Honor Previews AgenticOS AI Architecture Concept Focusing on Self Governing Agents and Future Integration with MagicOS 11
At the MWC Shanghai exhibition in China, Honor previewed its upcoming new OS architecture concept called AgenticOS, an evolution of the mobile operating system where the focus shifts from hardware oriented interaction system to AI oriented interaction system. During a key note speech by Honor Product Line President Fang Fei, he said
"In the next 10 years, mobile hardware will evolve from being vessels for applications, to being the dedicated operating environments for self governing AI agents"
As indicated in the executive presentation, the terminal has a strategic position among the technology ecosystem. On one hand, the terminal connects with the user on the other hand, it links with cloud model and computing power. AgenticOS is intended to serve as the critical transition, transpose these functionalities to users by stitching up local sensors, computing, and action features together.
The design foundation of AgenticOS is formed on 4 core traits that are new to conventional mobile software design. 1st pillar being intent driven, implying the complete system self architectures to the real requirements instead of making the user go through different packages. 2 nd pillar being multi dimensional natural interaction where the system can take commands through speech, gestures, eye movements, human movements and reduce the dependency on touch screens.
The 3 pillar is active intelligence provided by a core agent, which gives the system the native capacity of intermediate planning, assisting, and program execution against complex tasks with no need for step by step user commands, the 4 pillar is cross device integration, which uses a centralized unified intelligence engine to seamlessly cooperatebetween thousands of widely dispersed terminals with individual AI agents.
Honor is building this human centric OS on developing 3 basic capacities, namely perception, planning and action. In terms of perception, the company is improving its YOYO virtual assistant by enabling it to meet the needs of contextual understanding and user habits thereby making it transition from reactive software to pro active service provider. The processing task is offloaded from the phone to neighboring wearable devices such as smart watches and wireless headsets to scale.
In the planning phase, Honor uses an LMI of its own on device and cloud models, with the YOYO Claw framework orchestrating tasks over a range of different devices such as tabletop PCs, tablets and smart home appliances. In the action phase, the agents use the Magic GUI and Magic Agent model, which use the system sensors of a full stack to effect the information context in the virtual agent for carrying out multi step automized functions.
These technologies were previously evidenced by the company even at an early stage. The self developed Honor Robot Phone prototype used motion ability through embodied AI, and the company built a humanoid biped robot that competed in athletic contests within a month of development. Early features of its AgenticOS framework will be embedded into the yet upcoming MagicOS 11 which will also debut the first liquid glass visual design for Android ecosystem to boost UI fluidity.
Along with the software unveiling Honor partnered with the GSMA at the event to release an in depth research paper on the future 6G terminals. In the report they explore the use of AI in the new generation cellular network with detailing 4 key capabilities, 7 new categories of hardware and 4 prospective economic growth areas.
