Orange flames exploding from the heart of Lahaina send red-hot embers streaking across the sunset sky.
Anyone there on Aug. 8 would immediately know the image available on Adobe Stock is a fake. Where is the choking cloud of black smoke that turned day to night? Why aren't the palm trees being battered by cruel winds? Where are the sailing leaves, limbs and debris? Why isn't the ocean a wind-chopped froth?
The Lahaina picture was featured this week as part of a Washington Post story about images generated by artificial intelligence that look like prizewinning photojournalism. Other examples included children in war zones, a phalanx of protesters, bombs exploding in neighborhoods and a victim in the throes of domestic violence.
In response to the Washington Post probe, Adobe has reportedly vowed to crack down on misleading AI-generated images that seem to depict real news. We appreciate the words, but cannot help wondering if the speed and overwhelming force of the Lahaina disaster is an apt metaphor for the AI-generated firestorm to come. As the technology grows and improves, how long before it is impossible to tell reality from fake? How long until AI knows enough about each of us to spoon-feed exactly what we believe?
AI-generated images are supposed to be identified as such by those who use them. Here at the newspaper we have our own protocols for the rare pictures that are manipulated. We label them as photo illustrations and usually mention in the caption what was done. Unfortunately, according to the Washington Post article, AI images have been slipping in as real news on the web and by accredited media.
There are ways for consumers to verify. We can look for watermarks that are embedded somewhere in the image. We can also search for distortions or anomalies such as repeating patterns and unrealistic clarity. AI reportedly has a problem with hands and teeth. Ask any artist and they will tell you the same thing. They're hard to capture. As for deepfakes, which deliberately present falsehoods, often through the use of facial recognition technology, confirming the sources and story itself may help weed out the nonsense.
No wonder we feel fortunate to have lived so much of our lives when journalism was a bastion of the unvarnished truth. Stories were fact-checked and vetted. Opinions were saved for the editorial page where they belonged. Sure, there were fakes in tabloids like the National Enquirer, but those manipulations were done with scissors and paste and almost comedic in their inferior quality. It was as if the head-swaps and alien monsters living amongst us were presented with a wink.
The stakes are so much higher now. Real journalists form one of the last lines of defense against a deluge of disinformation, much of it designed to rip our society apart. Thank you for supporting this newspaper and its people.
