The tinfoil hat part of me says its like the "identify the stoplights" challenges.

We'll introduce this idea that sounds really tech, and hype. Maybe imply its so difficult that computers could never compete with humans.

We'll say its about improving security, and most humans will get a vague comforting feeling from the idea that there's some security feature, even if they find it kind of annoying to interact with.

Five years later, we say "surprise, it was really about training an eyeball recognition database that we can sell. Just like the image recognition datasets."

I do not even kind of believe these people are dumb.