AI Photos Fooling Eyes: Spotting Fake Images

Northwestern University researchers have developed a guide to help identify AI-generated images by highlighting common visual flaws. These artificial intelligence systems, while adept at creating realistic scenes, often fail to grasp fundamental concepts like human anatomy, physics, and object functionality. The research team cataloged several recurring mistakes to empower viewers to scrutinize suspicious visuals more carefully.

Four categories of warning signs have been identified: anatomical implausibilities, stylistic artifacts, functional implausibilities, and violations of physics. Anatomical errors can manifest as extra or missing fingers, malformed hands, distorted eyes, or asymmetrical teeth. Stylistic clues include waxy skin, oversaturated colors, and an overall 'overperfection' that deviates from natural photography. Functional errors occur when objects, like clothing or gear, fail to operate as they would in reality. Lastly, violations of physics are evident in mismatched shadows, illogical reflections, and warped perspective.

While these indicators are not conclusive proof of an AI origin, they serve as crucial prompts for closer examination. Researchers emphasize that an anomaly in an AI-generated image is a reason to investigate further rather than a definitive verdict. By paying attention to these visual discrepancies, individuals can develop a more critical eye when encountering images online and mitigate the spread of misinformation.