The evolution of AI in city planning

Abstract Artificial Intelligence (AI) is entirely coincident with the emergence of the digital computer. It was assumed from the start, some 75 years ago, that the computer had more than the required power to simulate human intelligence. This led to the speculation that ultimately computers would take over many of our own tasks which we once considered could never be modelled using machines. Here, we sketch the history and evolution of AI, note the different phases in this history, define distinctions between strong and weak AI, and emphasise the difference between generative and discriminative processes. There are many possible applications in city planning with the most suggestive and possibly the most disruptive being those related to the development of new methods for generating sustainable plans and designs. We make a key distinction between inductive and deductive AI, demonstrating these differences with methods of machine learning (ML), showing how early network methods based on the perceptron, can be linked to deductive procedures that enable us to build new models for city design. Our key illustration links urban simulation models to land cover built around geospatial data infused with ML. The aim of this paper is to chart the progress in AI and its applicability to city science and city planning from its first applications and speculate on future developments.