This tutorial provides a broad vision of AI-native joint communication and computation for 6G networks and beyond, emphasizing the integration of AI into modeling, resource allocation, and decentralized management. We begin by revisiting traditional RAN architectures, cloud/edge architectures, and multi-cloud networks so as to establish the foundational system model of the networks under consideration. Building on this foundation, we define a joint communication and computation provisioning architecture for next-generation wireless networks. Focusing on multi-tier computing, we then explore how AI-driven decision-making enhances latency-sensitive applications across the cloud-edge continuum, by shedding light on timely multiple access
mechanisms, e.g., rate-splitting multiple access (RSMA) and cooperative non-orthogonal multiple access (Co-NOMA). Such a cross-layer approach is especially poised to enable distributed intelligence, laying the groundwork for end-to-end AI in communication and computation.

The tutorial afterward examines the synergistic benefits of AI-native
hybrid networks, focusing on joint and distributed resource allocation
schemes. The tutorial progresses from a single-cell deep learning
framework for robust delay minimization to a multi-cell
interference-limited setting, ultimately leading to a federated learning
approach for distributed resource management. Throughout the tutorial,
we provide comprehensive modeling, analysis, and optimization of the
exemplified multi-tier networks, addressing their applications, enabling
technologies, and the mathematical methodologies required for their
systematic design. A major highlight of this tutorial is the crucial role of AI in spearheading joint communication and computation resource management toward seamlessly enabling future decentralized 6G networks and beyond.