Urban artificial intelligence is becoming an increasingly crucial aspect of urban governance. This is a trend that can be observed at various scales: from predictive algorithms and self-driving cars to entire cities managed by digital brains. Urban artificial intelligence is firmly rooted in the present and past of the city, particularly in the smart city model. This article illustrates the main connections and differences between the smart city and urban artificial intelligence. We argue that the application of urban artificial intelligence in the governance of cities differs considerably from the classic smart city model in three key areas: time, action and power. This is giving rise to new challenges, including ontological ones. By combining geography and philosophy, our interdisciplinary contribution aims to highlight and conceptualise a different ethics of artificial intelligence. This moves away from simple regulatory issues to focus instead on how the very essence of the city, and thus its ontology, is changing in the transition from smart cities to autonomous urban systems.