Can you tell if AI created the image you’re viewing? It’s becoming increasingly difficult these days, thanks to advanced AI image generators like DALL-E 2 and Midjourney.
However, there are several ways to spot AI-generated images. Here’s how you can do it:
Check the title, description, and comments section:
When sharing photos, some individuals may explicitly mention the use of artificial intelligence in the image description or comments. So, it’s always worth checking these sections for clues. Additionally, visiting the profile page might provide keywords like Midjourney or DALL-E, indicating that the images are likely AI-generated.
Look for watermarks:
Watermarks can be another helpful clue in identifying AI-generated images. For example, DALL-E places a subtle tag on every image downloaded from their website. However, removing the watermark is allowed by OpenAI’s guidelines as long as it’s not used to mislead others about the nature of the work. Midjourney, on the other hand, does not use watermarks.
Identify anomalies in the image:
AI-generated images often exhibit distinct visual distortions, which can be more apparent upon closer inspection. These distortions may resemble the brushstrokes found in impressionist paintings and can manifest as various anomalies in the image, such as missing or mismatched earrings, blurred backgrounds resembling textures, indistinguishable background text, facial asymmetry, painted-like dots, or glasses blending into the skin. Spotting these small details can help reveal whether an image is AI-generated.
Use a GAN detector:
AI image generators operate using Generative Adversarial Networks (GANs). While GAN detectors exist, their results can be mixed and unreliable. It’s best to use such tools in combination with other methods to verify if an image is AI-generated. It’s worth noting that future advancements may lead to more accurate GAN detectors, so keeping an eye on developments is important.
Looking ahead, AI generators will continue to improve, making AI-generated images incredibly convincing and difficult to distinguish visually. In such a scenario, relying on visual aberrations alone won’t suffice. Hopefully, future apps or websites will provide solutions to detect AI-generated images, similar to reverse image search capabilities today.
For now, creators using AI to generate images should adhere to OpenAI’s advice and be transparent when sharing their work.
To summarize, when identifying an AI-generated image, check the title, description, and comments for any mention of artificial intelligence. Look for watermarks or visual distortions in the image. You can also utilize a GAN detector, but consider its results alongside other methods. Combining these approaches will help you determine if an image is AI-generated.