Thanks everyone for a great semester and event. We will be starting up again in 2019 with open, weekly meetings on Columbia University’s campus. (You don’t need to be associated with Columbia to join.)
If you would like to receive announcements about these meetings and other future events, please sign up for our mailing list here.
A Cross-Disciplinary Reading Group
This website serves as a public record for a reading group on art and algorithms being held this fall 2018 for the Columbia community. On this page you’ll find an overview of five meetings held, as well as links to in-depth notes from our discussions.
- Meeting 1 (19 September 2018)
- Meeting 2 (3 October 2018)
- Meeting 3 (24 October 2018)
- Meeting 4 (7 Nov 2018)
- Meeting 5 (28 Nov 2018)
This event has passed.
Join us for a reflection on algorithms that generate art.
What is the difference between a set of instructions for a person and a set of instructions for a computer? Is training a neural network an artistic endeavour? How can we analyze a piece of music that can be generated infinitely? In what ways are our preferences mediated by algorithms?
Come see talks, demos, and performances inspired by a semester of debating and discussing the technical and critical issues surrounding algorithms, art, and the act of creation.
1-2pm: Finite and Infinite Art
Poet & professor of digital media Nick Montfort from MIT
Philosopher James P. Carse ends his book Finite and Infinite Games by stating “There is but one infinite game.” There are, however, many infinite works of art — actually, infinitely many. I will describe several types of finite and infinite artworks and relate these to algorithmic, computational artistic practices. There are many infinite (boundless) artworks that are not based on digital computation or even what people would usually describe as “technology,” while there are many finite (bounded) artworks that are computational. Distinguishing these helps explain how computation (as it is employed in music, visual art, and literary art) can be used to explore that which is bounded along with different infinities.
2-3pm: Rapid-fire lightning talks
- Audrey Amsellem: “You Are What You Stream” : Spotify’s Algorithms
- Katy Gero: “The best way to predict the future is to make it”
- David Watkins: Recreating Geometry Using a Limited View
- Jeanne Devautour & Samuel Boury: Random Walks Through Poetry
- Liz Bailey: Uncanny Display: Algorithmic Art at the Whitney
- Oscar Chang: Creativity in Automated Drug Discovery
- Eamonn Bell: Congratulations! You are now an algorithm!
- Elizabeth Case: 5 Minutes to Glitch Art
3-4pm: Catered reception
Winning art from the Columbia Data Science Institute Data Art Contest will be displayed.
This event is supported by:
- The Center for Science and Society at Columbia University
- The Center of Data, Media & Society at Columbia University
- The Brown Institute for Media Innovation
Meeting 1: “Algorithm”
This meeting introduces participants to various definitions of “algorithm,” from historical to contemporary mathematics and from computer science. Readings will be chosen to reflect multiple understandings of the term but also to serve as concrete examples of the diversity of modes of argument, standards of proof, notations, and intellectual priorities within the scientific and artistic communities.
- Striphas, Algorithmic Culture
- Cormen et al., The Role of Algorithms in Computing
- Karpathy, The Unreasonable Effectiveness of Recurrent Neural Networks
Additionally, we’ve selected some pre-computer artwork that is algorithmic in nature to peruse:
- James Tenney, August Harp for Susan Allen (1971) [recording]
- Sol Lewitt Wall Drawings
- Tristan Tzara, TO MAKE A DADAIST POEM (1920) [implementation]
Meeting 2: Garbage in; garbage out?
Some generative models of art require large amounts of training data (in this case, pre-existing artworks) before they can be used effectively. If this training data is of poor quality or does not accurately represent the domain being modeled, the algorithm may fail to produce convincing outputs. Readings will capture the difficulty of arriving at intersubjectively acceptable criteria for evaluating the quality of both input data and generated output.
