How to spot an AI-generated image

UC professor offers tips to Fox 19 on how to tell if an image is real or artificially generated

It's getting harder to tell the difference between real photographs and artificially generated images. In the near future, it could be impossible with just human vision, a University of Cincinnati professor told Fox 19.

Zag ElSayed headshot

Zag ElSayed, PhD, assistant professor in UC's School of Information Technology

Zag ElSayed, PhD, an assistant professor in UC's School of Information Technology, challenged Fox 19 journalist Megan O'Rourke to determine if images were AI-generated or real photographs. O'Rourke correctly identified the authenticity of some but not all of them.

“I predict in the next five years we will have 100% generated images AI that you cannot detect by just a regular human eye,” ElSayed said.

Using a simple python code and information readily available online, ElSayed also was able to generate images that looked similar to O'Rourke.

To detect the difference between real and fake images, ElSayed suggested three “truths” to watch for:

  • Physical law: Like close objects that are not focused or have blurry backgrounds
  • Logical law: Dealing with the overall composition of the image such as too many lighting sources
  • Natural imperfections

See more from Fox 19.

Megan O’Rourke headshot
AI-generated image using Megan O’Rourke's photo

With a simple python code that uses an OpenAI library (which a part of ChatGPT application programming interface) and DALL·E 3, Zag ElSayed created an artificially generated image on the right that is very similar to Megan O'Rourke, whose real photo is on the left.

Featured image at top: 3D render of AI and GPU processors. Photo/Igor Omilaev via Unsplash

Impact Lives Here

The University of Cincinnati is leading public urban universities into a new era of innovation and impact. Our faculty, staff and students are saving lives, changing outcomes and bending the future in our city's direction. Next Lives Here.

Related Stories