Statistical Pop Art?

One of the problems I think about far more than I ought to is the great river of images being created today. Suppose that one wanted to get a sense of what kinds of photographs are being made today by the masses?

You could look at flickr's Explore to get a sampling, and there are no doubt many many other ways to get a sampling of images being created on a day to day basis. However, as of this writing, flickr gets something over 1,000,000 photographs uploaded every day, and it's not the biggest. How many photographs does one need to look at to be sure that one is getting even a vague notion of what people are actually doing? If I looked at 1000 images a day, how likely is it that I could miss a smallish community of people doing something really interesting? A productive group of artists could easily produce 1000 images a year, and nonetheless be utterly lost in the noise.

It's certainly not necessary to be plugged in to any current zeitgeist, to produce art. However, if you do want to be so plugged in, and you do want your art to be informed by what the mass of imagery is doing, you have a definite problem on your hands. To grasp the current zeitgeist isn't an unreasonable desire, but it does seem to be an intractable problem.

You can simply dip into the river and see what you can pull out. I think it might be interesting to apply technology, however. There are ways to approximately classify images as similar, either by algorithmic examination of the picture itself, or through metadata like tags and surrounding text, or by applying social connectivity graphs (making the assumption that groups of friends are probably taking similar photographs). I think it would be interesting to use tools of this sort to classify and collect, roughly and approximately, large groupings of images.

By creating groupings of, in some sense, similar or related images, you could divide the river of imagery into stuff that has been so grouped, and stuff that has not been so grouped. Sampling a grouping containing a few million images might well provide a more artistically complete grasp of the group than sampling purely at random. Understanding the groups might give you a more thorough grasp of the snapshot zeitgeist than simply digging through flickr's Explore.

Then you've got the unclassifiable stuff, the stuff that you haven't been able to lump in with other stuff. This might be richer ore, in a way. This will surely be where the odd communities that are friends with nobody are experimenting in new ways. This might be where the unusual subjects and methods that nobody likes are in play. This might be where the really interesting stuff is to be found. There's probably still an utterly intractable number of images in here, but random dipping in here might dredge up more interesting work, on average.