The second aspect to my Error and Power proposal was for the artist to construct an image based on instructions from a computer.

If the artist were to follow the instructions laid out by the JPG algorithm what would be created? 

We’ve seen how an AI is an idiot, bludgeoning its way to a conclusion, interpreting things only on its terms, now the it is the turn of the artist to be the idiot.

A JPG sees an image broken down into extremely small pieces with the algorithm travelling a zig-zag pattern through these pieces, generalising as it goes along. This amount of information is probably too much for the artist to comprehend so to begin we must ask the computer to simplify things vastly.

A little program settles on a suitable amount of data, breaking the image down into 24 squares rather than the thousands it would usually. It then takes note of the red, green and blue from each square, does some magic and turns it into red, yellow and blue so the artist can work out how to paint from it.

So with this set of numbers, now what? The computer needs to present it to the artist in a way the artist can understand. For the example here it shows a square, much like the square it ‘saw’ earlier but instead sorts the red, yellow and blue into order of amounts and fills the square in blocks of these colours accordingly. It then passes through the simple image in the same zig-zag path it did earlier. 

The artist, an idiot, finds this harder than it should be to follow.