> Just why the 5G edge needs AI wasn't explained to The Register.
Variety of use cases available in the real-time optimisation category that vendors are happy to say are "AI enabled" to make them sound more modern and cutting edge.
Plus the posibility of deploying a model to be trained/tested/validated on bits of the network without having to backhaul loads of data to a central server (farm).
a) radio parameter optimisation (handover thresholds, antenna tilt adjustments)
b) mobile geolocation
c) traffic analysis/prediction
And that's all in the existing traditional fields, which can further expand when you start playing with multi-user-MIMO.
Regular analytical techniques can require a lot of modelling expertise - AI can possibly simplify this, assuming deployed correctly (and no doubt mistakes will be made...)