Art(ificial) Intelligence

10.12.2021 0 Like 0 kommenttia

Advances in artificial intelligence techniques in the last decade allowed artists to create new  approaches and paradigms in generative art. Instead of using symbolic programming to try to describe a musical style or a graphical quality, machine learning techniques allow creators to “teach” neural networks how to generate new works based on a particular model – a written style, a musical genre or even the pieces of renowned painters. This ability opened new pathways towards experimental setups in all creative fields, and indeed it has been embraced by artists working in the most different supports. 

In the first part of Kuva Research Days seminar on the 15th of December, we will see the works of three visual artists: Berlin-based Sofia Crespo, local Helsinki talent Jukka Hautamïki and the collective Taller Estampa, from Barcelona. Even if they all resort to the same type of machine learning training techniques, they managed to create very particular individual styles, based on the different approaches to AI and the input datasets used.

But the popularization of AI-generated content also gives room to concerns. Writers or musicians might fear losing their purpose as they could be replaced by programs. Yet under any perspective, this risk is negligible at least for the foreseeable future:  even with the best AI models, computers are still merely highly skilled copycat machines, that always depend both on a huge amount of previously human-created content to be trained, and on a careful curatorship of its output. It seems like AI applied to arts also suffers from what has been called the AI effect: when scientists manage to create machines that are able to perform a new task that until that moment was only feasible by humans, they realize the biological intelligence is much more than obtaining skills like mastering chess or the style of Rembrandt. Hence, the artificial general intelligence goalpost is moved a bit further.

Art practice seems particularly challenging for machines. Artists are always walking the fine line between creating masterpieces within certain constraints (like a musical scale) and breaking these rules to propose new configurations. Machines, however, are doomed to follow whatever instructions we programmed them with. Would Art convey a particular type of intelligence unobtainable by machines? In the afternoon, we will hear three different approaches to the limits of AI. Bruno Caldas Vianna, PhD candidate at Kuva, will talk about the hurdles to disobedience in programs. Jordí Vallverdu, professor at the Universitat Autónoma de Barcelona, will ask if a machine can deal with ethics, and finally Ninon Devis and Philippe Esling, from IRCAM_Paris will talk about Artificial Creativity.

Bruno Caldas

The program for the 15 December, KuvA Research Days

KuvA Research Days website