After "uncanny" is "undetectable"

It's official. We can't believe our lying eyes.

Here is a video which, to this moment, has garnered 1.7 million views; 1,830 comments and 44 thousand likes – but no one has commented on the practical certainty that it is AI-generated.

Don't believe it? Consider that this heroic young woman has no name, is completely symmetrical, is possessed of a perfect complexion and is presented to us by a channel which does nothing but post heartwarming imagery much more obviously AI-generated than this. The fact that it was generated in the past few weeks is a bad sign, because that is when such photorealistic shams have gotten very good indeed.

Still not convinced? Notice that the writing is nonsense. OK, maybe it is Cyrillic, but it is still nonsense. Yes, her fingers are pretty reasonable (though that thumb is a little long) but within the bounds of the plausible, so we can't rely on AI's notorious difficulty with our clever humanoid digits. Notice though, that while her watch has the winder on the correct side (indicating that the image has not been reflected for layout purposes) and it agrees with that stylized K on her red shirt – its handedness disagrees with the big sign back behind her head. Even if the sign is some Cyrillic neologism, it is reversed for no plausible reason.

Claude AI judges the likelihood this image is AI-generated as at least 85% and that is without considering various "tells" which I shall refrain from identifying at this time. There are some systematic failures which the perpetrators of this kind of deception seem to be unaware of and which we should not rush to help them understand. Join in the fun and Spot The Bot with some help from check lists and other resources we're curating.

The point of this post, though, is to establish that we are now well past the uncanny valley. Humans must be very alert, suspicious and even possessed of some expertise to be able to detect the handiwork of the bots and the humans who wield them. This particular deceit plucks at our heartstrings but is not, in itself, particularly dangerous. The capacity for wholesale deception which it demonstrates though (remember 1.7 million people were apparently bamboozled by this) should utterly terrify us.

The take-home point is – we can no longer believe anything our lying eyes might tell us, and – we are going to have to figure out what this means for politics, policy and technology.  

Here at the Noospheric Software Foundation we strive not only to contribute to this dialog but to the creation of fundamental technologies which might support the process of us collectively making sense of even the most complex puzzles.