A zoom in on our vertiginous present, the Forum Vertigo offers a series of international encounters between scientists and artists, engineers and intellectuals. After three editions devoted to the artistic uses of 3D, code, and living matter, this year's Forum Vertigo examines the multiple links between artificial intelligence and contemporary creation, on the occasion of the exhibition "Neurones, les intelligences simulés" (lit. Neurons, simulated intelligence).
Hugues Vinet, Director of Innovation at IRCAM and Forum Vertigo coordinator, explains how artificial intelligence has led to a revolution in artistic practices and their relationships with technology.
Hugues Vinet, the 4th edition of the Vertigo Forum is dedicated to artificial intelligence. Can you tell us more about what is involved in these meetings?
Artificial intelligence technologies have made spectacular advances in recent years with deep neural networks. These networks represent a radical paradigmatic break in the processing of digital information between the usual approaches by modelling and those by analyzing massive data, surpassing the latter in several fields. We are just beginning to glimpse their potential for artistic creation: processes for generating and hybridizing totally new content, producing artifacts with autonomous behaviors, etc. Beyond the renewal of forms and materials for creation, the deployment of artificial intelligence techniques on a large scale in society—personal data management, face and object recognition, autonomous behavior— also induce critical positioning in the artistic field. These meetings bring together some of the most advanced artists and researchers on these subjects to present the state of the art of their work and discuss the issues at stake in this ongoing revolution.
Will computers be able to become creative and succeed in replacing artists, or is it pure fantasy?
We are a long way from that, and a possible method for the distinction between the possibilities offered by a machine and a human being lies, for me, between the notions of generativity and creativity: there is indeed a substantial increase in the generative power of machines in proportion with the scale of the learning corpus used, but in the end they remain configured by humans. The challenge for artists, as it has always been in the history of technology, is to appropriate these radical new methods in the service of their expression: creating a work of art is a creative act that cannot be reduced to the machine.
A more forward-looking question to conclude: what are the most promising prospects for scientific research in AI and especially for the arts?
Returning to the comparison between modelling and data analysis that I mentioned at the beginning, today we still have a limited understanding of the representations of the world that are specific to deep networks: they are black boxes that produce unforeseen artefacts without really being able to assimilate their logic into our cognitive categories. Better understanding these representations, formalizing theories and making them accessible to non-computer scientists, in particular to artists, is an exciting scientific and technological challenge.