- Elgammal et al., CAN: Creative Adversarial Networks, Generating “Art” by Learning About Styles and Deviating from Style Norms, 2017
- Wiggins and Bhattacharya, Mind the gap: an attempt to bridge computational and neuroscientific approaches to study creativity, 2014
- Bolukbasi et al., Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings, 2016
- Parrish, Poetic Sound Similarity Vectors Using Phonetic Features, 2017
- Visual Chatbot Demo
- Text-to-Image Generation Demo
- Sunspring A Sci-Fi Short Film Starring Thomas Middleditch
- Parrish, Exploring (Semantic) Space With (Literal) Robots, Talk at Eyeo 2015
- Firth, The Techniques of Semantics, 1935
- Harris, Distributional Structure, 1954
A timely article from 25 Sept, 2018!
- Schneider and Rea, Has Artificial Intelligence Given Us the Next Great Art Movement? Experts Say Slow Down, the ‘Field Is in Its Infancy’, 2018
Meeting 3: Artificial interlocutors
Artists and corporations have used artificial interlocutors in their live performances, installations, and advertising campaigns, raising interesting questions about agency and artistic identity in an age of algorithms. Conversational agents also lie at the heart of a growing number of communication systems: customer support lines, chatbots, and voice recognition interfaces. Readings in this meeting will explore what artists have learned from the creation of algorithmic collaborators and what designers of conversational agents can learn from the study of art.
- Lewis, George E. Too Many Notes: Computers, Complexity and Culture in Voyager. 2000.
- Weizenbaum, Joseph. ELIZA—a computer program for the study of natural language communication between man and machine. 1966.
- Newton, Casey. Speak, Memory: When her best friend died, she rebuilt him with artificial intelligence.
- Mccarthy, Lauren. social turkers: crowdsourced dating
And here are some performances and demos to play with:
- George Lewis “Interactive Trio” for Trombone, Two Pianos, and Interactive Music System, 2011
- Magenta music improv demo at NIPS 2016
- Browser implementation of ELIZA
- Sandy Speaks, a chatbot exploring racial injustice, and a short write-up.
Meeting 4: Everything new is old
Recognizing the iterative nature of research entails reckoning with its history of provisional and failed attempts to produce art algorithmically. Readings will examine how the resuscitation of “old science” is facilitated by the appearance of new applications and, relatedly, how classical (and potentially problematic) applications and evaluation regimes—canonical generative task framings, old datasets, and competitions—are used to legitimize new algorithmic arts.
- Interactive and generative music: a quagmire for the musical analyst by Michael Young (2016)
- Art in the Sciences of the Artificial by Kenneth Stanley (2018)
- THE COMPLEXITY OF POETIC PATTERN: Recreating Early Work in Machine Translation by Nick Montfort (2018)
And here are a bunch of demos and interactive examples related to these readings:
- Continuator: A musical instrument that answers you in your style
- DeepBach: a Steerable Model for Bach Chorales Generation
- Concrete Perl: A set of four concrete poems realized as 32-character Perl programs, by Nick Montfort
- Tracery: generate text, graphics and more
Meeting 5: Opening black boxes
Contemporary technological developments mean that decisions made by many kinds of algorithms cannot be always be simply interpreted as a chain of reasoning leading from input to output. Readings explore the concept of “interpretation” in both the statistical and hermeneutic sense, and will examine whether new algorithmic techniques ought to cause us to update our understanding of closely related concepts: causality, explanation, explainability, and even moral accountability.
- Black GUI Universe by American Artist (2018)
- Algorithmic Accountability by Nicholas Diakopoulos (2015)
- Grad-CAM: Visual Explanations by Ramprasaath R. Selvaraju et. al. (2017)
- (Or an information write-up of Grad-CAM by Kirsten Menger-Anderson)
- Zach Lieberman Interview on Open Source by Kyle McDonald (2011)
And here are some demos/short reads:
- Grad-CAM demo (slow! ~8 seconds for a response)
- Simulate world systems by Nicky Case
- Scientific Communications as Sequential Art by Bret Victor
- Playboy centerfold model used as image processing benchmark info from Wikipedia