Edge AI for 6G Networks
The roundtable debate titled "Edge AI for 6G Networks", organised by IEEE Communications Society, is interesting. The participants are a select group of people from industry and universities. The debate gives insights into the role and future of edge AI in telecommunications, both in (1) operational functions of the networks (e.g., the RIC function in Open RAN, physical layer support of massive MIMO beamforming, etc.) and (2) in making money with edge AI by the providers for offering/hosting AI services in the edge (base stations) for Over-the-Top (OTT) service providers (e.g., Netflix).
The operational application is not at all controversial and is necessary to further drive up the quality of network services and capacity, so this is happening.
The second role of AI, as a revenue model is seen as problematic, as customers, OTT companies, have traditionally been reluctant to share their revenues with telecom "data pipeline" providers. Investment in AI edge infrastructure will therefore be difficult to recoup.
There is a consensus that there does seem to be a future in application-enabled edge-AI in private 5G/6G networks targeting specific verticals, e.g., the healthcare sector for enabling "remote and robotic surgery". This requires intensive collaboration with these verticals. Traditional operators are not seen as a primary actor here.
An interesting point of interest that is touched upon is the energy consumption in base stations, a major cost when they start doing edge AI computing.
There is also discussion about the architecture, particularly the balance between what happens in the cloud for certain applications and what happens at the edge. It is considered likely that many applications will require an optimised hybrid architecture: the heavy work happens in the cloud and the time-critical work in the edge.
Check it out further on
