These days, even experts struggle to detect fake images and videos on the Internet. With AI coming along as far as it has, deepfakes are becoming more commonplace, with fake faces and voices being indistinguishable from the real thing to the naked eye.
Deepfake refers to videos that are digitally altered, often using AI, to create fake video scenes with a person’s face or voice. As you can imagine, these have been used by content creators, sometimes to the detriment of the actual person’s life.
FakeCatcher is Intel’s answer to the problem at hand. FakeCatcher is an AI that can almost instantly spot a deepfake with 96% accuracy.
The internet is great, but it’s also riddled with misinformation that continues to pose a threat to authentic information. Basically, you just can’t trust what you see on the internet anymore.
Intel says FakeCatcher has been trained to detect real subtle human movements in a video, as opposed to most fake news detectors that attempt to analyse the raw data in the video files. The AI can also track bloodflow in the pixels it’s analysing. Something that other deepfake detectors haven’t been able to manage yet.
If all goes well, Intel plans to eventually make the web-based platform available to the general public.