This is totally a Doctor Who plot waiting to happen; it turns out, that all the faces we see in all the random places in the world, are actually ALL THE SAME FACE.
Edible by vegetarians.
I wish I’d never said “enhance” 
The artist knows he is making pictures of the face detection algorithm (per his notes on the site), so the claim “with no human faces as inputs” is a little nuanced. The definition of a human face embodied in the algorithm is also an input.
For input sets with no photos of humans, the output is actually made from exactly one face. It (obviously) should be heavily biased by the face detection algorithm that is chosen.
Still, seriously cool results.
I came here to make that point, but I think you perhaps made it more eloquently anyhow , so …thanks!
I had to read your comment a couple of times, but I’m glad I did. This is a really interesting and compelling example to remind people that the tool used for analysis, and the questions asked in analysis are part of the resulting answer. This theme runs through all the reporting on the biases that exist within algorithmic systems, including everything from market analysis to police databases, etc…The headline gif for this article should be included in all stories with this theme…
This topic was automatically closed after 5 days. New replies are no longer allowed.
