The 2012 success of the AlexNet deep neural network at the ImageNet image-recognition contest demonstrated that large neural networks trained on big datasets and GPUs could outperform older methods, triggering a wave of deep-learning research. Subsequent advances extended machine learning to prediction, language and, with large models, the generation of text and images at scale. Cities adopted these tools for traffic-signal control, video surveillance and facial recognition, demand forecasting and planning analytics. The same techniques also powered private platforms operating within urban space.
It provided the analytical engine behind smart-city sensing, automated monitoring and predictive planning. At the same time it raised serious concerns about algorithmic bias, mass surveillance and democratic accountability in how cities are governed.