Tuesday, September 5, 2006

The Networked Image

I first picked up on Google's Image Labeler two days ago (via). In a nutshell, Image Labeler addresses a semiotic problem: the indexing of hundreds of thousands of images based on semantic assignments in the visual field of each image. Indexing an image depends upon the assignment of keywords that correspond to the objects represented. Google Image Labeler makes this process into a game of peer review: in this two person game, a player win points by registering a descriptor that also appears on the other person's list.

Tracing a few links (succumbing, that is, to the beckoning of a surprising curiosity), I briefly started to follow the life of this conversation in computer science and art. Most intriguing in this regard was the talk embedded below, a talk called "Human Computation" given by Luis von Ahn at Carnegie Mellon.

I find von Ahn's talk fascinating on several levels. He explains gaming (i.e., "Games with a Purpose") as a solution to the labor-production dilemma of developing a gargantuan repository of indexed images, of reconciling the gap, using the most basic set of terms, between image and word. Implicitly, he sets up a way of thinking about "writing the image" as collaboration, as writing that connects (in the allure of consensus, agreeing, that is, on the equivalence of Bush's photo and "yuck" (around 21:00 in the video)) and produces. In keeping with the title of his talk, he explains human computation--"Running a computation in people's brains instead of silicon processors."--premised upon "anonymous intimacy," the pleasure of coming to terms with strangers about the verbal evoked in the encounter with the visual. He also refers to his impressive research projects The ESP Game and PeekaBoom, predecessors, it would seem, to Google's Image Labeler.

Is it going too far to invoke Barthes here? If not the studium, exactly, there is something studium-like in Google's Image Labeler. The image-index undertaking is a project akin to establishing studium as a database. The collaboration between a labeler and a validator (partners in the game) devalues the intense singularity and instead reduces the image to human-generated language agreements, its lexical mutuality. There is a networked quality to the image in the way it is treatment here. Enigmatic thinking won't win in this game. The image is domesticated by the process and submitted into the most generic realm of culture, which, as Barthes puts it, "is a contract arrived at between creators and consumers" (28). It's useful for image searching, but is also has implications for habituated seeing and collaborative image work.

Posted by Derek Mueller at September 5, 2006 12:50 PM to Networks