RNN Descriptions

All of these paintings, drawings, and sculptures are inspired by the output of a neural network - a TextRNN Recurrent Neural Network trained on a collection of artwork descriptions (most of which were things like "man looking at pair of birds at left"). After training and generating a series of results, I hand-picked a corpus of a few thousand "interesting" descriptions that were then fed back into the RNN for additional training. This process was repeated 6 more times, with every successive training producing stranger and stranger results biased towards certain elements that I found entertaining (things like the triangle figures, the gold mask cult, and the term "geometry"). I take the resulting text descriptions and paint or draw what I imagine them to describe, just as (my frequent collaborator) Thorin has done when he wrote his book, One Long Panel of Stones