Up in the sky, it’s a bird, it’s AI

A bird is merely a bird until it is spotted outside its natural range. Then it becomes a sign to be decoded. A photograph taken in Brazil of a red-winged blackbird, which is native to North and Central America, was recently posted on an online forum, leading to much speculation among scientists about what its unusual presence could possibly mean. Was it a vagrant, blown off course by some miscalculation of its internal GPS? Or an indicator of how greater forces — climate change, habitat destruction, etc. — were impacting the behaviour of different species? A little detective work helped uncover the truth: The photographer who had submitted the image had captured an epaulet oriole, common in Brazil, and asked an AI platform to make the picture “look better”, resulting in a single modified detail — and a false sighting. A recent article in Nature argues that this is precisely the kind of “AI slop” that is making life difficult for scientists who rely on public tracking platforms to study wildlife: Not the obviously fake images of tigers in the savannah or parakeets in the Arctic, but the subtly manipulated ones that just might — and often do — pass for documentary evidence. The study of nature is not merely about pretty pictures or awe-inspiring vistas. It is about understanding ecological shifts that often portend changes in the world humans inhabit, from warming temperatures and disappearing habitats to the spread of disease and invasive species. While many wildlife photographers use technology to tweak certain parts of a photograph in order to get the “perfect” image, AI tools, with their tendency to add or “hallucinate” details, can, literally and figuratively, lead researchers on a wild-goose chase. Because in science, even the smallest fabrication can end up obscuring a larger truth.