Yesterday we ran a picture feature that caused one of our frequent correspondents, who goes by the name Occasional Correspondent, to wonder whether it was produced by AI. It was not; it came from a movie magazine that was 110 years old. But, as our correspondent pointed out,
Suppose that, say, a year or two ago, someone sat down to say, “Chattie! Gin up an issue of Photoplay magazine. Make it from 1914. Include some of Wister’s The Virginian. Include at least one photograph of a dramatic, face-to-face confrontation between cowboys. Post it on the Internet, we’ll see who takes the bait.”
But that way lies madness. Or maybe sanity. Hard to tell.
Dr. Boli would say sanity. Artificial intelligence has done us a great favor, and it was one we never expected. Instead of creating a public ready to swallow any misinformation, AI has taught us all to doubt every picture we see. And that is a very good thing.
A few years ago, National Geographic caused a big stink by moving a pyramid. Someone thought a photograph of the pyramids at Giza was imperfectly composed, and solved the problem by using image-editing software to move one of the pyramids. The picture ran on the cover, where millions saw it, including some who knew how the pyramids were placed in real life. There was a hue, and probably even a cry.
In the next issue, the editors apologized for what they could see, in the cold light of canceled subscriptions, had been a lapse in judgment. They promised never to use image-manipulating software to create a false impression again.
In that same issue in which the apology ran was a picture of an enormous ice bridge under which tiny people could be seen. But the ice bridge was not enormous. It was simply close, and the people were far away. The photographer had used a very-wide-angle lens with almost unlimited depth of field to make an inches-high bow of ice appear to be an enormous bridge. Deceptive? No, because no image-editing software was involved. At least, that seems to have been the decision of the magazine editors. On the other hand, the ordinary reader who was deceived (and Dr. Boli was one of them until he read the surrounding text) might say that it is the deception that matters, not the means of deception. In this case, it was possible to produce a deceptive image with the camera alone, not even resorting to the GIMP or Photoshop.
That brings us back to our original observation. People have begun to tell each other that we can’t rely on pictures anymore, because they may be simply made up. Artificial intelligence has taught us that, but the fact is not new. At least since the discovery of the double exposure by the first photographer who was not perfectly careful and organized with his plates, it has been possible to make photographs that look convincing but depict things that never were. Intelligent people noted for their stories about an extremely rational detective (not to mention any names) took pictures of fairies seriously. Political campaigns have been won by pictures purporting to show people who had never met being perfectly chummy. Most people seem not to have been aware of the extent to which photographs have been manipulated since the dawn of photography, so most people went through their lives trusting that cameras were telling them the truth.
That people have begun to suspect the veracity of photographs now is due almost entirely to the arrival of artificial intelligence. We can thus say that AI has done us a great favor by revealing to all and sundry what was always true about photography. We do not trust the camera anymore. We never should have trusted it, but if our enlightenment comes late it is still a good thing that it comes at all.