Here’s How Experts Say You Can Spot A Deepfake

( – Being able to spot a deepfake is an important part of being on the internet and being aware of misinformation. With artificial intelligence taking over, there has been a lot of concern when it comes to deepfakes and being able to spot what’s real and what isn’t.

There are some detection tools that you can use commercially and in research labs which will help to tell you if it was made by AI or not. These deepfake detectors are used by uploading the image or video and the detector will give you a likelihood of whether it’s made by AI or not. It’s very similar to how artificial intelligence detection is done with writing.

With this said, people’s normal reactions and senses to an image or video are also usually right, when looking at and analyzing media. There are some key giveaways that we see often with these artificial images and videos that can help you spot them before having to put them into a detector.

Siwei Lyu made the DeepFake-o-meter, which allows users to upload media and run it through different labs to get an idea of if it’s AI-generated. There are also some concerns as some AI-detection models are run by artificial intelligence, which could cause a bias when detecting the potential of a fake image or video.

Says Lyu, “I think a false image of reliability is worse than low reliability because if you trust a system that is fundamentally not trustworthy to work, it can cause trouble in the future.”

Lyu stressed that a human operator should be brought in to do the analysis on the media to give a more accurate response on whether it’s AI or not.

There have been a lot of concerns regarding media in the form of not only video or images but also audio. One example of an artificial intelligence audio recording was the robocall that circulated through New Hampshire using President Joe Biden’s AI-generated voice.

Experts have said that detecting audio as a deepfake is more difficult because it relies simply on hearing. When examining photos, it’s important to zoom in and look closely at any inconsistencies or areas that don’t look normal. When looking at video, it is often harder to recognize videos without people as AI, but generally, if there are people in the video, then looking for discrepancies is one way to identify an artificial intelligence-created video. Some cues in the video included unnatural eye blinking and some movements that showed pixelated areas.

Copyright 2024,