6G Artificial Intelligence Service

Description

The 6G AI Service is a service concept introduced in 3GPP Release 20, as specified in TS 22.870. It represents a service provided by the 6th Generation (6G) network where AI functionalities—such as AI model training, AI model management, or AI inference—are made available to a subscriber/user or an authorized application on the User Equipment (UE) or on an application server (AS). This service is a cornerstone of the 6G vision, which aims to integrate AI natively into the network to enable intelligent, autonomous, and personalized services.

The 6G AI Service encompasses a wide range of AI capabilities that can be offered by the network. AI model training involves the network or the UE training machine learning models using data collected from various sources, such as network telemetry, user data, or sensor data. AI model management includes the lifecycle management of AI models, including deployment, updates, and monitoring. AI inference refers to the execution of trained models to make predictions or decisions, which can be performed at the UE, at the application server, or within the network itself. The service can be tailored to different use cases, such as autonomous driving, smart manufacturing, healthcare, and immersive entertainment.

Architecturally, the 6G AI Service is expected to be supported by a distributed AI framework that spans the UE, the edge, and the cloud. The 6G core network will include AI-related network functions, such as an AI Service Function (AISF) or AI Management Function, which coordinate the provision of AI services. These functions interact with other core network functions, such as the Session Management Function (SMF) and Policy Control Function (PCF), to ensure that AI services are delivered with the required QoS and security. The RAN may also play a role in AI service delivery, especially for low-latency inference at the edge.

The service is designed to be flexible and scalable, supporting various deployment models. For example, an AI service can be hosted entirely on the UE, on an edge server, or in the cloud, depending on factors such as latency, privacy, and computational resources. The network can dynamically decide where to execute AI tasks based on context, such as network conditions, user preferences, and application requirements. This requires seamless orchestration and management of AI resources across the network.

Security and privacy are critical aspects of the 6G AI Service. The service must ensure that AI models and data are protected from unauthorized access and tampering. It also must comply with privacy regulations, such as GDPR, by implementing techniques like federated learning and differential privacy. The 3GPP is defining the security and privacy requirements for AI services in 6G, which will be essential for building trust and enabling widespread adoption.

The 6G AI Service is expected to enable a new generation of applications that are intelligent, context-aware, and proactive. It will transform how users interact with the network and how applications are developed and deployed. By making AI a native service, 6G will democratize access to AI capabilities, allowing developers to create innovative solutions without deep AI expertise. This service is a key differentiator for 6G compared to previous generations, which treated AI as an add-on rather than an integral part of the network.

Purpose & Motivation

The 6G AI Service exists to fulfill the vision of a fully intelligent and autonomous 6G network that natively integrates AI capabilities to serve a wide range of applications and users. In previous generations, AI was often implemented as an over-the-top service or as a network optimization tool, but not as a core service offering. The 6G AI Service changes this by making AI functionalities available as a service to subscribers and applications, enabling new business models and use cases.

The motivation for this service comes from the increasing demand for AI in various domains, such as autonomous systems, personalized services, and real-time analytics. Traditional approaches to AI deployment face challenges such as high latency, limited scalability, and privacy concerns. By providing AI as a network service, 6G can address these challenges by leveraging the network's distributed computing, storage, and connectivity capabilities. For example, AI inference can be performed at the edge to reduce latency, while model training can be done in the cloud using vast amounts of data.

The 6G AI Service also solves the problem of fragmentation in AI deployment. Currently, AI models are developed and deployed in silos, leading to inefficiencies and lack of interoperability. By standardizing AI service delivery in 3GPP, the industry can ensure that AI models and applications can be ported across different networks and devices, fostering a vibrant ecosystem. This is essential for achieving the scale and ubiquity required for 6G.

Furthermore, the 6G AI Service addresses the need for trust and security in AI. By integrating security and privacy mechanisms into the service, 3GPP ensures that AI services are reliable and compliant with regulations. This is crucial for applications in sensitive areas such as healthcare and finance. The service also enables new revenue streams for operators, who can offer AI-as-a-Service to enterprises and consumers, differentiating their offerings in a competitive market.