SCMS 2019 Panel— Neural Media: On Neural Networks and New Data Practices

deadline for submissions: 
August 10, 2018
full name / name of organization: 
Ranjodh Singh Dhaliwal (University of California, Davis) and Théo Lepage-Richer (Brown University)
contact email: 

Neural Media: On Neural Networks and New Data Practices

A Proposed Panel for the 2019 conference of the Society for Cinema and Media Studies

(March 13–17, 2019, Seattle)

 

After decades at the fringes of computer science, neural networks are now recognized as one of the most effective architectures for digital systems to develop generalized representations of the data that are given to them. These networks ‘learn’ by establishing probabilistically weighted connections among specialized ‘neural’ units that mirror the key statistical dependencies that can be identified in training datasets. From Google Home and Amazon Alexa to self-driving cars and search engines, neural networks are today ascendant in most, if not all, machine learning frameworks and ‘artificially intelligent’ systems. On the one hand, then, neural networks enable some of the defining functions of contemporary digital infrastructures, directly informing how the institutions using these techniques perceive, classify, and operate upon the objects and subjects they engage with - concerns that have been central to the recent work of media scholars such as Wendy Chun, Orit Halpern, and Adrian Mackenzie. On the other hand, neural networks are also key components of the popular cultural imaginaries around AI and inspire many aesthetic explorations in the field of media art, as recently discussed by the likes of Blaise Aguera y Arcas and Matteo Pasquinelli.

 

This proposed panel welcomes scholars working at the intersection of media studies, science and technology studies, history of science, art and visual studies, algorithm studies, and/or critical data studies to kickstart a discussion around the epistemic conditions entangled in the development, design, training, and/or implementation of neural networks and, more broadly, contemporary machine learning or AI techniques. What type of epistemic assumptions were necessary in the fields of cognitive science, mathematics, and computer science for neural networks to become the convenient analogy for cognition and perception that it is today? What kind of schemes of knowledge and expectations are embedded in such visions of the world? How can neural networks’ key applications (computer vision, predictive analytics, generative modelling, etc.) illuminate the type of representations these media technologies acquire? What does the prevalence of such technical forms mean for today’s visual culture, which increasingly borrows from these technological implementations (e.g., Google Deep Dream, Dota 2’s first professional AI team OpenAI Five, the work of media artists such as Trevor Paglen, Constant Dullaart, and Terence Broad)? How are data captured as objects of knowledge by contemporary media technologies? How do neural networks transform and alter the broader models of surveillance, capture, and control (biometrics, data collection, predictive modelling, etc.) in which they operate? How can art practices and resistance tactics, such as adversarial attacks, uncover counter-narratives and -histories about neural networks?

 

If you are interested, please send a 150-300 word abstract along with a short bio to Ranjodh Singh Dhaliwal and Théo Lepage-Richer (rjdhaliwal@ucdavis.edu, tlricher@brown.edu) by August 10th, 2018.