An Initial Analysis of the Shortcomings of Conventional AI and the Benefits of Distributed AI Approaches in Industrial Use Cases

Abstract

The centralised approach of IoT (Internet of Things) applications leveraging cloud infrastructures to address constraints at the level of end and edge nodes is no longer viable, especially for applications with hard real-time requirements and increasing AI (Artificial Intelligence) usage. This paper presents an initial analysis of the shortcomings of such centralised AI approaches applied to the five industrial use cases considered in the ITEA3 MIRAI project and discusses the expected benefits that distributed AI approaches will bring to these use cases, namely, to lift constraints such as computing power, bandwidth, latency, security and privacy.