leaderobot.com reports on Guangxiang Technology CEO Zhang Tao’s remarks on embodied intelligence

On October 4, 2026, leaderobot.com reported on remarks about embodied intelligence by Guangxiang Technology CEO Zhang Tao. The article addresses views on the relationship between autonomous driving and embodied intelligence.

The article reporting Zhang Tao’s remarks suggests that the development process for embodied intelligence may not be directly modeled on autonomous driving. If autonomous driving-derived data pipelines and virtual validation can be used while contact-based manipulation evaluation must be handled separately, the allocation of development investment may change.

This appears to present the question of where embodied intelligence can draw on autonomous driving and where it differs, challenging the view that it is simply an extension of autonomous driving. The article addresses not just technical issues but also how processes such as data collection, cleaning, end-to-end learning, virtual validation and real-world testing should be reorganized for robots. In particular, the distinction that “methodologies” can be transferred but manipulation involving hand contact and force control does not fit the timeline of autonomous driving could prompt changes to development plans. Previous reports, such as [Tesla continues review of humanoid robot supply chain in the Yangtze River Delta, focusing on existing automotive parts companies](/news/tesura-chouzou-kaku-de-hitogata-robotto-kyoukyuu-ami-no-shin-c7ef86) and Tesla begins sourcing Chinese components for U.S.-made humanoid robots, show that automotive supply chains and parts procurement are beginning to connect with embodied intelligence. The issue raised here goes further, to how training data and manipulation data should be built up separately. The focus therefore shifts to how far sensors, virtual environments and evaluation metrics optimized for vehicles can be repurposed for robots. In particular, because grasping and assembly involving contact cannot be measured by driving distance or road scenarios, the next question is how much real-world test design is needed. Consult the original source for details.