Deep Learning Kubrick (2016). Making use of current image-recognition software, Deep Learning Kubrick explores the idea of AI trying to make sense of stories and fictions; in this case snippets from classics by the film director Stanley Kubrick.

Taking stills from these snippets at four-second intervals, the algorithm analyses and tries to describe what it sees on the image. Without any knowledge of what happened before and what will happen after each still, and without any cultural context of cinema, let alone of a Kubrick film, an alternative narrative emerges out of its descriptions. As humanity comes to terms with the existence of other forms of intelligence, our ability (and evolutionary advantage) to understand and believe in fiction might be the next frontier for AI to master.

The first iteration of Deep Learning Kubrick makes use of an image recognition software by the University of Toronto (which contributed towards a nomination for the Tate IK Prize 2016). For this year’s Ars Electronica Festival I revisited the piece, made use of Microsoft’s new image-recognition bot and used different excerpts from Kubrick films.

Video and Images Courtesy of Michel Erler.