Are you sure the image you just saw is real? In the age of artificial intelligence, distinguishing truth from falsehood has become a superpower. Deepfake photos are everywhere, and their quality is impressive. But all is not lost. Indeed, there are clues that can help you unmask the deception.
Are you ready to test your skills? Take our quiz to find out if you’re a master of image recognition.
The Technique Behind the Deception
How are deepfakes created? Their creation relies on two AI algorithms that work together: Generative Adversarial Networks (GANs). One of these, called the “generator,” is responsible for creating the fake image. The other, the “discriminator,” analyzes it to determine if it is authentic. This process repeats millions of times until the generator can deceive the discriminator. Consequently, the result is an incredibly realistic image.
The Clues AI Can’t Hide
Artificial intelligence systems are not yet perfect. Therefore, they often leave traces that a careful eye can spot.
For Images with People
- Eyes, Teeth, and Ears: Deepfake images often show anomalies in these areas. Eyes may have unnatural reflections, while teeth appear blurred or crooked. Similarly, ears may have irregular or asymmetrical shapes.
- Shades and Shadows: Shadows in a deepfake image are not consistent with the direction of the light. AI still struggles to replicate the physics of light, producing flat or unnatural lighting.
- Inconsistent Details: You may notice defects like “faded” hair in the background, distorted jewelry, or an unnatural blending of a person with the background.
- Facial Movements: In videos, AI has trouble replicating natural movements. Expressions and blinks are often unnatural, robotic, or completely absent.
For Images with Objects and Landscapes
- Physical Anomalies and Perspective: Image generation algorithms can struggle with real-world physics. For example, water might have strange ripples or reflect light unnaturally. Objects in the distance might not follow the laws of perspective, appearing too large or too small for the context.
- Illegible or Distorted Text: Artificial intelligence still has great difficulty generating coherent text. If you notice blurred letters, deformed characters, or nonsensical words on signs, books, or labels, it’s a strong sign that the image is not authentic.
- Irregular Colors and Textures: Sometimes, the colors and textures of an object don’t blend realistically with the surrounding environment. You might see a car with paint that seems to “float” or a brick wall with a pattern that repeats too perfectly or unnaturally.
- Repeated Details or Anomalous Patterns: AI tends to replicate patterns in an inorganic way. Look at backgrounds like grass, tree leaves, or tiles. If you see identical motifs that repeat in a serial manner, it’s a red flag.
Why It’s Important to Spot Deepfakes
Deepfakes are not just a game. They can be used to spread misinformation, manipulate public opinion, or damage people’s reputations. Their proliferation represents a threat to society. Therefore, it’s crucial to develop critical awareness and learn to distinguish reality from digital fiction.