They believe they’ve managed to create a system capable of identifying criminals from non-criminals simply by analysing pictures of faces. Seriously. That’s what they’re actually saying.
After feeding their AI 1,856 mugshots, they claim to have found “some discriminating structural features for predicting criminality, such as lip curvature, eye inner corner distance, and the so-called nose-mouth angle.”
It’s almost as if criminals and non-criminals are slightly different species of human, as if criminals were born to do bad things. Plug incoming…
It’s clearly utter garbage, but at least the conclusions are entertaining.
“By extensive experiments and vigorous cross validations, we have demonstrated that via supervised machine learning, data-driven face classifiers are able to make reliable inference on criminality,” the report reads. “After controlled for race, gender and age, the general law-abiding public have facial appearances that vary in significantly lesser degree than criminals.” [Arxiv via RT]