Mario Klingemann 
Appropriate Response, 2020

125 Letter Split-Flap Display, Kneeler, Computer running custom trained GPT-2 model.
75 x 250 cm x 200 cm

A reflection on meaning by the AI-art pioneer, Mario Klingemann 

How much can be expressed in just 125 letters? Appropriate Response, by Mario Klingemann, addresses this question. Inspired by the power of words, this interactive piece explores meaning, expectation and relationship with artificial intelligence. 

The installation features a wooden kneeler and split flap display, which shows a random selection of continuously changing letters. When a person uses the kneeler, the installation’s built-in artificial intelligence responds by presenting a short sentence on the display. Each phrase is written by the machine’s neural networks and is entirely unique; no two visitors will ever receive the same line of distilled wisdom from Appropriate Response.

The power of words

From religious principles to marketing slogans or self-help, pithy phrases abound as a source of inspiration and guidance. Yet with today’s technology, it can be difficult to discern whether certain texts have been produced by humans or machines. Within this context, Klingemann’s artwork raises pertinent questions about authorship and the significance attached to written language.

Physicality, ritual and interpretation

For Appropriate Response, Klingemann enhanced the GPT2 neural network with a further 60,000 quotes to create a model capable of producing short texts. These appear on a split flap display, chosen by the artist for its aesthetic appeal, distinctive sound and connotation of waiting.

The kneeler turns interaction with Appropriate Response into a ritual-like experience. “We fear AI but also hope it might help us,” says Klingemann. “That balance between hope and fear is related to religious experience, so I felt kneeling was very fitting.”

Context and expectation are also key. Viewers participate by kneeling but also by processing the text shown on the display. Appropriate Response generates coherent aphorisms but it is human viewers that furnish them with meaning.