AI APPLICATIONS
cars, or animals. But these networks were hungry for data. The modern era of computer vision was truly started by the creation of ImageNet, a colossal dataset of over 14 million labelled images. As Dr. Fei-Fei Li, the visionary behind ImageNet and Professor at Stanford University, famously said:“ I came to believe that if we wanted to build a machine that could think, we needed to teach it to see.”
The applications powered by this sight are already driving huge efficiencies. In healthcare, algorithms now screen for diabetic retinopathy with an accuracy matching human ophthalmologists, providing critical care in underserved regions. On factory floors, cameras monitor assembly lines for microscopic defects and perform predictive maintenance by spotting wear and tear on machinery before it fails. The technology is even being used for environmental monitoring, with satellite imagery analysis tracking deforestation, plastic waste in oceans and illegal mining operations in near real-time.
Yet this rapid adoption isn’ t without friction. Debates around privacy, consent and algorithmic bias are more fierce than ever, facial recognition in public spaces raises significant ethical questions and biases in training data can lead to
96 November 2025