If you’ve been on the web lately you’ve likely come across a deepfake. Some are terrifying, some are amusing, but without a doubt, they are becoming a big topic of discussion. You’re likely wondering what exactly they’re all about, if they’re dangerous, and if so how to protect yourself. So, we’re here to give you an overview of this new form of artificial intelligence (AI).
What is a Deepfake?
In short, a deepfake is a fraudulent piece of media that has been manipulated by artificial intelligence. The word comes from the combination of two words, “deep learning” and “fake”. This occurs when a video or audio is produced to make it look like a person did or said something that did not actually occur. Deepfakes are created by artificial technology reviewing hours of video footage detecting a person’s facial movements, speech patterns, and more, and then replicating it. As mentioned before, deepfakes are not only videos, as is typically imagined, but they can be fraudulent audio recordings as well.
Are Deepfakes Dangerous?
Deepfakes are often used for very nefarious purposes. For example, they can be used during a presidential election to slander a candidate’s character to make it seem like they said or did something that they didn’t do. Many prominent political figures such as Obama and Trump have their own deepfakes floating around on the internet. The biggest problem with this is of course that deepfakes are a powerful tool for those who wish to spread misinformation and blur reality around important political issues. Deepfakes also pose challenges to the legal system as fake videos and audio recordings could potentially be submitted into evidence for instances such as child custody cases.
Deepfakes can also be used for fraudulent activity by tricking the victim into thinking that they are speaking with someone else. An example of this was last year when an employee at a U.K.-based energy firm was scammed out of 220 thousand pounds, when he thought he was speaking with his boss, the CEO, over the phone. In the entertainment industry, deepfakes can be used for more beneficial purposes such as improving the dubbing of foreign language films, or more controversially be used to feature actors as iconic characters (looking at you, Luke Skywalker).
How to Spot a Deepfake?
Big tech companies such as Microsoft and Facebook have started their own initiatives to help detect and mitigate deepfakes on their platforms. According to Reuters, the two companies have announced they will be working with leading universities around the United States to launch a contest to better detect deepfakes.
Ironically, AI technology which is used to create deepfakes is also being used to spot them. Deepfakes are becoming increasingly more difficult to detect though as technology continues to advance. In 2018, US researchers discovered that deepfake faces didn’t blink normally. At first, this seemed like a catch-all that could help solve the problem of detection. But as soon as the research was published, deepfakes started to appear with blinking. The Catch-22 of this quickly evolving technology is that as soon as a weakness is revealed, it is fixed.
Although, there are still a few key characteristics that can help give away deepfakes, such as unnatural hair movement, patchy skin coloration, strange positioning of hands and body, poor lip-syncing, strange lighting, and voices that sound robotic and unnatural speech patterns.
To be certain we are only at the beginning of what deepfake technology can do. Like Photoshop, as time goes on people will ultimately become more attuned to spotting these altered videos and audio recordings. The truth is Deepfakes are probably here to stay, and I for one am interested to see where this technology goes and hope that it will be used for more positive forces in the world rather than negative.