What is Semantic Communication? A View on Conveying Meaning in the Era of Machine Intelligence

In the 1940s, Claude Shannon developed information theory focusing on quantifying maximum data rate that can be supported by a communication channel. Guided this fundamental work, main theme of wireless system design up until fifth generation (5G) was maximization. Shannon's theory, semantic aspect and meaning messages were treated as largely irrelevant to communication. The classic started reveal its limitations in modern era machine intelligence, consisting synergy between Internet-of-things (IoT) artificial intelligence (AI). By broadening scope communication-theoretic framework, article, we present view (SemCom) conveying through systems. We address three modalities: human-to-human (H2H), human-to-machine (H2M), machine-to-machine (M2M) communications. latter two represent paradigm shift computing, define article. H2M SemCom refers techniques for meanings understandable not only humans but also machines so they have interaction "dialogue". On other hand, M2M effective efficiently connecting multiple such effectively execute specific computation task network. first part article focuses introducing principles including encoding, layered architecture, approaches: 1) layer-coupling design; 2) end-to-end using neural second discussion different application areas [including human AI symbiosis, recommendation, sensing care, virtual reality (VR)/augmented (AR)] (including distributed learning, split inference, consensus, machine-vision cameras). Finally, discuss approach designing systems based knowledge graphs. believe comprehensive introduction will provide useful guide into emerging area is expected play an important role sixth (6G) featuring connected integrated sensing, communication, control.