Towards Named AI Networking: Unveiling the Potential of NDN for Edge AI

Thanks to recent advancements in edge computing, the traditional centralized cloud-based approach to deploy Artificial Intelligence (AI) techniques will be soon replaced or complemented by the so-called edge AI approach. By pushing AI at the network edge, close to the large amount of raw input data, the traffic traversing the core network as well as the inference latency can be reduced. Despite such neat benefits, the actual deployment of edge AI across distributed nodes raises novel challenges to be addressed, such as the need to enforce proper addressing and discovery procedures, to identify AI components, and to chain them in an interoperable manner. Named Data Networking (NDN) has been recently argued as one of the main enablers of network and computing convergence, which edge AI should build upon. However, the peculiarities of such a new paradigm entails to go a step further. In this paper we disclose the potential of NDN to support the orchestration of edge AI. Several motivations are discussed, as well as the challenges which serve as guidelines for progress beyond the state of the art in this topic.