Open Access

Matthias Grund, Lasse Scherffig

• The question “What’s hidden in the hidden layers?” has been asked in artificial intelligence research since at least 1989. Likewise, today's answers to the challenges of deep learning remain the same as they were 35 years ago – from backpropagation to feature visualization and synthetic training data. Even the applications remain the same: nonlinear classification and self-driving cars. Once again, this reveals the surprising continuity of methods, applications, and rhetoric in the field of machine learning, that claims perpetual novelty, even though its roots stretch all the way back to cybernetics and its origins in control engineering. But the secrets of the hidden layers, it has become clear, are not only what is revealed by looking at their weights or their activation in response to particular inputs. This collection of texts explores the key themes and discussions of the second edition of the Hidden Layers conference, held at Köln International School of Design from June 12 to 15, 2024. Offering various perspectives on ‘hidden layers’, the conference was organized in two thematic tracks: “AI, Code & Material” and “AI, Society & (Visual) Culture.”