Panic Breeder (2019) - Samir Bhowmik Jukka Hautamki

VIDEO DESCRIPTION:

Play, download and edit the free video Panic Breeder (2019) - Samir Bhowmik Jukka Hautamki.

Panic Breeder (2019)
Samir Bhowmik and Jukka Hautamäki

Panic Breeder consists of an AI-generated video animation and a sculpture - Machina Baltica (the Beautiful Mind of), that together comment on the impacts of Machine Learning and Artificial Intelligence to the Visual Arts in the European context as well as the environmental damage to the Baltic Sea.

With more than 85 million people living in its catchment area, the Baltic Sea has been exposed to an extensive environmental degradation since the beginning of the industrialization of the region in the late 19th century. It is often regarded as the most polluted sea in the world. This condition has negatively affected its biodiversity by directly endangering animals and plants or by damaging their habitats.

Thus, building on a HELCOM (Helsinki Commission [Baltic Marine Environmental Protection Commission]) Red List of endangered (threatened and declining) species, Panic Breeder presents a quasi-nature documentary of imaginary fantastical species as the future inhabitants of the Baltic Sea. It takes inspiration from An Evolution of Species, theorized by Charles Darwin, and comments on the iterative and adversarial capabilities of AI to probe the surreal and terrifying future of “survival of the fittest”.

By quantifying the energy consumption of machine learning that goes into such suspect desires, from among the machine-generated breeding, a singular creature is manifested in physical form. In this fictional scenario, the loss of species is counter-balanced by the possibilities of AI-generated creatures to act as a temporary, if not futile, antidote to climate change panic and anxiety.

Panic Breeder was exhibited in the Aalto Artificial Intelligence Exhibition, from 26.11.2019 - 15.01.2020 at Dipoli, Aalto University, organized by the Finnish Center for Artificial Intelligence (FCAI) and Aalto Digi Platform. The exhibition received over 1000 visitors.

Software
The artwork uses GAN (Generative Adversarial Network), a class of machine learning systems, where two neural networks (Generative and Discriminative) compete with each other for several training sessions. The trainings utilize a custom dataset of extinct and endangered birds to generate new images. The animation created from the new images presents the transformation and evolution of the species. The artwork is subtitled with a running commentary, with other numerical metrics such as energy consumption and carbon emissions displayed on the side.

Hardware
Triton Cloud Computing Cluster, Aalto University. Triton is part of Finnish Grid and Cloud Infrastructure.
Our installation involved using 224 hours of machine training in 9 NVIDIA GPUs using Triton.
This resulted in 225 kilo-watt-hours or 159 kilos of CO2 (calculations based on various scientific papers).

Additional Commentary
Panic Breeder changes the way we look at Machine Learning and Artificial Intelligence in relation to the Visual Arts. It raises several questions and dilemmas attached to the use of machine learning in the making of contemporary art, the role of the artist, aesthetics, as well as the environmental impacts.

Who is the artist? Us or the neural network? In this work, the artist’s role was relegated to being simply a curator of images, a preparer of datasets. In fact, the artist’s tasks were merely to facilitate the process by which machine learning could take place. As the Machines spit out data in the thousands, the artist has to go through thousands of images, having to make aesthetic decisions of choosing the best images.

Where is this aesthetic judgement coming from? Art history, cultural background, social bias, beauty, hate, anger? Not only the AI-generated images that we might call art, are deceptive in a sense that they are almost impressionist (according to visitors) they also make one believe the work is of a human. Although the network does not contain an ”impressionism” code.

What are the environmental impacts of machine learning? Running GPUs for machine learning over nights and days, for weeks to create an artwork has a significant energy footprint, not to say of greenhouse gas emissions. The making of Panic Breeder resulted in 225 kilo-watt-hours or 159 kilos of CO2.

Finally, what ethical implications has this work that revives endangered and extinct species of the Baltic Sea? How would one balance the desire to re-create extinct species versus the energy and environmental costs of AI and machine learning? The work thus challenges us to examine the gradual destruction of the Baltic Sea, the growing number of endangered and extinct species of birds versus the frivolous desire to re-create them through artificial intelligence-generated imagery.

Download, play and edit free videos and free audios from Panic Breeder (2019) - Samir Bhowmik Jukka Hautamki using RedcoolMedia.net web apps