Keynote: Edge Intelligence - Emerging Solutions and Open Challenges

Abstract

Recent advances in AI have led to a number of intelligent applications, such as intelligent personal assistants, context-related and personalized buying advice, automated video surveillance and smart appliances. Such applications quickly rose to prominence and gained tremendous popularity in a short time as they have had a positive impact on people's lives by improving the productivity and generally efficiency of human-based processes. With the proliferation of mobile, pervasive computing and the IoT, the trend that had seen the cloud as the largest data concentrator is reversing. In fact, bringing the enormous volumes of data to a completely centralized AI is unthinkable given the size of the phenomena taking place. The edge ecosystem plays a fundamental role in enabling the creation of intelligent applications and systems that start from the periphery and integrate with centralized AI systems. It is therefore necessary to push the boundary of AI towards the edge ecosystem that basically lives in the last mile of the Internet. This, however, is an extremely difficult task due to the non-trivial issues of performance, cost, privacy and integration. Transferring huge amounts of data across Wide Area Networks raises costs, slows down operations, and makes ensuring privacy extremely complex. Energy efficiency also plays a key role in this context. In fact, AI services require computing power which must therefore be shared in the periphery, appropriately guaranteeing all systems, even the limited ones, to be able to use them. It is therefore necessary to exploit the physical proximity between the sources of processing and generation of information through decentralized architectures capable of cooperating: this allows to obtain numerous advantages compared to the traditional processing paradigm which is entirely based on cloud. From this close integration between edge computing and AI, a new research area has arisen, namely “edge intelligence”. Edge intelligence aims to exploit peripheral and pervasive resources. The world's leading IT giants, including Google, IBM, Intel, Apple and Microsoft, have presented projects of various kinds to highlight the advantages of edge computing and the impact it can have on society in a wide spectrum of AI applications, ranging from live video analytics, to cognitive assistance, to smart home, precision agriculture, industry 4.0.