Sunday, November 18, 2018

Whence Greatness, Again

The title here refers to a post I made some time ago which long-time readers may recall at least a little. That post is still perfectly spot on, in that it more or less perfectly encapsulates my thinking on this. It might also provide relevant background for these remarks. Like every word I have ever written, it is pure gold, worth revisiting regularly.

Consider, if you will, the following thought experiment. Suppose one constructed a simple robot, which simply rolled about the streets of a reasonably sized city, taking square photographs with a normal length lens from about waist height. Yes, black and white, of course, what sort of animal do you take me for? Let say about one exposure, autofocused on something roughly near the center of the frame, every 5 seconds.

Turn it loose for for 700,000 seconds. Call it 20 days if we stick to about a 10 hour shift during the daylight every day. This produces an archive of, what hey, 140,000 pictures.

Will there be some good ones? Even stellar ones? Yes. There will be. Likely you would be startled by the occasional quality that would turn up. Before you get too het up, let us consider how these pictures would differ from those of the amateur with 100,000 photos on a huge RAID system, who to be honest probably has far fewer really good individual pictures. Our amateur, regrettably, almost certainly has any number of bad habits. He almost certainly "works" a scene, taking 10 bad pictures where 1 would do. He almost certainly has some stupid ideas he has borrowed from flickr, or instagram. He is quite likely burning exposures "practicing" things by photographing his cat using his new beauty dish.

Our robot, on the other hand, takes no duplicates at all. Because it is not fumbling with autofocus and exposure modes it does not understand, it generally gets pictures that are more or less properly exposed and properly focused. But mostly, in its little robot heart, every picture it takes is a brand new unsullied and completely earnest attempt to take a good picture of whatever is in front of the lens.

This is how I know that the robot would produce good work, albeit at a very very low rate: I have made myself into that robot. I have set aside my preconceptions, my bad ideas, my stupid tropes, and simply blasted away. The results were fascinating. If you have not performed this experiment, I highly recommend it.

Based on this, I feel comfortable asserting that while the rate will be low, there will be many hundreds of pictures which, with perhaps a moderate crop to account for the robot having not even the rudiments of framing but simply pointing at random, are pretty good. This one might look a bit like that one Stieglitz, that one a bit like an Arbus, and so on.

Now, there is no artist here. So of the three options I suggested in my previous remarks: 1. Find the artist, 2. Create a fake artist, 3. Pull out the greatest hits, only the last two are available. I am certain that the last two would be completely available, though. You could certainly pull out 10 or 20 things that are sufficiently reminiscent of "great photos" and mix them in to another 100 things that are generically pretty good and vaguely in the same style, and produce a book indistinguishable from Vivian Maier: Street Photographer. You could also produce probably several books that appeared to be coherent bodies of work by selected accidental repetitions of tics, tropes, ideas, and sequencing the resulting pictures carefully. The latter books would be far less marketable.

What is unlikely to happen is that you are able to extract a portfolio of work that is both coherent and appealing. God knows that the vernacular photography crowd can manage coherent, and occasionally they hit upon a particular tic that is humorous enough to carry a project. I am thinking in particular of one project (book? show? maybe just someone's inspired tumblr?) in which every photograph includes the cast shadow of the photographer. This, it is worth noting, means that all the photos are lit the same way, in addition to "including" the photographer, so there's a great deal of stylistic coherence here. And, it's kind of witty. But it certainly isn't great, or even important, or much interesting past a few chuckles and a flip-through.

Why do I think it would be hard to extract a portfolio that was both coherent and appealing?

It boils down, essentially, to statistics. If 1 in 100 photos is, at least a little bit appealing then we have 1400 candidates. Say portfolio needs to contain at least 50 photographs to be a meaningful "body of work" for an imagined artist. It's pretty easy to pull a coherent collection of 50 when we have 140,000 photos to choose from. It's lot harder when we have only the 1400 (or fewer) appealing ones.

