Biography

Leonardo Impett read Engineering at St. John’s College, Cambridge, with a master’s thesis on the use of machine learning in experimental violin acoustics. He then joined the RAINBOW group at the Cambridge Computer Lab under Alan Blackwell. After briefly studying nanotechnology at the University of Tokyo, he worked on image aesthetics for Microsoft Research in Cairo, and built electronic instruments - including the ‘mephistophone’ in Matthew Herbert’s Faust at the Royal Opera House. He is finishing a PhD in digital humanities at the EPFL under Sabine Süsstrunk and Franco Moretti, focusing on gesture in art. He joined the Biblioteca Hertziana in 2018. He is a Fellow of the Royal Society of the Arts and a Member of the IET.

Project Summary

In the last half-decade, convolutional neural networks have radically transformed the ways in which computers are able to understand images, including the precise deduction of human poses and hand gestures. This project attempts to transfer these techniques to art history and historical anthropology, exploring live and depicted gestural practices in fourteenth- and fifteenth-century Italy. Building on his earlier work on Aby Warburg and the Pathosformel, Leonardo’s project traces micro-gestural movements and temporalities in large collections of Christian iconography. He is also interested in the epistemic role of such computational techniques, currently treated largely as empirical forensics or as tools for reconstruction and publishing. These views ignore the technician’s perspective: that programming is largely a craft practice rather than an experimental apparatus, and thus that computational criticism is a form of practice-based research, perhaps in the vein of experimental archaeology or artistic research. The computational avoids the need for the mediation of images through language. Current scholarship largely frames this in terms of serendipity, such as the ineffable similarities returned by an image-search algorithm. This project proposes a Brechtian theoretical framework for such image-machines; their usefulness is precisely a product of the distance between human and machine vision.