Edge Computing and Intelligence: AI for Edge and AI on Edge (blog.sparsh.dev)

A recently published paper on Edge Intelligence has categorized this emerging field into two distinct branches: AI for Edge (IEC) and AI on Edge (AIE). Edge Intelligence represents a shift from traditional cloud computing, where data is sent to centralized servers for processing, to a decentralized model that enables devices at the network edge to process data locally. This transformation is critical as it addresses challenges such as network congestion, high latency, and privacy concerns, particularly with the exponential growth of data generated by IoT devices—predicted to reach 90.3 ZB by 2025. The paper outlines key technical implications and potential future research directions, including how AI can optimize resource allocation in edge computing and the importance of running AI models directly on edge devices. Techniques such as Federated Learning and model compression are emphasized as ways to enhance performance while respecting privacy constraints. Overall, this research provides a comprehensive framework for understanding Edge Intelligence's dual nature and sets the stage for further innovation and collaboration in the AI/ML community as it adapits to the rapidly evolving landscape of edge computing technologies.