Conferenza "Il binomio cervello-mente non scatta foto"

The technological wave of artificial intelligence has now permeated every corner of our society: from education to healthcare, from work to the economy, and much more. Even the world of contemporary photography has not remained immune to this revolution, surpassing simple mechanical and electronic automation.

Today, cameras do not merely passively capture light; they aspire to replicate the perceptual, neurobiological, and cognitive processes of the human brain. Thanks to the architecture of neural networks and modern imaging processors equipped with neural processing units (NPU), systems have been developed that reflect the function and structure of the visual cortex in living beings.

The goal is to explore the multiple similarities between how humans perceive the world, through a multi-level decoding and interpretation system that begins from the primary visual cortex V1 and extends to neuronal processing in other areas of the brain, and the new algorithms employed in artificial vision. Indeed, initial visual processing, which primarily occurs at the retinal level, allows a deep neural network to analyze a continuous flow of raw sensory input. Structurally, modern sensors approach the way the retina collects information. Just as the capture and processing of primary data, such as edge orientation, spatial frequency, and chromatic contrast, occurs in simple and complex cells of the striate cortex, the deep neural network of the photographic processor also combines these fundamental characteristics to create a high-level conceptual representation. Thanks to artificial intelligence, smart cameras are now able to "understand" what they are capturing, simulating the way the human brain uses selective attention and neural prediction.

The smart camera has the remarkable ability to recognize the order of objects, creating a sort of hierarchy among them. It can predict how they will behave in the future, thanks to what we call optical flow, and it can also fill in the gaps in the image, based on probability patterns learned over time. This fusion of computational neuroscience and engineering not only radically changes the hardware possibilities of imaging systems but also sparks a lively debate on the interaction between artificial perception and aesthetic cognition. In fact, the computational device now processes information in a way that closely resembles the functioning of a biological system rather than that of a simple optical machine.