"My Agent in Your Hand"... Smartphones Join PCs as AI Control Centers

Greater AI Convenience on Personal Devices
For Full Performance in Disconnected Environments
Advancements Needed in AI Semiconductor Performance and Miniaturization

Features that allow the use of artificial intelligence (AI) directly on personal devices without connecting to external clouds are expanding. While heavy computations are handled by the cloud or a PC, simple and personalized tasks can be performed on mobile devices, gradually increasing the convenience of AI.

On June 30, OpenKlo released its official iOS and Android apps, enabling users to control AI agents from their smartphones. Users can converse with the AI agent on their smartphones, check ongoing tasks, and issue instructions. The app also provides features that link smartphone functions such as the camera, location information, and contacts with the AI agent.

Last month, OpenAI added the "Codex" coding agent feature to the ChatGPT mobile app. This allows users to check ongoing development work handled by Codex or request new tasks directly from their mobile devices. The AI coding platform Cursor has also released features for managing agents in mobile environments, further enhancing AI accessibility.

Running AI at the personal device level offers the advantage of processing sensitive data internally, without sending it to external clouds. Another benefit is faster response times, as there is no need to go through the cloud. By handling repetitive AI tasks on the device, cloud server usage is reduced, which helps companies cut operational costs. For individual users, real-time AI features such as document summarization, translation, and voice recognition become more accessible, improving overall AI accessibility.

Within the industry, a "hybrid structure" is being adopted, where the cloud is responsible for large-scale training and complex inference, while on-device AI handles personalized AI and real-time responses. However, some experts say on-device AI is still in a transitional stage. They point out that there are technological limitations to fully implementing high-performance AI in environments with disconnected networks or in internal closed networks.

Jinho Park, Professor of Computer AI at Dongguk University, said, "The convenience has increased to the point where AI can be used at the personal device level, such as on smartphones, but it is not yet at a level where AI can be fully utilized in closed networks or completely disconnected environments." Professor Park added, "If, in the future, advances in AI semiconductors such as graphics processing units (GPUs) or neural processing units (NPUs) lead to miniaturization and performance improvements, AI will be able to learn independently within devices and share content with each other, greatly accelerating learning speed and usability."

This content was produced with the assistance of AI translation services.