Confusing Reality for AI
The Boston Globe recently published an article about this video. The very short version: the author thought the video might be AI, and he investigated.
From the Globe article by Spencer Buell:
I’m a reporter, so I’m skeptical by nature. My eyes were drawn to the uncanny nature of what I was seeing. Did the herky-jerky movements of the people on the unicycles seem human, or robotic? Did the oddly shaped shovel one of them is holding look somehow wrong?
And come to think of it, wasn’t a scene of a jester and his family shoveling snow on unicycles exactly the kind of thing you’d expect someone to type into an AI generator as a prompt?
Before you read any further, watch the video for yourself and see what you think.
Indeed, the front shovel is weird, but I've seen that shape at Home Depot. The motion is jerky, but that's the nature of riding a unicycle, especially when doing so slowly. My friend once had to bring his unicycle across town while the rest of us were on foot (long story), and he jolted along to accommodate our pace. The music may be ear-grabbing, but most short-form video benefits from that for the same reason slop does. At the very least, we have to admit the clip is plausible.
Unfortunately, I never got to try evaluating the video for myself. I only saw it because my aforementioned friend sent me the article. He knows a lot about unicycles, and even without going to the reporter's lengths he could tell the unicyclists were real. Especially because he's the one in the middle.
He, his brother, and their father Alex have been doing stunts like this for years. The original video is from 2017! Hopefully, that will convince the most hardened of cynics.
It's also worth watching their Halloween video from the same year.
So, the video is difinitively real. And yet, it apparently had some reporters swearing it was fake until further investigation.
We talk more about the opposite. AI is getting good enough to sometimes fool even very competent spotters, at least briefly. That's a central facet of the "post-truth" phenomenon. Consequently, the savvy have become skeptical of a lot of online media. As generative AI gets better at fooling us, we get warier. In the long run, that mistrust just might be more toxic than the slop itself.
Trusting nothing is surely better than trusting everything. Still, I have to wonder where it will lead us. The everyman already has waning trust in anything he percieves as an instutution. The now-mainstream right wing in the U.S. rejected, and continues to reject, scientific consensus — look where that's getting us. And because truth is ephemeral, the regime will happily shoot one face and lie about it to the next.
The cat is out of the bag. How do we build a new trust? How do we earn it?