Stuff like locating stuff for beamforming or MIMO has a non-zero cost, and newer standards are getting even more complex trying to drag signal out of noise and

Stuff like locating stuff for beamforming or MIMO has a non-zero cost, and newer standards are getting even more complex trying to drag signal out of noise and error correction and stuff. The first is exactly the sort of thing where a cheaper "Good Enough Estimate" may be a great solution - which is what a fair bit of "AI" hardware does today. Though I don't know if firing up a separate AI block instead of an integrated to the radio "hardware" solution would be better, but it may be there's some crossover of techniques that can help.