Judging a successful image
So far the AI program has generated tens of thousands of images, ranging from a hazy noise of green and blue to, sparing the oddly wobbly bath, almost believable bathrooms. But which of these images are ‘successful’.
This question is troubling and is down to the artist selecting what they deem to be ‘interesting’ or caches their attention. What the algorithm deems as successful is something else entirely. Where the element of power lies in terms of image selection therefore shifts from the computer to the artist. Each round of image generation the program spits out a number of images that meet the requirements to be demanded successful at that time. The artist selects from all rounds at the same time, looking for their mistakes as a measure of success.
Below are a selection of images deemed ‘successful’ by the artist and not the algorithm. They are selected because of their errors, removing any power the computer had in deciding if they were correct. They themselves have been through a selection process based around their successfulness.