Teaching a GAN the Alphabet

Teaching an alphabet to a GAN (Generative Adversarial Network) is attached to an art-historical lineage. Baldessari’s 1972 work titled “Teaching a Plant the Alphabet” was once described by Coosje van Bruggen as “quite perverse, futile, a conceptual exercise” in its time. It was an absurdity about learning and recognition. In this video, Baldessari extends a piece of paper with a letter of the alphabet towards an ordinary houseplant. He repeats the letter aggressively and quickly (A-A-A-…) and after about thirty seconds, moves on to the next letter in the alphabet.

This work may seem illogical and foolish to the viewer, but it’s possible to examine it beyond ordinary themes and irony. Baldessari executed this piece in response to Joseph Beuys’ 1965 performance “How to Explain Pictures to a Dead Hare.” The logically executed illogical educational task provides the viewer and the plant with a greater experience. Taking this idea a step further, I am teaching a computer, a lifeless being like a dead wild rabbit or a houseplant, the English alphabet using thousands of different fonts and styles through the method of self-learning (machine learning) using the Generative Adversarial Network (GAN) technique. To implement the algorithm, I am using a platform called playform.ai developed by Ahmed Elgammal, a computer science professor at the same university as me.