots, expecting to achieve a major software breakthrough by mid-2027 that would let machines understand spoken commands and perform complex physical tasks. Spirit AI, valued at $2.9 billion and backed by battery maker CATL and retailer JD.com, is betting that the next frontier in artificial intelligence isn't just language understanding, but embodied AI, where software enables robots to learn from the physical world and act within it.

What Is Embodied AI and Why Does It Matter?

Embodied AI refers to artificial intelligence systems that learn and operate through physical interaction with their environment, rather than purely through text or data. For robotics, this means developing software that allows machines to understand how humans move, manipulate objects, and navigate spaces, then replicate those behaviors. The challenge is significant: while hardware advances have allowed Chinese humanoid robots to sprint, dance, and perform backflips, the software that determines robot intelligence remains the weakest link in the entire robotics stack.

Spirit AI's co-founder and chief scientist Gao Yang explained the company's ambitious timeline during a recent visit to the company's Beijing offices.

"We anticipate reaching the GPT-3.0 milestone by mid-2027. You will be able to speak to a robot in natural language, and it will execute a series of reasonable physical actions to attempt the task," said Gao Yang.

Gao Yang, Co-founder and Chief Scientist at Spirit AI

This milestone would represent a transformative moment for the robotics industry, similar to how OpenAI's ChatGPT launch in 2022 brought generative AI to mainstream audiences and demonstrated its commercial viability.

How Is Spirit AI Training Its Robot Brains?

Spirit AI's approach to training robot intelligence relies heavily on real-world human movement data rather than computer simulations. The company employs approximately 1,000 contractors nationwide who wear data-collection sensors while performing everyday tasks in homes and factories. These workers repeat motions such as opening refrigerators, unlocking safes, and cutting vegetables with knives, generating the diverse movement patterns needed to train robot models.

The company's current robots have achieved a 90% success rate for simple tasks in structured living-room environments, and Spirit AI already has tens of its Moz1 wheeled humanoid robots deployed on production lines at major companies including CATL and JD.com. However, significant challenges remain in perfecting fine-motor actions and handling unexpected scenarios.

• Real-World Data Advantage: Spirit AI prioritizes actual human movement data over simulations because flexible objects like deformable electric cables cannot be reliably trained in virtual environments, requiring physical interaction to learn properly.
• Dirty Data Strategy: Rather than requiring operators to repeat movements 50+ times to capture one "clean" motion, Spirit AI found that using "dirty data" with diverse, less-precise motions enabled models to improve faster and generalize better.
• Current Deployment Focus: The next one to two years will mark the initial window for industrial applications, with robots expected to be deployed in commercial service settings doing simpler tasks before eventually entering homes.

Gao Yang noted the rapid progress the company has made since its founding in 2024.

"Progress is extremely fast. When Spirit AI was founded, a robot could perform only one isolated task well, like pouring water or folding a piece of clothing. Today, robots operate across large spatial areas and execute continuous complex workflows," said Gao.

Gao Yang, Co-founder and Chief Scientist at Spirit AI

The company has raised over $670 million since its founding, making it one of China's most rapidly capitalized embodied intelligence firms.

What Safety Measures Are Being Implemented?

As robots become more capable and begin interacting with humans in unpredictable environments, safety concerns are becoming increasingly important. Spirit AI has implemented baseline safety features including whole-body force control that triggers automatic emergency braking if a robot encounters excessive interaction force with its environment.

However, Gao acknowledged that the risk of an AI system going rogue is less pressing for robots in the physical world at the current stage of development, since the software models are still too immature.

"Once foundation models reach a mature, autonomous 'GPT-4.0' era, researching advanced AI safety and alignment will become much more actionable," said Gao.

Gao Yang, Co-founder and Chief Scientist at Spirit AI

This suggests that as robot intelligence advances toward the GPT-4.0 equivalent milestone, the industry will need to invest significantly more effort in safety research and alignment protocols to ensure robots behave predictably in human-centered environments.

Spirit AI's trajectory reflects a broader shift in the robotics industry, where companies are increasingly recognizing that hardware breakthroughs alone are insufficient. The real competitive advantage lies in developing intelligent software that can learn from diverse real-world data and enable robots to understand natural language commands and execute complex, multi-step tasks. With a mid-2027 target for its GPT-3.0 equivalent milestone, Spirit AI is positioning itself at the forefront of the embodied AI revolution, betting that the next transformative moment in artificial intelligence will happen not on a screen, but in the physical world.