Artificial Intelligence and Machine Learning are all around us. They are the key technologies that drive digital transformation in various industries and society at large. As we can see with recent technical breakthroughs like ChatGPT, we are still at the beginning of exploring the full potential and impact of AI/ML.

In mobile communications, AI/ML could upend how products are designed, built and deployed. The design of such solutions requires augmenting the engineering process with a data-driven approach that considers data quality, selection and governance. In other words, data will become a pivotal element of solution engineering. The benefits are solutions that can be constantly retrained and improved, taking advantage of powerful AI/ML models and continuous integration and deployment (CI/CD).

In the next generation of mobile networks, the integration of AI/ML will be a fundamental design principle so that these technologies will be able to fulfil their full potential.

5G-Advanced will pave the way, and 6G will be the first truly AI-native system

3GPP is the world’s leading standardization specification organization for mobile networks. Its ongoing work on the 5G-Advanced study of AI/ML in air interface offers the first glimpse of AI/ML that is embedded across device, radio and RAN. It will create the foundation for AI/ML features for all future releases to come, including 6G. The study focuses on a general framework, enabling a family of use cases utilizing AI/ML techniques. These include enhancements in channel information feedback, beam management and accuracy of device positioning.

In Rel-19, we can expect further AI/ML use cases to be addressed in air interface, RAN and system architecture. In Rel-20, 6G will be studied with AI/ML as an integral part of the system. Advanced technologies like distributed learning, in conjunction with AI deeply embedded in the network and devices, will boost performance and usability. In that sense, 6G will be the first generation of data-driven mobile networks.

The integration and embedding of AI/ML in mobile networks and standards will unleash a technological journey that the whole industry needs to take on together. It is a transformational task that touches on nearly all aspects of the mobile system. A careful and rigorous study is now needed to make the right decisions for coming future network generations, with trustworthiness and sustainability as key design principles.

The promises and challenges of AI/ML

The main promise and game changing property of AI/ML as a technology is that it can be better than human-created algorithms or can find solutions for problems that have yet to be solved. To satisfy the demand of future networks, the AI-native system will employ AI/ML-based solutions based on deep learning, reinforcement learning and distributed learning to provide superior performance and to adapt to changes and continuously improve in an agile manner. 

Examples of AI/ML solutions include:

• Enhancements in the devices and network for more efficient radio transmission and radio resource management. Ongoing work in 3GPP reports significant gains of up to 30% for cell edge throughput and other significant KPIs.
• AI/ML baseband solutions which outperform conventional approaches already today by up to 3dB in terms of receiver performance with throughput gains of 30%.
• Decreased power consumption and improved energy efficiency of network nodes and devices.
• Improved end-user perception for high-demand applications like AR/XR by reducing latency and network link failures.
• Full end-to-end automation of network management, including faster and more accurate predictions and reactions to network conditions and faults.

Additionally, AI/ML will increase operative efficiency by introducing MLOps as an extension of the DevOps software development and operations paradigm by integrating ML model training, development, deployment, and integration into a common and highly automated process.

However, the industry still has some challenges to overcome before the full potential of AI/ML can be unleashed. In standardization, an efficient and flexible operational framework for ML-based solutions needs to be defined. This includes life-cycle management for distributed or federated AI/ML with potentially joint inference in different network nodes and devices that will enable true end-to-end ML solutions and high-performance gains. Another challenge is to keep the effort for data collection tractable while still ensuring high quality input for model training. Finally, performance requirements need to be properly defined considering the adaptive nature of ML-based solutions.

These studies will be fundamental for the further adoption of AI/ML in mobile communications from 5G-Advanced to 6G and beyond.

The Nokia approach embraces sustainability and trustworthiness as key design principles 

At Nokia, AI/ML is not just about performance. Any improvement comes at a cost, so it is important for us to evaluate gains as part of our larger vision of guaranteeing sustainability and trustworthiness. 

Sustainability is a strategic goal of Nokia, and AI/ML will be key in reaching our sustainability goals by improving energy efficiency in network and devices alike. For example, AI/ML can be applied to reduce energy consumption by optimizing device channel measurements. In the network, AI/ML can be used to switch off unneeded components and ensuring seamless and superior user experience by incorporating the timing and context of past and future actions. However, the holistic energy footprint of AI/ML solutions, including training operations, also need to be further considered and optimized.

Nokia is similarly committed to Trustworthiness. Our mission is to apply AI/ML in a responsible, ethical and trustworthy manner. This includes rigorous and thorough evaluation and understanding of AI/ML solutions. That’s why we launched our 6 pillars of responsible AI platform.

AI/ML is already a fundamental cornerstone of mobile communication networks. But its full potential has yet to be reached. Overall, Nokia is a driving force to create the highest quality standards for AI/ML that are going to be the foundation of the AI-native communication networks to come.