Last Friday I attended a talk featuring Max Matthews and a new conceptual work by local artists as the Gray Area Foundation for the Arts (GAFFTA). GAFFTA is an intriguing new organization and space for the intersection art, design, sound, and technology. They are “dedicated to building social consciousness through digital culture.”
I had last seen Max Mathews, considered by many to be the “father” of computer music, at an 80th birthday tribute at the Computer History Museum in 2007, and before that delivering the keynote address at ICMC 2006. It is great to see him still going strong and engaged with technology and supporting others’ creative work. His talk primarily focused on his work at Bell Labs, and in particular the history and technologies surrounding his 1962 computer rendition of the song “Daisy Bell” (aka “Bicycle Built for Two”). It was an early example of physical modeling synthesis, where sections of the vocal tract were modeled as tubes, and sound generated directly from physics equations. His version of the song was popularized in the film 2001: A Space Odyssey, although Mathews revealed that the version Kubrick used in the film was not his recording. He also presented another famous example of computer-generated vocals, a performance of the “Queen of the Night” aria from Mozart’s The Magic Flute. This piece used formant synthesis in which focus on recreating spectral characteristics of the sounds (i.e., the formants that characterize vowel sounds) without necessarily modeling the physical processes that allow humans to create those sounds. The voice is quite compelling (if a bit dated), and demonstrates that the most realistic sounds are not necessarily those generated from physical models.
[Mathews, Koblin and Massey. Photo by Vlad Spears.]
Mathews’ presentation of “Daisy Bell” served as an introduction for a new project “Bicycle Built for 2000” by Aaron Koblin and Daniel Massey. Koblin has been working on a series of conceptual pieces that utilize the Amazon Mechanical Turk, a framework for harnessing human intelligence to solve large problems. There are some things that humans are quite efficient at and computers are very poor at, such as recognizing distorted text (think of the CAPCHA codes that we all deal with on websites, including here at CatSynth). The Amazon Mechanical Turk, which derives its name from an 19th century hoax where a supposedly mechanical chess-playing machine turned out to be a human hidden inside a box, provides a framework and API for defining tasks to be solved by humans, recruiting people to work on them, and then compensating them for their efforts according to a fixed budget. One of Koblin’s pieces provides the instruction to “draw a sheep” for a few cents. He then collected the resulting drawings of various sheep (and non-sheep) from around the world and compiled them into a larger mosaic work, The Sheep Market. You can see the overall mosaic as well as click on individual sheep and even see an animation of how they were drawn by the individual contributors.
[Screenshot of Bicycle Built for 2000, by Aaron Koblin and Daniel Massey. Click to enlarge.]
For “Bicycle Build for 2000”, Koblin collaborated with Daniel Massey on a work that focused on sound and music. The Max Matthews rendition of Daisy Bell was decomposed in a sound segments no longer than one syllable, and for each segment an Amazon Mechanical Turk task was created for someone to sing back the sound. Koblin and Massey then reassembled the sounds to re-create the song as sung by the participants. The initial version, in which only one singer was used for each component, was almost unrecognizable, though quite interesting. A larger version which included a chorus of voices on each segment better represented the original qualities of the song, and clearly recognizable. You can hear the full version at the project website, along with an interesting visualization of the assembled recordings. The fact that the larger ensemble produced a more recognizable result, essentially averaging out the various sung renditions, is perhaps an example of the oft touted “wisdom of crowds”.
There is also a slightly more ominous set of questions from these works concerning exploitation of individuals who don’t know the overall purpose of their tasks, or about hive mentality and such, but I still find it quite interesting and am inspired to try out the Mechanical Turk for a future art project.