Embodied AI shifts focus toward industrial utility and autonomous interaction

The development of embodied AI is shifting from performing human-like movements to achieving practical utility in complex environments. While humanoid robots often demonstrate skills like dancing or folding clothes in controlled settings, the industrial sector faces a gap in automation for high-variability, small-batch manufacturing where traditional robots struggle due to rigid programming.

To address this, EVE Robot and its incubated entity, Qizhi (Wuhu) Intelligent Robot Co., Ltd., are pursuing a dual-track strategy of autonomy and intelligence. They focus on self-developed hardware and a technical foundation—including the Openmind OS and HumanGPT world model—designed to act as a universal base for various robot configurations. A key component of their approach is the use of real-world, multi-modal data collection via the HALO skill suit, which captures human expertise in tasks like welding and spraying to train robotic models.

Beyond physical tasks, the industry is facing a new challenge regarding autonomous interaction. As robots move into homes and care facilities, the focus is shifting toward defining levels of interaction autonomy. This involves moving from passive response (L1) to context-aware understanding (L2), where robots must learn to navigate social boundaries—knowing when to intervene, when to offer assistance, and when to remain silent to avoid overstepping human emotional and decision-making boundaries.