This applies far more broadly than our little notional robot, by the way. With all these photographers running around with digital cameras, we see tons of people grinding out more or less appealing pictures that are more or less copies of other appealing pictures. The hand of the photographer is largely invisible in these, completely overwhelmed by the hand of whatever influences are driving the stupid pictures. More rarely, we see photographers grinding out coherent collections of pictures built around some sort of shared notion, some sort of an idea and very often these pictures are not particularly appealing.

This, fundamentally, is why the pre-1950 idea of the single great "image" is an idea that deserves to die.

In 1850 a good photographer was one who could reliably get any kind of picture whatever to appear on the plate. 50 years later this is now easy and the bar rises, you have to be able to place forms and tones and so on, to create a single appealing picture. This sticks with us as the underlying conceit, even though increasingly artists are in fact producing coherent portfolios of appealing (in some sense) pictures. There's a bit of a glitch when color becomes dominant, and then digital shows up. The bar rises again, because now literally anyone can take as many "good ones" as they like. Exposures are free, so just make a lot of them.

The fact that so many photographers somehow manage to produce no "good ones" at all is a social problem, as noted above, not a technical one nor one of talent/lack thereof.

Even I, saddled with my own set of bad habits and foolish tics, can manage to produce a couple of appealing pictures, a couple "good ones" in a year of a few thousand exposures, pretty consistently. Very few of them are worth a damn, though, because my ability to combine appeal with idea remains low.

No, the bar has risen, and the measure of a good photographer is now: can produce bodies or work that are both more or less coherent and appealing, around ideas that are at least roughly interesting.


  1. Your robot already exists. It's called the Google Car. There are whole websites dedicated to amusing still pictures pulled out of the Google Car video stream.

    Susan Sontag saw this coming a long time ago when she wrote “there are pictures taken by anonymous amateurs which are just as interesting, as complex formally, as representative of photography’s characteristic powers as a Stiegliz or an Evans.” This does seem to support the point you're trying to make here. Substitute "anonymous robots" for "anonymous amateurs" and there's the future.

    Instead of wasting time worrying about this (is it really debatable?) I think we need to sort out what this means for photography made by humans. What really distinguishes "great" work by human photographers from the work of lucky amateurs and robots working the probabilities? You've advanced two criteria: coherent and appealing. That's a start, but I'm not confident that coherent and appealing are going to be the exclusive preserve of humans for much longer. The things that make a photo appealing are well understood (not by every photographer, obviously, but they're out there). That sounds to me like the basis for an algorithm that could be used to guide your robot. Coherence is a bit trickier, but if it's as simple as similarities in viewpoint, style, content, etc., within a group of pictures that gives the impression of some kind of organizing idea, then coherence ought to be programmable too.

    Where do we go after after the robots can do appealing and coherent as well or better than most photographers? I'm not sure. I hope that the individual vision, insight, idea of really good, thoughtful photographers will remain more interesting than what's behind the algorithms. But I'm not confident that's always going to be the case.

    Anyway, I appreciate your efforts to think these things through Andrew.

    1. Google's street view is actually not as good as my robot since it is at pains to eliminate as far as possible the fascinating interplay of human interaction -- and yes, it still produces pretty good pictures from time to time!

      The robots are certainly coming. There's no doubt you could, will soon be able to, teach a neural net to take appealing pictures on a theme, and then it will be quite difficult to distinguish the work of humans from the work of robots.

      Perhaps the bar will rise again, to what I cannot imagine. Perhaps not!

  2. This is exactly how I take photographs. Like a monkey with a typewriter. I like to imagine that my ability to throw away trash - and very occasionally, when I overcome my almost infinite laziness, my ability to hack away superfluous image aka crop - gives rise to collections that are somewhat coherent and somewhat interesting. I always find them extremely satisfying and they are primarily for me so WIN. But the start point is basically the monkey-robot.

  3. Replies
    1. Not sure what "that" refers to! But I appreciate the sentiment, thanks.

    2. Aspirational/invidious obsessing over "greatness".

    3. Well dammit. and here I thought you were talking about my monkey-bot approach. Or maybe you were?? It's deliciously unclear!

  4. Pair your photobot with an Instagram bot and a random inspirational quote generation script and you might even sell a few 'I Shoot BOT' wooly hats or a street pho-togging workshop.