"Nobody much complains about human Police spotters identifying trouble makers at football matches. So is this just luddite objections?"
I have no experience of Police spotters at football matches as I've assiduously avoided football for my entire adult life. So I'm open to contradiction in this. But as I understand the comment spotters would be looking for known individuals who shouldn't be there. If that's the case then their visual identifications would be subject to more robust checks on identification. If automatic facial recognition is being applied to the lousy images from typical CCTV as the only evidence then it's far from being a Luddite objection. The initial ID, human or AI, needs to be tested and distrusted until it's tested. A violent arrest as in the Denver case linked above really shouldn't be conducted on the basis of such ID.
It's really like the blood test kits we used to give to SOCO - and used in the lab as well. The technical term was a presumptive test (it worked by detecting peroxidase activity which haemoglobin shares with several other substances such as fruit juice). The presumptive test highlights something to be followed up to decide not only is it's blood but also what species.
Facial ID seems to be at the presumptive level - useful to tell you what's worth looking at but too liable to give false positives to be used solo.
I see from the linked article that the FBI have ditched hair evidence. I never could understand why they were keen on it. It always seems hopeless. In 14 years I only ever used it once and that was to give an ID of a body where there were 3 alternatives; fortunately I could exclude 2 as couldn't be and that left one that could have been.