en.Wedoany.com Reported - On June 2, organizational adjustment information regarding ByteDance's Seed team drew attention. Zhou Chang, ByteDance's head of multimodal AI, has further expanded his management scope. The Seed Robotics team, previously led by Li Hang, has been reporting to Zhou Chang for over a month. Li Hang now serves as an advisor responsible for academic collaboration. ByteDance is also recruiting a technical lead for embodied intelligence to oversee the overall planning of the robotics business, reporting to Zhou Chang.
The signal from this adjustment is quite clear: embodied intelligence is moving up within ByteDance's artificial intelligence system, with the robotics business transitioning from relatively fragmented research directions into a management chain that is more centralized and closer to core model capabilities.
Seed is an important team at ByteDance focused on general intelligence research, covering areas such as large language models, speech, vision, world models, foundational architecture, AI infrastructure, and next-generation AI interaction. Zhou Chang was previously primarily responsible for multimodal interaction and world model directions. Over the past year, visual generation businesses such as the text-to-image model Seedream and the text-to-video model Seedance have also been placed under his purview. With the robotics business now incorporated into Zhou Chang's management scope, a more direct organizational connection has been formed between ByteDance's multimodal AI, world models, and embodied intelligence. For the robotics business, embodied intelligence is not just about connecting large models to robotic terminals; it requires models to understand physical environments, handle spatial relationships, decompose tasks, plan actions, and complete continuous interactions in the real world. Multimodal capabilities address "seeing" and "understanding," world models handle environmental prediction and action reasoning, and the robotics business translates these capabilities into executable physical tasks. By integrating the Seed Robotics reporting line into Zhou Chang's management scope, ByteDance is attempting to bridge the technical boundaries between video generation, spatial understanding, interaction models, and robot control, reducing the previous R&D fragmentation across different teams.
The recruitment of a technical lead for embodied intelligence also indicates that ByteDance aims to strengthen the overall planning capability of the robotics business at the organizational level. This position is graded at L8, comparable to Alibaba's P10 to P11 level, with candidates primarily coming from technical leads at leading embodied intelligence startups.
Embodied intelligence has become a crucial direction for large model companies competing for the next-stage entry point. Pure text, image, and video models primarily remain within digital content production and software workflows, whereas the robotics business targets task execution in the physical world, involving mechanical hardware, perception systems, motion control, embodied large models, data collection, simulation training, and safety constraints. If ByteDance continues to increase its investment in this direction, it may initially focus on R&D organization, technical lead recruitment, and core model capability reuse in the short term, before gradually moving into hardware collaboration, scenario validation, and commercialization route design. Compared to humanoid robot startups, ByteDance's advantage lies not in existing robot hardware manufacturing, but in its accumulated strengths in multimodal models, video understanding, interactive products, data engineering, and recommendation systems. However, for the robotics business to truly enter industrialization, it still needs to fill gaps in real-world physical data, hardware engineering, control algorithms, and scenario delivery capabilities. This adjustment, which incorporates embodied intelligence into Seed's core business scope, reflects more of a strategic resource concentration and technical closed-loop construction, rather than a single product launch.
Subsequent variables are concentrated in three areas: first, whether ByteDance can find a technical lead with experience in robotic systems to truly integrate model R&D with robotics engineering; second, whether Seed Robotics can form a verifiable embodied intelligence technical roadmap centered on multimodal AI and world models; and third, whether the robotics business will choose to develop its own hardware, invest in partnerships, or start by focusing on models, toolchains, and simulation platforms. This round of organizational adjustment at ByteDance indicates that the robotics business has moved from an exploratory direction to a position closer to the core AI business, but its industrial implementation still requires multiple validations across technology, hardware, and scenarios.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
