A.I.

In recent years we’ve become accustomed to seeing computer-generated images that look a lot like what we used to know as photographs. We often call them photographs (sometimes “A.I. photos” or “A.I.-generated photography”) even though they don’t involve a camera at any stage; an image made by a computer program that was trained on photos is not itself a photo, if that means a real recording of real objects using real light. When A.I. makes images that look to us like things in the real world, it’s learned how to do that from existing images, not from the real world itself. At the moment A.I. is not too far removed from reality, because its source material is directly anchored there, but that will probably become less and less true in the future, when (under the “model collapse” scenario) the pool of data A.I. learns from will be ever more polluted by the increasingly degraded and unrealistic imagery A.I. itself has created. A.I. photography is an imitation, not even of the world that the optical–chemical medium of photography was invented to record, but of photography—yet we can’t necessarily distinguish photos from photo-like images amid the great mass of stuff we consume daily. Actually, fakery is nothing new in photography. A terrific exhibition at New York’s Metropolitan Museum, “Faking It,” made this point back in 2012: photographers have been “improving” and even fabricating images for as long as the medium has existed. They’ve done this to make art, to entertain, to fudge, and to falsify; subterfuge is sometimes part of it and sometimes not. Some photographers use whatever technical means are at their disposal to make the output of their cameras look more like what they saw with their own eyes; but they sometimes have notions of “truth” that go beyond optical literalism. Other photographers never were interested in what they saw with their own eyes. If we believe that images involving a camera must be veracious, must correspond to reality in some direct way, well, we should never have believed that. But lately our confidence in the veracity of photography has been shaken as never before: A.I.-produced “photography” has become so pervasive that everyone is aware of it. Like the photographic fakery that’s been around forever, A.I. fakery isn’t always intended to deceive; it’s sometimes obviously “creative.” But what else is going on? We are now uncomfortably aware that A.I. may be behind images we once would have thoughtlessly accepted as direct representations of reality. These days we can’t look at a photo that’s meant to convince us of something without wondering about how it was made. And so the purpose of photography in many of its applications is changing before our eyes. Photography as proof or evidence is being replaced by photography as illustration or decoration. In the age of A.I., we have to have seen something ourselves in order to be completely sure it actually happened or looked like this or that. And we won’t be able to get other people to share our conviction by taking a picture. In sum, we no longer trust photography very much. The new technology, which we don’t trust at all to represent reality as it is, has made the old technology untrustworthy, too. But snapshots are the most trustworthy of photos. Earlier (here) I made the point that snapshots are more trustworthy than other kinds of photos because we know they were taken for simple documentary reasons. They are rarely tendentious: the photographer’s purpose was just to get a record of a scene or a moment. But there’s another reason we trust snapshots, and that’s the fact that we know they’re technically primitive. There’s just no opportunity for fakery to creep in. You may want to doubt that this example is showing us something that actually happened, but you really can’t. It happened. It's kind of shocking to consider that an ultra-crude photographic technology has what it takes to convince us, while the whiz-bang modern one has apparently forfeited all claim to believability. Snapshots, almost all of them family photos, were never intended to be seen by anyone else. But when they do make it into the outside world, they communicate straightforwardly the factuality of their mundane scenes; whereas media photography aimed at a public of millions contains less and less that anyone has to put any stock in. Like so much of modern technology, our photography is isolating. It purports to be about reality but, increasingly, isn’t; by communicating less and less of that kind of truth, it effectively locks us into our own ideas of what the world is like. By contrast, snapshots are entirely intersubjective, even though they were meant to be private.

Roll over to enlarge