在上海舉行的世界人工智慧大會上，數百個仿生機器人成為全場焦點。儘管外觀極度逼真，但大多數機器人在精心編排的展示之外仍然行動遲緩且僵硬，這凸顯出雖然機器人的硬體軀殼建構已取得重大進展，但要開發出能自主思考的大腦仍然是一項巨大的挑戰。

為了突破這項瓶頸，中國科技巨頭如騰訊、阿里巴巴以及字節跳動等，正將發展重心轉向「世界模型」。這些模型旨在超越現有的大型語言模型，試圖賦予機器人理解物理世界並具備常識的能力，讓它們能像人類嬰兒一樣，透過觀察、行動與回饋來適應充滿未知與不確定的環境。

中國在具身智能領域擁有獨特的競爭優勢，不僅受惠於其深厚的製造業供應鏈基礎，該領域更被列為國家政策的重點發展方向。儘管專家評估世界模型要取得重大突破可能還需要數年甚至更長的時間，但中國正積極利用其豐富的硬體資源與開源策略，為後 ChatGPT 時代的人工智慧新競賽做足準備。

At the World AI Conference in Shanghai, hundreds of humanoid robots became the center of attention. Despite their highly lifelike appearances, most robots remained slow and jerky outside of carefully choreographed demonstrations, highlighting that while significant progress has been made in building robotic hardware bodies, developing a brain capable of autonomous thought remains a massive challenge.

To overcome this bottleneck, Chinese tech giants such as Tencent, Alibaba, and ByteDance are shifting their focus toward developing "world models." Designed to go beyond existing large language models, these models aim to equip robots with a common-sense understanding of the physical realm, enabling them to adapt to unknown and uncertain environments through observation, action, and feedback, much like a human baby.

China holds unique competitive advantages in the field of embodied intelligence, benefiting not only from its deep manufacturing supply chain foundation but also from the sector's prioritization in national policy. Although experts estimate that major breakthroughs in world models may still take years or even longer, China is actively leveraging its abundant hardware resources and open-source strategies to fully prepare for the new AI race in the post-ChatGPT era.