YOU ARE AT:6GAI in 6G: A 'pervasive' view

AI in 6G: A ‘pervasive’ view

Artificial intelligence is expected to play a major role in future 6G systems, to the point that 6G is already being talked about as an “AI-native” technology, with a holistic and pervasive approach to integrating and using AI within the network, as well as enabling more AI adoption.

But what does an AI-native telecommunications system actually look like, and how does the industry build 6G so that it will achieve that level of AI integration?

Dr. Yue Wang, chief technologist for network and AI at China Telecom, said during the recent 6G Forum that AI-native is really about “pervasive” AI, and offered a vision on how that translates at a system level. “For me, pervasive AI is the deep and wide-spread integration of AI throughout the 6G network. It’s operations and services — it basically means AI is becoming an integral part of both the network infrastructure, and the application that runs on top of it, that enables the autonomous, more efficient network management, as well as service delivery,” Wang explained.

She went on to say that there are two main aspects of pervasive AI in a 6G context: AI within the 6G system itself, where it would be supporting network operations in order to promote energy-efficient operations or enable intelligent service provision, for example.

The second aspect would be the interaction of the network with multiple other AI systems, which could be either internal or external to the network: edge AI, cloud AI, or AI within devices, for example.

“The network will need to facilitate this seamless communication between these AI systems and enable them to work together in a real-time environment,” she added.

AI efforts within standardization grounds are initially focused in multiple areas, Wang said, ranging from the Radio Access Network and taking multi-cell approaches for things like mobility management, energy savings and coverage and capacity optimization; genAI is also being used to reduce operational costs through the use of custom chatbots. But for 6G to be a pervasive, AI-native system, AI has to move into the network with an end-to-end view rather than one that is use-case or application-specific.

“When we consider the use of AI in the system, we want it to be holistic,” Wang pointed out. That also may mean that if standards groups want to allow for integration and utilization of AI, they, too, will have to look across domains with an end-to-end view. While Wang noted that she was not commenting on any specific standard, said said that from a broad technology perspective, “I think we need to streamline the AI operation and allow a holistic approach in the network—not just for one use case, not just in one domain.”

At the moment, she explained, the industry is taking a “piecemeal approach of integrating AI” on a use-case-by-use-case basis. As use cases are patched in to the network to achieve specific goals, sometimes two patches aren’t going to work well with each other, requiring even more patching. An AI-native approach to 6G, then, implies that not only will AI be commonly found, and able to be accommodated, within and outside of telecom networks, but that the various AI systems will be able to collaborate with one another to function optimally—and maybe even organically handle uses and demands that aren’t manually built out or anticipated by humans.

“When we say AI-native, when AI application becomes widespread in the network, we actually do need to figure out out how they can work together, work interactively and collaboratively across-domain and end-to-end, and to address [many] more use cases simultaneously,” Wang reflected, going on to add that whether AI models are in the RAN, the core or the cloud, they need to be managed in a unified manner, operated and life-cycle-managed across the network, with the computational and communications resource needs of the models factored in.

Further complicating the standards work is that AI is changing to rapidly. How do you build a new technology on a shifting foundation that has clearly not yet reached its full potential—but will probably be closer to it, by the time 6G actually emerges in the 2030 timeframe?

“I think we do need to bear in mind, when the network is deeply integrated with technologies that are outside of our traditional domain—which is communication technology—they don’t necessarily follow the pace that the standard would like to take,” Wang concluded, making note of the relatively rapid releases of AI models. AI, she went on, “is obviously a different system, and a different nature of innovation, but I think now that the network is becoming more software-ized and more virtualized, more cloud-native, I think that give opportunities to us to strike the right balance … between the standard and allowing room for innovation.”

ABOUT AUTHOR

Kelly Hill
Kelly Hill
Kelly reports on network test and measurement, as well as the use of big data and analytics. She first covered the wireless industry for RCR Wireless News in 2005, focusing on carriers and mobile virtual network operators, then took a few years’ hiatus and returned to RCR Wireless News to write about heterogeneous networks and network infrastructure. Kelly is an Ohio native with a masters degree in journalism from the University of California, Berkeley, where she focused on science writing and multimedia. She has written for the San Francisco Chronicle, The Oregonian and The Canton Repository. Follow her on Twitter: @khillrcr