Edge AI and its Impact on Resilient AI Fabric Design: Distributed Intelligence and Data Locality

Abstract

As Artificial Intelligence (AI) transforms various sectors of industry and all spheres of everyday life, there is more necessity to make intelligence relocate towards the place where data is being generated at the edge. This shift is termed Edge AI, and it allows gadgets, including smartphone, sensors, and cost-automated machines, to determine decisions on the ground, devoid of the continuously reliant connection to geographically distant distributed cloud servers. It is expected that it would augment response rates, data security as well as autonomy, but it also raises profound issues on how we would be able to construct resilient systems that are made of distributed agents with AI capabilities that could work effectively under real world conditions.
In this paper, I will explore how deployment of AI at the edge may impact the topology, the resilience of AI fabrics, the complex interdependence of compute, data and learning systems that enable intelligent decisions. We revisit the rationale of Edge AI and how latency minimization and control of data sovereignty have led to it, and the issues of resource bottlenecks and latencies, synchronization and security in a distributed setting. We also make comparisons between the current solutions of data locality, collaborative intelligence, and federated learning and we also endeavor to find answers as to how an AI infrastructure can become flexible and dynamic in the processes of a decentralized world. We would like to establish a reputation of the principles of being smart, resilient, secure, and human aligned AI through this work.