The idea behind picvolution.org is to reverse engineer pictures by adapting them to our mental expectations. Every human mind on this planet might have a different definition of what is art or what makes a great picture. However, if the decision that one picture is better or more beautiful than the next is taken by a large number of people independently and nevertheless shows a similar tendency, this could mean there is some agreeable content of art in that picture. So by looking at all possible pictures and comparing them side by side, we could establish a global ranking of pictures, with the one that gets the most likes as the universal piece of art. Of course the number of possible pictures that can be generated even on a small canvas is almost infinitely large and hence this undertaking would take a long time...
picvolution.org uses an evolutionary algorithm to establish the pictures that have accumulated the most hits over a certain period. These pictures are more likely to get copied into child pictures with random errors. Out of all the child pictures, there might be one that is even more likeable than its parent and ends up winning the next round of selection. The expectation is that over time the pictures get more likeable, as we move upwards in the design space of art.
picvolution.org relies on you, the crowd, to select a picture. The directionality of the selection is implied with the question "Which picture looks more like a face?" Human minds are good at seeking patterns, especially faces. You are thus kindly asked to click on the picture that most reminds you of a face. This should speed up the trajectory of improvement. However, what the pictures eventually evolve into is entirely up to your selection.
And remember: Evolution is a very slow process. So if you are dissapointed by the lack of improvement, come back in a million years or so ;)
picvolution.org is set up and maintained by Tobias Großkopf. For contact please use this link.