Geekplus Founder Wang Xingxing: Embodied AI Needs 1-3 Years for Millimeter Alignment Before ChatGPT Moment

At the WIC 2026 forum, Huang Qingqiu of Geekplus stated that large models alone cannot control robots due to latency and closed-loop needs. He emphasized using on-device execution, non-intrusive data collection like Ego and UMI, and real-world validation for embodied AI progress.

The 2026 World Internet Conference Wuzhen Summit hosted a robotics and embodied AI forum on November 4, where companies like Geekplus and JD.com shared data-driven strategies for real-world robot deployment. Experts emphasized that practical application requires coupling models with hardware and high-quality operational data from complex scenarios like hotels before scaling to homes. The event focused on bridging frontier technology with actual use cases through open collaboration and rigorous safety assessments.

At the ninth CIIE forum, Geeks' founder Wang Xingxing stated that current robot challenges stem from input-output misalignment with the physical world rather than planning limitations. He noted that while large-object grasping approaches 100% success, precision assembly remains difficult due to these final-millimeter deviations.

At the Hongqiao International Economic Forum, Geekplus founder Wang Xingxing explained that robot combat entertainment value has decreased because enhanced stability reduces dramatic falls. He identified precise alignment between multimodal models and the physical world as the primary challenge for embodied AI, noting it will likely be resolved within one to two years.

At late October events, Geekplus founder Wang Xingxing integrated GPT-6 Astra into its G1 robot for zero-shot tasks like painting, while Agibot's Zhihuijun deployed GO-3 Preview for planning but retained self-developed force control. This marks a strategic split where one path bets on external general models evolving quickly, and the other prioritizes physical reliability and real-world data collection through proprietary hardware.

At the 2026 World Internet Conference in Wuzhen, experts highlighted emerging liability challenges for embodied AI robots, noting that current laws lack specific risk management frameworks. The discussion emphasized the need to define clear responsibilities among developers, manufacturers, and deployers before physical deployment.

The robotics sector continues to gain momentum, highlighted by Unitree Technology's revenue growth and Xpeng Motors' readiness for robot mass production. At the 2026 conference, Ubtech's Joe Jiaochao emphasized humanoid robots' unique ability to replicate full human labor capabilities. Unitree founder Wang Xingxing predicted a critical industry tipping point next year, focusing on physical AI breakthroughs. A new joint venture between CNPC Kunlun Manufacturing and Unitree targets industrial embodied intelligence solutions.

The 2026 World Internet Conference in Wuzhen highlighted robotics advancements, including Ant Group's pharmacy robots deployed in Shanghai stores and JD.com's data-driven approach to embodied intelligence. Experts noted that while demonstrations are impressive, stable real-world operation and cost reduction remain critical hurdles for widespread adoption.

The 2026 World Internet Conference in Wuzhen featured embodied AI robots performing pharmacy sorting and personal assistants executing multi-step tasks, alongside a 100k-card AI cluster. Experts noted that while demonstrations are promising, stable real-world deployment requires overcoming high costs and adapting to complex environments beyond fixed factory workflows.

At the Wuzhen Summit, twenty organizations jointly released an AI Agent Self-Discipline Pact outlining eight measures to enhance trust, interoperability, safety, and user rights while opposing algorithmic exploitation and illicit activities.

At the 2026 Wuzhen Internet Conference on Nov 4, Geekplus founder Wang Xingxing stated that embodied AI is approaching its ChatGPT moment, with 2027 as a critical industry inflection point. He noted that while top models are close to handling 80% of tasks in unfamiliar scenarios via voice commands, a key barrier remains the model's alignment with physical contact, which he estimates could be resolved within one to three years.

At the 2026 World Internet Conference Wuzhen Forum, ICTO Secretary-General Muhammad bin Ammoun stated that large-scale embodied AI deployment requires overcoming challenges in trust/safety, data infrastructure, and interoperability. He emphasized that future robotics depends on safety and scalability rather than intelligence alone, calling for international cooperation to share knowledge and standards.

At the 2026 World Internet Conference in Wuzhen, Geekplus and other robotics experts discussed embodied AI progress. Secretary-General Mohamed Ben Amour of the Arab ICT Organization highlighted three critical challenges—trust and safety, data infrastructure, and interoperability—that must be solved before large-scale deployment. He emphasized that future success depends on safe, beneficial robots and international cooperation rather than technology alone.

At the 2026 World Internet Conference, Geekplus founder Wang Xingxing stated that 2027 is likely a critical turning point for robotics. He noted that while manufacturing thousands of robots is feasible, efficiently utilizing data for AI self-evolution remains a key challenge, with current models struggling in fine-motor tasks like threading needles.

At the 2026 Wuzhen Internet Conference, Wang Xingxing of Unitree stated that global top models are nearing a "GPT moment" where robots can complete roughly 80% of tasks via voice commands in about 80% of new scenarios, though key technical barriers remain. He noted physical AI breakthroughs may occur within the next one to three years, while industry data shows rapid iteration rates for embodied models from companies like Zhiyuan and Alibaba.

At the 2026 World Internet Conference Wuzhen, industry leaders discussed AI governance for agents and robots, emphasizing that trust, authorized access, and physical safety are prerequisites for real-world deployment. Experts proposed isolating agent actions from user credentials via trusted gateways and shifting security from static compliance to dynamic, real-time risk assessment. The discussion highlighted the need for clear responsibility boundaries among device makers, model providers, and network services to ensure human oversight as AI systems gain autonomous capabilities.

At the 2026 World Internet Conference, Geeks' Robotics founder Wang Xingxing stated that embodied AI is near a "ChatGPT moment" but requires 1–3 years to resolve physical-world alignment issues like fine motor tasks. He noted industrial robots currently succeed in fixed stations with high accuracy but lack generalization across arbitrary scenarios. The company expects significant improvements next year as models approach handling 80% of tasks in unfamiliar settings via voice or text commands.