Dialogic Machines is an ongoing research project in which an Artificial Intelligent system (GAN) is trained as an analytical tool to understand on an objective level the nuanced ways in which we construct, draw, and represent architectural design. The pair of AI models, left and right, are in a state of continuous conversation, allowing them to both propose and respond to what they believe an architectural drawing might be – having been trained on a dataset of over 8000+ unique orthographic architectural drawings. Through this generative conversation and response to each other’s interpretations the machines begin to find differences and commonalities that they believe objectively are worthy of representation and discussion. These images rarely find common ground as fully resolved architectural drawings to the trained eye, instead allowing the machines to become the arbiters of their own understanding of architectural drawing, unbiased by a predetermined representational ideal of the architectural author. As the machines flicker between these abstract images an audio interpretive model begins to signal binary changes as to how they believe their processes are converging and diverging, signalling to the viewer an interpretation of their discourse.