Wang Xingxing predicts a possible robotics turning point in 2027
At the 2026 Wuzhen Internet Conference, Unitree founder and CEO Wang Xingxing said 2027 was likely to be a turning point for the robotics industry or robot models. He identified aligning model inputs and outputs with real robots as a core challenge for embodied AI. Unitree has begun internal research into self-evolving physical AI models that can write code, train and iterate, while humans evaluate results and set goals, and Wang displayed a “Physical AI Robot Self-Evolution V1.0” research plan. He said trained robot models achieve success rates close to 100% in some single-process or fixed-workstation tasks, while precision tasks such as phone disassembly, assembly and threading remain limited by physical alignment and tactile contact—the “last few millimeters” problem. Unitree plans to focus on robot-brain technology and seek global cooperation.
The remarks identify Unitree’s focus on robot-brain development, model self-evolution and physical-world alignment, while highlighting the precision and tactile limitations that remain in embodied AI.
(本文作者为 产联社CLS,钛媒体经授权发布)
文 | 产联社CLS
“明年很大概率会是整个机器人行业或者机器人模型的一个临界点。”11月4日,宇树科技创始人、董事长兼首席执行官王兴兴在2026年世界互联网大会乌镇峰会“互联网企业家”分论坛上表示,当前全球顶"尖"人工智能模型已具备较强的理解与规划能力,但机器人在精细操作及与物理世界交互方面仍存在技术瓶颈。
他透露,宇树科技已开始探索物理人工智能模型的自主进化,让模型自行编写代码、训练和迭代,人类主要负责评价和设定目标。
展望未来,他认为,除数据和算力外,芯片产能以及背后的能源、矿产资源也将成为全球科技产业发展的关键资源。
2027年或迎产业临界点
王兴兴表示,过去几年,宇树科技持续推动机器人进入工厂等实际作业场景。目前,在部分单工序或固定工位上,经过充分训练的机器人模型作业成功率可以接近100%,但作业效率仍有进一步提升的空间。
#Unitree Robotics#Wang Xingxing#GPT-6#Intel#G1 humanoid robot