The 10 Year Meme Is Not A Trick - CPG Rant
Recently there have been a number of people attempting to suggest that the 10-year meme is a trap, attempting to create some kind of training data set for a theorized facial-invariant recognizer model. This has been a response in The Discourse on Facebook, on Twitter, and on Facebook copying from Twitter. There is absolutely no reason to believe this is true.
[My opinions are my own and not that of my employer.]
The technology to do facial invariant detection like this on humans is old news. (At this point, it even works on pets!)
Facebook doesn't need this: they have access to your first profile picture and your most recent profile picture. Google doesn't need this: They shipped facial invariant detection at Google I/O more than 5 years ago. Others don't need this: The techniques that Google and Facebook use are largely common knowledge as the result of published research.
Even if there were a use for a model like this, getting the information out of a platform like Facebook is not significantly more difficult than just extracting public photos in the first place, so any actor other than Facebook would be just as well off going and grabbing some of your older photos from your public photo rolls; this training data literally offers nothing new that can't already be done .
This entire idea of this being a trap is ridiculous. It's just a fun meme. Do it, don't do it, but don't avoid it because of fear of some algorithmic influence: That ship sailed already a long time ago.
This is a CPG Rant, a brief interlude routinely featured in my livestreams which comments on some social, political, or technical happening in the world, giving you my honest -- and usually shouty -- opinion.