CONTINGENT SYSTEMS: ART AND/AS ALGORITHMIC CRITIQUE IS A BI-WEEKLY PANEL SERIES THAT EXPLORES CRITICAL INTERSECTIONS BETWEEN CREATIVE PRACTICE AND ALGORITHMIC CULTURE. EACH PANEL BRINGS TOGETHER INTERDISCIPLINARY ARTISTS AND SCHOLARS TO EXAMINE IN GREATER DETAIL A THEME ALIGNED WITH THE CORRESPONDING EXHIBITION. PAPER PRESENTATIONS WILL BE FOLLOWED BY A MODERATED DISCUSSION PERIOD AND Q&A.
OCTOBER 15TH | MACHINE LEARNING
Unnaturally Speaking: Exploring Algorithmic Bias in Speech Technologies
Halcyon Lawrence, Towson University
Speech technologies are routinely hailed as “revolutionary” because they successfully recognize and replicate natural language in a growing number of domestic and industry applications. However, for speakers with a non-standard accent, a non-native accent, or a dialect, virtual assistants are often unresponsive. In fact, for these speakers, speech technologies are disciplinary in nature, and do nothing to change existing “social biases and hierarchies,” but instead reinforce them. Therefore, while they might be useful to some, they are in no way revolutionary. The need to address the lack of representation of accent varieties in speech technologies is imperative as voice is increasingly becoming a medium of surveillance and policing (in prisons, schools, banking, etc.) of our most vulnerable members of society, even as marginalized voices are minimally represented in speech training corpora. In this talk, I argue that to move beyond token, gimmicky, celebrity voice representations, there is a need to understand the socio-economic context in which speech technologies are developed and the long history of assimilation—rooted in imperialist, dominant-class ideologies—those non-native and non-standard speakers of English have had to practice in order to participate in economic and social systems. These assimilative practices result in the sustained marginalization and de-legitimization of non-standard and foreign-accented speakers of the English language. I examine a few projects that challenge existing industry practices around speech and language technologies.
Language in the Age of Algorithmic Reproduction: A Critique of Linguistic Capitalism
Pip Thornton, University of Edinburgh
In an age of digital technology, language has become far more than a means of human communication, creativity or expression. Increasingly written for – and ‘read’ by – algorithms, when words become data, they carry more than linguistic meaning, and as such are valuable commodities in the advertising marketplace. Nobody knows this better than Google, which made its fortune from the auctioning of words through AdWords; a form of ‘linguistic capitalism’ (Kaplan, 2014) in which the aesthetic value of language is negated at the expense of its exchange value. But what are the residual cultural or political effects of this algorithmic exploitation of language? What is gained and what is lost when words become data? I argue that the liquidation of language into data, and the speed in which it can be processed, reproduced, interpreted and capitalised, has consequences that are both linguistic and political. In addition to the critique, I would also like to offer a means of resistance against the algorithmic acceleration of linguistic discourse in the form of an artistic intervention. In order to make visible – and interrupt – the workings of linguistic capitalism I have developed a research/art project called {poem}.py which uses poetry, the Google AdWords keyword planner and a second hand receipt printer in an attempt to rescue language from the algorithmic marketplace; re-politicise it (Benjamin, 1936), and reclaim it for art. This talk explains the genesis of the project, including demonstrations of the different modes of intervention I am currently exploring.
Computational Haze
Zhenzhen Qi, Columbia University
Today’s digital interactive medium is faced with an imperative – the negotiation between functionality and openness. Digital communication is facilitated by computers. Once designed, a computational device has a finite amount of total computational power that can only afford a finite number of signals to be transmitted in a given amount of time, which is free from the degradation of noise. Therefore, we are faced with a dilemma, precision at the expense of variety. When we build an electronic scale, we are concerned with the scale precisely calculating and outputting the amount of weight it is measured. However, when we build a musical instrument, we desire a variety of outcomes, with some room for imprecision and unpredictability. We will describe an electronic scale as informative, but we will never say it is poetic, energetic, or creative. Machines that have a high level of control of information over their surrounding environment offer a low level of choice and possibilities. Machines that offer variety and chance offer an ineffective rendering of the single immediate reality at hand. If the history of the internet is seen as a process of balancing the informational and energetic outflow of the computational device, then this history has overwhelmingly favoured efficiency at the expense of openness – a necessary condition of the creative act.
The machine can only highlight the quality of an object, so long as that quality can be represented and communicated via the predefined protocol. Digital interactivity is designed in a way that an object’s value can only be communicated via a pair-wise interaction. Interactivity implies there is a sender and receiver, and the sender can only make its existence felt upon by sending a series of electronic signals which can be perfectly received and comprehended by the receiver. When we operate a machine or program software, our communication with it is bounded by the same predefined communication protocol.
There is an urgent need to reimagine a radically different mode of communication through digital interactive media, so that interaction enabled through computational object have the capacity to be both absent and present. This dual existence will create a condition for communication that is non-reduced and will force the self-referential way of communication to ascend to a higher level. What we need is a digital haze, that allows one’s digital presence to be felt upon, approximated but not exposed or approached.
BIO | zzyw is a research collective formed by Yang Wang and Zhenzhen Qi in New York, 2017. It produces software applications, simulations, and text as instruments to examine the cultural and educational imprints of computation. They were recently technology resident artists at Pioneer Works, Center for Art and Innovation, and members of NEW INC, the culture and technology incubator led by the New Museum. In addition, Zhenzhen Qi is an educator, researcher, mathematician, and technologist based in Brooklyn, New York. She is a candidate for Doctor of Education(EdD) at Teachers College, Columbia University, currently researching artistic interventions to unloop the software effect. Yang Wang is a computational media artist, designer, and software developer based in Brooklyn, New York. He works as a creative technologist at architecture firm Rockwell Group.
Panel organized by Crystal Chokshi & Dr. Mél Hogan (Environmental Media Lab, University of Calgary)
FEATURE IMAGE: ZZYW, LENNA, 2018 — Image Credit: Carolina Vasquez-Lazo
