Humanoid Robots Aren't Quite Ready for Prime Time, Says Unitree CEO
The Reality Behind Those Viral Robot Videos
Those mesmerizing clips of humanoid robots dancing or navigating obstacle courses have flooded social media lately. But according to Wang Xingxing, founder and CEO of robotics leader Unitree Technology, we shouldn't expect these machines to become household helpers quite yet.
Speaking at the 2026 China Online Media Forum, Wang delivered a sobering yet optimistic assessment: The true "GPT moment" for embodied intelligence - when robots become genuinely useful in our daily lives - remains 2-3 years away.
What Would a Robot "GPT Moment" Look Like?
Wang outlined three key benchmarks that would signal robots have reached this transformative stage:
- Environmental adaptability: Walking into any unfamiliar home or office and figuring things out on its own
- Task mastery: Completing 80-90% of complex jobs with just simple voice commands
- Natural interaction: Understanding ambiguous requests and showing physical common sense - more like a helpful friend than a programmed machine
"Right now," Wang explained, "our robots are like talented students who've mastered textbook problems but struggle with real-world surprises."
The Road Ahead: Breakthroughs on the Horizon
The CEO remains bullish about near-term progress. Unitree expects major technological leaps in 2026-2027 that will accelerate mass production. The company aims to ship 10,000-20,000 units this year alone.
Recent achievements hint at this potential. Unitree's robot recently made history by autonomously walking 130,000 steps in -47°C Arctic conditions - proving these machines can handle extreme environments.
Building Robots With "Common Sense"
The holy grail? Developing what Wang calls a "general-purpose humanoid robot foundation model" - essentially giving robots a universal brain that understands the physical world like we do.
"Creating this would be like winning the Nobel Prize in robotics," Wang joked during his talk. While challenging, such breakthroughs could finally bridge the gap between viral sensations and practical assistants.
Key Points:
- True robot intelligence requires understanding environments naturally
- Major advances expected within 2-3 years
- Extreme weather tests prove hardware reliability
- Next goal: developing AI that gives robots "common sense"
