Maybe they'll explain to TV producers that facial recognition software doesn't work by showing each face it's checking
It doesn't, and it shouldn't but it could do something along those lines by way of feedback.
I've written a number of systems that have lengthy batch processes that flashes up record information as the system moves through; it doesn't show every record it passes as the screen updates would slow down the system enormously, but it updates a couple times a second and shows for example, every 10000th or so giving them real feedback that something is actually happening without slowing it down at all.
The progress bars that are time 'calibrated' and do not bear any reflection to actual progress are the bane of my existence; where the process hangs at 50% at stops dead, but the time calibrated progress bar just drifts along to 99% and then eventually reports that an error happened.
Those updates that actual data is being processed are good user feedback that its actually doing something.
Of course I don't think facial recogntion is done with a cursor search from the start of a database to finish, the way a system batch processing transactions would be. Instead I imagine they work more like traditional databases, breaking the images down to collections of indexable information and searching the indices; so a record-by-record walk wouldn't be necessary, or perhaps would only be necessary as the final pass through a returned set to further score and sort the results.
In any case, the trouble with TV facial recognition portrayals is less the software itself (because I can handle a dramatization of a computer search like that), I'm more offended by the portrayol of the results. There are no false positives (finding the wrong people) and false negatives, (failing to find people who ARE in the system), or multiple results. No its always either... face goes in and perp comes out... or face goes in and computer declares the person doesn't exist.