AI in 6G development: direction and purpose

AI based on large language models offer considerable potential for developing communication networks. AI is being used in connection with communication networks at many levels. It boosts the efficiency of communication, provides a new platform for AI-based applications, and shapes the direction in which networks are developed to take advantage of faster and more extensive data transmission at higher frequencies. The recent publication, entitled, Large Language Models in the 6G-Enabled Computing Continuum: a White Paper, is edited by Lauri Lovén, Director of the Future Computing Group at the University of Oulu, who coordinates the Distributed AI research line in the national 6G Flagship research programme.

"During the first four generations of wireless, we got used to seeing data rates skyrocket to new heights at each turn, transforming the way people use the network. With 5G, the improvement didn't feel quite as radical, even though it was taking place. The focus of network development had shifted beyond improving the user experience, introducing various machine-to-machine communication scenarios and implementing network services using modern software development methods, making them easier and better to manage in many ways. Put simply, 5G integrated cloud services and edge computing into mobile networks. With 6G, wireless networks will be combined with artificial intelligence," Lovén explains the history of the six Gs of mobile technologies, using past tense for 5G. As a researcher, he is already looking to 2030s.

He adds that from the AI perspective, there are three key areas and viewpoints in 6G development.

"The focus areas are AI as a tool to support network operations (AI for RAN), AI-based applications and services that require mobile networks (AI with RAN), and adapting AI to the network (AI on RAN)."

Technology terminology is often not written in Finnish, so knowledge of English prepositions is an advantage. RAN is the acronym for Radio Access Network. It refers to the way in which different devices in a wireless communication network are connected to other parts of the network via a radio link. At the acronym level, the working title of the researchers' publication was LLM WP. Forty-six researchers helped write this open-access paper. The main purpose of the paper is to consider and demonstrate how 6G networks will use modern artificial intelligence methods to provide communication links and to offer AI-based services between individual devices, densely distributed base stations, massive data centres and cloud services. These networks make data available and create services for data collection and transmission within them.

“We focus on the technical aspects but also briefly discuss regulatory and application dimensions, as well as security, and resilience. Research on AI methods and ethical concerns is excluded from this publication. The writing process was demanding, but we are pleased with the final result. During the writing, there were differences of opinion. A small group of authors decided to withdraw their contributions altogether because they felt the changes we proposed were too extensive. No conflicts occurred, however. The rules of the game were clear from the start: the editors have the final say on content," concludes Lovén.

The content editors were Lauri Lovén, Miguel Bordallo López, Jaakko Sauvola and Sasu Tarkoma, who is also a professor at the University of Helsinki, and Roberto Morabito from EURECOM, a French technology research centre and university, who works in the 6G Flagship research programme at the University of Oulu.

For the wider AI debate, the LLM WP is a clear statement of direction, setting out a vision of where the development of a brand-new technology is going, where it fits in and what it offers. It is an opening for further research, projects and publications.

"Of course, the technologies for different AI models are developing at a dizzying pace, and so are the applications. On the tele-networking and computing continuum, from local devices to cloud computing centres, these may change not only the content of data transmitted and processed but also the methods of managing the network and its computing capacity. Meanwhile, we may soon have personal AI helpers, or even a whole army of them, and we can already see the enormous computing and data transmission requirements that will come with them."

But research looks beyond the hottest bubbles. That's its job.

"This was necessary and worthwhile to get published. In AI research, five years is a very long time. A lot can happen in a year! But data network cycles are slower. And standardisation takes a long time. One G takes about ten years. 6G is expected to hit the consumer market sometime in the early 2030s. That's why we've already been researching it for years. The development of artificial intelligence has introduced a whole new dimension. In the first generations of mobile communications, the network was essentially like the telephone network. It allowed people to communicate with each other. Then, alongside came human–machine interaction and, especially with 5G, autonomous machine-to-machine data transfer. The next steps will involve human-to-AI and AI-to-AI communication, with many questions still to be answered before reaching the stage of standardisation," concludes Lovén.

Download the entire 6G Flagship White Paper series for free.