Curating the World
Toby Ord
How much information is in a photograph?
The standard answer views the photograph as the image file your camera produces. A mid-range DSLR these days might have a sensor with 20 million pixels, each of which is represented with 14 bits of information, giving a 35 megabyte RAW file. Lossless compression may take this down to 20 megabytes. Or with some almost imperceptible lossy compression you might get it down to 10 megabytes. These are the kinds of numbers that are relevant when storing the images or sending them to a friend. As a photographer, this is how I usually see things.
But there is a deeper way of looking at the question. How much information is in a photograph? The first answer started with the image file, but we can widen our view and take in the whole act of photography. We can see a photograph as an act of communication from the photographer to their audience. As with any art form, the photographer has something they want to say (or show, or express) and they do so by making an artwork then sharing it with their audience. This communication is limited by the sensor and the file size — but it is also limited by how much is under the photographer’s control, and this turns out to be much more significant.
When taking a photograph, the photographer chooses a time, a place, an angle. They shoot with their choice of camera, lens, and (at times) film. They adjust a range of settings, such as ISO, exposure length, aperture, focal length, and focus distance.
Most of these are continuous choices, which we can represent via just 12 numbers:
1 for the time
3 for the place (longitude, latitude, altitude)
3 for the angle (heading, elevation, roll)
1 for the ISO
1 for the exposure length
1 for the aperture
1 for the focal length
1 for the focus distance
Let’s be very generous about how we represent those numbers, allowing wide ranges and extremely fine-grained measurement within them:
time: a million year range, to the nearest millisecond — 3 x 10$^{16}$ possibilities
place: anywhere within 100 km of the Earth’s surface to the nearest millimetre — 5 x 10$^{28}$ possibilities
angle: measured to the nearest arc-second — 2 x 10$^{18}$ possibilities
ISO: from 1 to 1,000,000 — 10$^6$ possibilities
exposure length: from 1/10,000 seconds to 1 million years in 0.1% increments: — 40,000 possibilities
aperture: from ƒ/0.5 to ƒ/500 in 0.1% increments — 2,400 possibilities
focal length: from 5mm to 5000mm in 0.1% increments — 7,000 possibilities
focus distance: from 1mm to ∞ in 1 billion increments — 10$^9$ possibilities
That is about 10$^{90}$ different possible combinations.
To these we need to add the discrete choices of camera, lens, and film. Let’s generously assume:
camera: 100,000 different camera bodies
lens: 10,000 lenses compatible with each camera
film: 1,000 kinds compatible with each camera
This gives a trillion different discrete configurations, taking us up to 10$^{102}$ different possible photos.
10$^{102}$ sounds like a lot. It is a lot. But isn’t much information.
You only need to write down 102 decimal digits to specify any possible photograph. That’s just 339 bits of information; just 42 bytes. There are fewer distinct photographs that could be taken than there are 19x19 grids of black and white pixels — fewer photographs than there are arrangements of stones on a Go board.
This is very little information compared with other art forms. Natural text contains about 1 bit per symbol. So a photograph contains about the same information content as a few sentences of text: a short paragraph or a stanza of verse. Art forms that allow expression through continuous curves or continuous modulation of one’s voice contain much more information. 42 bytes couldn’t describe even a few brushstrokes of a painting or a few seconds of singing.
What’s going on here? Why is there such a gap between the information needed to specify all the pixels in a photograph and the information content that can be expressed by the photographer?
One reason is that a lossless image file allows for all possible combinations of pixel values. If you randomly sampled from the space of image files, almost all of them would look like coloured static rather than a real scene — there is no process of pointing a camera and releasing the shutter that would produce them. The file size is so high in part because the file formats allow for astronomical numbers of non-realistic photos.
But there is, I think, I deeper way to understand this gap between file size and information content. Many art forms (like painting, singing, writing, or sculpture) are acts of creation — they are about adding something new to the world. Photography is fundamentally about curation. Classic examples of curation are carefully selecting works of art to exhibit in a gallery, or assembling just the right sequence of songs for a playlist. Photography can be seen as selecting from among all the scenes in the world and choosing just a few to bring to the audience’s attention.
Curating usually involves far less information than is contained any one of the works being curated. Think of how you only need to specify the artist, album, and track titles to pick out a song, but need a multi-megabyte file to store it. You could think of each selection as indexing into an existing database. Instead of sharing an entire item in that database, you only need to share the pointer to it. Indeed the information involved in taking a photograph much more similar to the size of a memory pointer (64 bits) than the size of what that points to; more similar to the size of a hyperlink (~50 bytes) than to a whole webpage. A photograph is a pointer into the world.
Photographers are seeing the world then telling you where to look, and how. The photographer is bringing 42 bytes to the artwork; the rest is being brought by the world. This doesn’t diminish photography, it just gives you a different lens through which to view it.
My most emotional moments as a photographer feel like this — times when I’m surrounded by aching beauty, but beauty which few others can see, and I’m fluttering around trying to capture the fleeting moments before they fade so that I can share them with others, helping others see. This feels much more like striving to do justice to what is already in the world than to making my own creation.
A photograph, in this sense, doesn’t contain much information. But not all information is equally important, and those bits the photographer brings are exceptionally dense in their importance. They show you where to look, and when, and how. Photography is fundamentally about curation — not curating a gallery, but curating the world.
Notes
Post-Processing
Our calculation of the number of distinct photos doesn’t account for post-processing. Does that change things? Adjustments made by one dimensional sliders change the story very little. Choosing one of 256 positions just adds one byte. A curves adjustment is controlled by a small number of handles, each of which adds 2 bytes, and a crop just involves choosing two pixels, for about 6 bytes. The same is true if these are applied to a feathered selection (just specify one of 20 million pixels times one of 256 feathering levels, or 4 more bytes). But if you spend a long time with selection or dodging brushes, you can effectively paint any picture you like on top of your photo, which does allow vastly more ‘photographs’. Though at that point, it would be fair to say it less a photograph and more a painting on top of a photograph.
Staged Photographs
Similarly, we might question whether a photographer is merely viewing a scene. If a photographer is staging a photo, then the photo could convey more than 42 bytes. For example, they could arrange a 100x100 grid of white or black stones and take a photo of it. There are more such images than 42 bytes can express. Even simple portrait photographs often involve substantial verbal direction and body language from the photographer, shaping the pose and expression of the subject. It would be easy for the photographer to exert more control over the image through their directions than through timing, angle, and camera settings. It is best to think of the calculations above as applying only to unstaged photography — where the photographer is as close as possible to being a mere observer. When it comes to staged photography, you could divide the artistic work into the staging and the photograph. The staging is the artistic act where a lot of information may come in, not the photograph.
6 September 2026