6G and AI Convergence: Network Evolution from "Connectivity" to "Intelligence-Native"

I. From 5G to 6G: Not Just Faster, But Smarter

While 5G networks are still being deployed, the outline of 6G is becoming increasingly clear on the global communications landscape. In June 2023, the Radiocommunication Sector of the International Telecommunication Union approved the "Framework and Overall Objectives for the Future Development of IMT for 2030 and Beyond," defining six major application scenarios and four design principles. Unlike 5G's "golden triangle," 6G extends to new dimensions like "ubiquitous intelligence," forming a hexagonal vision.

However, the real transformation of 6G goes beyond specifications; it represents a fundamental change in network form. As one industry expert observed, the core change lies in the shift from a "connectivity network" to an "intelligent platform."

II. Three Phases of AI and 6G Integration

Based on relevant academic research, the convergence of AI and 6G networks can be divided into three progressive phases:

Phase 1: AI for Network (AI Empowers the Network)

This phase uses AI to enhance network performance, improve efficiency, and optimize user experience. It operates on two levels:

Network Element Intelligence: For individual network elements and functional entities, AI optimizes traditional algorithms. For example, AI-based channel state information feedback demonstrates significant advantages in performance and throughput over traditional codebook methods. AI-driven beam management selects the optimal beam without scanning all beam pairs, substantially reducing scanning overhead.

Operations and Maintenance Intelligence: AI automates tasks across network layers and operations, including network traffic detection and congestion control, traffic prediction and scheduling, and dynamic load balancing and interference avoidance.

Phase 2: Network for AI (Network Supports AI)

The network is not only optimized by AI but also actively provides architectural support for AI operations. This requires 6G networks to possess:

• Distributed Intelligence Architecture: Supports flexible deployment and coordination of AI tasks across cloud, edge, and terminal

• Multi-Dimensional Resource Integration: Unifies scheduling of computing, data, and communication resources to provide a complete runtime environment for AI workflows

• Digital Twin Support: AI learns and predicts complex dynamic environments, bringing digital twin systems closer to reality

Phase 3: AI as a Service (AI as a Native Service)

Future 6G networks will natively provide AI capabilities as a service, supporting applications like immersive communications, intelligent industrial robots, and embodied AI. The service targets will also expand from human users to "machine users" such as robots and intelligent vehicles.

III. Large Models and Wireless Networks: A Key Driver for 6G

Compared to large language models, wireless network large models have their own distinct characteristics. They need to process wireless air-interface data such as channel state information, inter-cell interference, and multipath delay, as well as network data like KPIs and operational logs. Their application directions cover channel prediction, wireless channel modeling, intelligent routing, and service identification.

Key challenges in building wireless network large models include: how to deploy large models in communication networks with stringent real-time requirements? How to balance model capability with inference latency? The answers to these questions will determine whether the vision of "native intelligence" in 6G can be truly realized.

IV. Perspectives for Industrial Communication Equipment Manufacturers

With 6G still in the research and pre-standardization phase, understanding the trends in AI-communication convergence can help industrial communication equipment manufacturers prepare in the following areas:

1. Edge AI Capability Pre-research: Future 6G industrial routers may evolve from simple "data pipes" to edge intelligent nodes capable of local inference. Real-time industrial control and anomaly detection could be processed locally, reducing cloud dependency and transmission latency.

2. Semantic Communication Monitoring: Semantic communication aims to "transmit not just bits, but meaning." In industrial settings, this could mean compressing transmission bandwidth while ensuring accurate delivery of control commands, potentially revolutionizing communication efficiency in industrial IoT.

3. Device Intelligence Evolution: As networks move toward "native intelligence," industrial communication devices will also become more intelligent—evolving from "configurable" to "adaptive," automatically optimizing parameters based on network conditions and reducing manual intervention.

V. Conclusion

The convergence of 6G and AI represents a fundamental shift from "AI-assisted networks" to "AI-native networks." The three progressive phases—AI for Network, Network for AI, and AI as a Service—outline a clear technological evolution path.

For the industrial communication sector, while 6G commercialization is still years away, understanding this trend and preparing in areas like edge intelligence and protocol optimization will help secure a favorable position in the next wave of communication technology transformation.