"not post enough pictures that allow training an accurate representation" - well then you just have to post no pictures, and even that won't be sufficient as yo

"not post enough pictures that allow training an accurate representation" - well then you just have to post no pictures, and even that won't be sufficient as you're likely to end up in photos that others post, like the original article example of Andres Rafael Viola who was identified because he just happened to be in the corner of the background on someone else's photo. A few years ago I had a project that among other things needed to detect which politician is in a particular photo, and an accurate representation needed something like 5-6 photos from different angles back then; more pics helped, of course, but you could get quite good results with just a few images even back then. I presume that we're not far from tech where 1 or 2 at most would be sufficient, so if any photos at all exist on social media (e.g. a buddy has taken a group photo and marked you in it from their contact list) then you are or will be linked to your photos. And IIRC that recognition is quite stable as you age if properly trained, so if there's a ten year old photo floating around somewhere, then that part of your privacy is already gone.