Thumble
Thumble screenshot
This page is still under construction. In the meantime, see the latest screenshots... a simple mock-up version of the application: Thumble Mock-Up (click on the thumbnail)... and in its final form as a Google gadget: Thumble early beta (click "Add it now" — please note that the gadget is being developed on Firefox 3 and Safari 3 for Mac, and has not yet been tested on other browsers and platforms).
Abstract
Consider the apparently simple notion of a random image. At first one would assume that, by definition, the very last question about it one should be able to answer with any confidence is: "What do you expect it to look like?" Yet, when we further consider the practicable ways of generating random images, one realises that we all know full well what a two dimensional array of uncorrelated pixel values would be like. In fact we would simply say: "It'll be just like TV snow". Evidently, for (nearly) all intents and purposes, random distribution of colour values across a surface does not equate to an aesthetically relevant notion of a random image.
Consider the following question instead: "What would you expect an image to look like if pooled at random from the Google's image directory?" Clearly, one's confidence in answering this question in turn would drop right down: it could be any one of the many kinds of aesthetically significantly distinct images that populate web pages. Sure, images on the web in general and those referenced by Google in particular do not include all conceivable aesthetic types nor are the featured types equally represented. Yet, any such statistic that may be harvested is meaningful and of great interest to researchers and artists, which is why there are already quite a number of online applications that allow users to browse image directories, such as Google's, at (allegedly) random.
Thumble screenshot
Thumble is a Google gadget (i.e. a small application that sits on the user’s iGoogle homepage; most likely among other gadgets and RSS feeds) that allows cumulative traversal of the space of search terms and therefore something like “creative browsing” of the Google’s image directory. Thumble’s initial screen displays a search string, or “cue”, generated by adding terms pooled from an English dictionary — with the probability distribution reflecting their relative frequency of use — until Google matches less than a thousand images (this allows Thumble to access all the image results). The cue, therefore, may sometimes be two, sometimes ten words long. Above the cue is displayed a thumbnail image, or “nail” indexing one of the images that matched the cue, say, image result 278 out of 789 found. The nail and the cue (thumbnail and the query string) together make up a “thumble” and are arranged as a figure and a caption.
Although (quasi) randomness features at all stages of the process, the nail and the cue may be related in any of a number of ways that Google uses as matching criteria. The Thumble user is thus presented with a perceptual conundrum pregnant with meaning. These initialised thumbles may be variously attention grabbing, as chance would have it, but the user has a series of functionalities at her disposal to add to, remove from, or change the nails and/or the cue. In this way, the user quickly homes in on thumbles that feel increasingly her own. Eventually, she may choose to save a thumble so that it appears as one in the list of cues that index all saved thumbles any of which can be displayed at will. In future, moreover, she will also be able to share the thumble with other Thumble users.