Paolo Marocco

My path crosses — in a non-linear way — mathematics, statistics, drawing, and film criticism. In the twentieth century it was not easy to frame them within a single conceptual border. Today, thanks to a kind of homogeneous envelope of AI, I think it is easier, or rather more easily communicable. A carousel of practices that, beyond the technical details, resonates in a Galilean sense as the place where mathematics (in the broad sense) has finally found a creative harbour, moving beyond the patterns of numbers and formulas.

I hold a degree in mathematics and a PhD in economics. To be clear: I am not an economist — I only know something of the mathematics and statistics that underlie economics. Beyond those variants, I spent 20–30 years as a developer of applied processes — in this millennium, Machine Learning and then Deep Learning pipelines. I include my LinkedIn page here, though it is very general and not updated on the work presented on this site.

As for cinema, I have written around a hundred articles of film criticism for various print magazines and newspapers. I also wrote two books: a monograph on Eric Rohmer and one on Hitchcock's Vertigo. Two filmmakers who, in different ways, always closely combined creativity with mathematics. I abandoned that activity about 20 years ago. But the film market background continues the attraction of my interests, and I would like to try writing some posts on the subject here.

I would like to use this site as a platform for writing technical and educational posts, with a particularly didactic character, and some reflections on the evolution and market of the models I work with here (in particular generative image models). There are thousands of them, of course. But I believe that, since the introduction of vibe coding, there are increasingly more professionals interested in crossing the boundary from graphics into code development. One current approach is to integrate the traditional workflows of graphical environments with the automation of modular node graphs (as in environments like ComfyUI, to which I will dedicate some posts). I dwell on this point because I am fairly critical of the easy communication around the infinite possibilities of AI through pre-packaged tools that, as if by magic, generate AI reels and high-quality shots with two prompts and three clicks. There will probably be a convergence toward classes of easily generated images and video, but I do not believe in a clear improvement in experiential quality by relying on this approach alone.

A small section of the posts would also like to address that sector, not so spread in Italy, known as film industry analysis. At least for now, my aim is not a methodological data-mining and prediction exercise, but an intuitive approach to the socio-cultural reasons that can turn a low-budget film by a little-known director into a revenue multiplier of 20–50 times its production costs. In short, to investigate the insights of a work in order to plunge into the unpredictability of revenues (a master of this science is Stephen Follow).