The A-Z of AI

Fakes

Fake images, audio or video made using AI.

AI has given rise to new ways for technologies to create hyper-realistic “deepfake” images, speech, music and video.

Deepfakes work by studying existing real-world imagery or audio, mapping them in detail, then manipulating them to create works of fiction that are disconcertingly true to life.

These technologies, considered impossible just a handful of years ago, have found a wide variety of applications across everything from CGI in Hollywood movies to music production and pornography. While many are intended to be entertaining and imaginative, used inappropriately they can also generate misinformation that may be harmful to society: convincing fakes injected into public discourse can do significant damage to public knowledge and trust.

Fakes can be convincing, but there are some telltale signs that distinguish them from reality.

A diagram called "spotting a deepfake" shows a visual of a man, signposting features that indicate he may be AI generated.

In fakes of recent years, visual renders of people may look a little glitchy, like computer games. Voices might sound a bit robotic. Characters also blink less or repeat their hand gestures, as it can be tricky for creators to find, map and replicate source material of people with their eyes shut or hands exposed.

As fakes become more sophisticated, the methods and technologies used to detect them must also remain in a constant state of evolution.