Recognize AI-generated pictures and videos: These 10 characteristics reveal Deepfakes

The Boss

Recognize AI-generated pictures and videos: These 10 characteristics reveal Deepfakes

AIGenerated, characteristics, deepfakes, pictures, recognize, reveal, videos

Artificial intelligence (AI)

More and more AI generated fake images and fake videos flood the Internet. Even experts are becoming more and more difficult to recognize them. But there are certain information.

In many cases you can (still) uncover AI-generated fakes without appropriate software. You have to pay attention to a few certain things.

In many cases you can (still) uncover AI-generated fakes without appropriate software. You have to pay attention to a few certain things. (Source: Terovesalainen/depositphotos.com)

  1. General tips from the federal government

  2. Recognize abnormalities of AI fakes

They sprinkle disinformation, flood video platforms with garbage and spread conspiracy theories in chat groups. AI-generated fake content on the Internet has increased rapidly since the upswing of Chatgpt and such Deepfakes are not always to be understood as a fun picture assembly. They are often used specifically for dodgy purposes.

For example, deep -recognizable frauds are carried out or false information about politicians and celebrities are distributed. Therefore, it is all the more important for every internet user to know which pictures and videos the AI ​​helped.

Unfortunately, this is easier said than done, because artificial intelligence continues to develop. AI fakes are becoming increasingly difficult to recognize. Like AI generated texts and audio content, AI-generated pictures and videos are often not perfect and have a few bizarre features. They are difficult, but not impossible to recognize if you know what you have to search.

Before the dangers of Deepfakes The German federal government also warns And give some tips with which you can get AI fakes.

General tips from the federal government

  • Examines the image in the largest possible resolution on the largest possible screen in order to better discover abnormalities.
  • Pay attention to the facial expressions. AI programs still find it difficult to imitate certain reactions in the face area of ​​people (such as frowns and blinking)
  • As usual with every research, a source test also helps. For pictures, a backward search is recommended (camera symbol on the right in the Google search bar).

We also rummaged in our wealth of experience and would also like to give you a few tips on the way with which you can protect yourself from Deepfake fraud and disinformation by AI-generated image and video content.

Recognize abnormalities of AI fakes

  • With certain body components of people such as hair or fingers, many AI still make mistakes. Unusual fingers or hairstyles through which a false background color shimmers should let you notice.
  • When many people can be seen in the picture, the AI ​​likes to copy certain characteristics or even whole faces. Pay attention to such copy paste content.
  • Missing or strange shadow throws are another characteristic for AI fakes, especially in pictures of outdoor areas.
  • Many AIs are difficult with official logos and writing. Pay attention to strange characters and false logos.
  • Transitions between people or in the picture or video of very present objects can also provide information. A well -known example of this is unusually long legs and feet if a person to be seen is about an object in front of your hip.
  • Always think logically when looking at pictures and videos. If the content shows a particularly unusual event or unusual behavior of a person, you should at least note.
  • Angle and background are other indications. An impossible position of the photographer in particular can easily expose a fake. A location whose entry is prohibited for unauthorized and press should also encourage you at least to take a closer look.
The AI ​​fake is easy to see here. The gentleman in the foreground has only four fingers on the right hand. In addition, the lamp protrudes unusually from the shelf in the top left. The shadow throw in the back also looks a bit too dark.

The AI ​​fake is easy to see here. The gentleman in the foreground has only four fingers on the right hand. In addition, the lamp protrudes unusually from the shelf in the top left. The shadow throw in the back also looks a bit too dark. (Source: from Netzwelt with the help of Leonardo AI generated AI content / assembly: Netzwelt)

Unfortunately, these characteristics do not expose all AI fakes alike. Larger AI programs such as Chatgpt always create more realistic pictures and videos. For example, the AI ​​programs in the representation of fingers have now become significantly better. Earlier mistakes such as hands with seven or eight fingers are therefore much less common.

Basically, these tips give you the ability to distinguish AI content more easily from real recordings. Basically, you should always remain skeptical when surfing the network if you discover the above -mentioned features on pictures or in videos.

Mandatory AI markings would be an option to contain this fake flood. However, they would only work if they were also implemented uniformly worldwide. After all, at least the European Union has taken a step in the right direction with the labeling obligation in the EU AI ACT.

  • » Tipp: The best VPN providers for more security and data protection
  • »Buy the balcony power plant: Compare the best solar systems

Nothing missed with that NetworkNewsletter

Every Friday: the most informative and entertaining summary from the world of technology!

Leave a Comment