The self-portrait of Dorian Gray

We asked an Artificial Intelligence (AI) to make a self-portrait.

By Kevin Erkeletyan

“/imagine the realistic self-portrait of an artificial intelligence.” The prompt was deliberately open-ended. The aim was not to obtain a precise order, but to discover what is hidden in the “unconscious” – so to speak – of Midjourney, the artificial intelligence (AI) that generates new images from all those it has already ingested. What image does an AI have of itself? How does it represent itself?

The software gets going: an initial sketch appears, then the result becomes clearer. It offers us four graphic (or should we say artistic?) styles of the same reality. We choose one. Based on this, Midjourney creates four versions of an AI’s self-portrait.

DO YOU RECOGNISE IT?

Four faces take over the screen, representing a progression. A young man with a mechanical pallor seems to be nibbled away by technology: we can make out pixels, electronic chips, a network of neurons, then an assembly of parts and a bionic eye. His blue eyes seem dreamy, sad, attentive and determined in turn. Apathetic to begin with, he gradually steels up.

Is this quadriptych scary, reassuring or simply fascinating? AI paints a fairly unsurprising picture of itself halfway between man and machine, directly reflecting what we are used to in science fiction. We recognise Robocop, Terminator and the precogs of Minority Report. Perhaps it is trying to remind us that an AI’s self-portrait is above all our own… Which one would you pick?