Neural Architecture

Matias del Campo is a registered architect, designer and educator. Founded together with Sandra Manninger in Vienna 2003, SPAN is a globally acting practice best known for their application of contemporary technologies in architectural production. Their award-winning architectural designs are informed by advanced geometry, computational methodologies, and philosophical inquiry.

• Acknowledgments10 Preface, Mario Carpo16 Prologue, Matias del Campo22 Chapter I – This Building Does Not Exist —An Attempt at a Theory of Neural ArchitectureWhat does a Melange have to do with Artificial Intelligence?Familiar but Strange: Bits, Pieces, Features & NeuronsWith or Without you - Dependency between Concrete and Abstract ObjectsNeural Networks are Abstract Objects.PropertiesIt’s complicated – about a RelationshipAbout Wild Features and How to Capture ThemThings, Facts, and the Ontology of Neural NetworksWhat About Aesthetics, Agency, and Authorship?Aesthetics of Neural ArchitectureThe Sensibility of Neural ArchitectureAgency in Neural ArchitectureNeural Architecture is a New ParadigmReferences48 Chapter II – The Robot GardenThe Robot GardenHow to Test a RobotPosthuman Design is here.Big Data, AI and Architecture DesignReferences62 Chapter III – In the Valley of the Hallucinating Machines.Computational Vision as Design MethodWhat Would Turing Do?Architecture’s Empathy Towards Images and RepresentationHow to Recognize a Gothic ColumnThe Nature of Neural NetworksLearning Architectural FeaturesFountains, Figures, and Features – or How to Confuse an AIWhat it Means to Be a PixelMachines Hallucinating ArchitectureReferences84 Chapter IV – Not a Question of Style—Style, Artificial Intelligence and ArchitectureA Closer Look into the Suspicious Noun StyleWhen Aesthetics Collide with TechnologyReferences110 Chapter V – Talking ArchitectureTalking ArchitectureMaterials and MethodsInitial ModelingAttentional Generative Adversarial NetworkThe Urban Context of the DesignSigns, Scripts, and Codes—A Theory of the ArtificialReferences130 Chapter VI – How Machines Learn to Plan—A Critical Interrogation of MachineVision Techniques in ArchitectureHow machines learn to seeA Posthuman Trajectory for Plan FormationEstranged - but in a good wayNeural Networks and Learning the 2-D Visual WorldModeling the Style of the Real WorldThe Defamiliarization of the City, or An Alternative UtopiaReferences156 Chapter VII – Space: The Final Frontier (How to Wrangle a Neural Network to Deal with 3D Models)Terms of Engagement: Aesthetics, Agency, Sensibility, and Other Nasty ProblemsAestheticsSensibilityAgencyAuthorshipDatabase ConstructionNeural Optimization FrameworkExperiments and ResultsSo, Can a Neural Network Learn a Sensibility?References