For those not yet aware of this phenomenon, deepfake technology generates videos that look real from artificial intelligence (AI). In general, this content shows celebrities and politicians doing or saying something that did not happen, to generate more audience and instill curiosity and shock to the viewers.
Deepfakes are essentially false identities created through deep learning through a human image synthesis technique based on artificial intelligence. It is used to combine and overlay existing legacy images and videos and transform them into “original” images or videos using Generative Adversarial Network (GAN) technology.
The term came when a Reddit user jokingly published videos recorded with celebrities faces. Since that publication, video editing has already generated satirical and even criminal content through false identity and untrue information.
Public figures such as Barack Obama, former US President or Mark Zuckerberg, one of the founders of Facebook, and Donald Trump, the current US president, were some of the main targets of deepfakes. In addition, actress Gal Gadot and singer Taylor Swift also had their video-related face used.
In 2020, not only can your identity be stolen, but most importantly, you can be framed, defamed, and scammed with deepfake videos, which can be hard to prove fake.
Deepfake technology can convince a crowd that the content displayed in the video is real. Although there are sharp differences between a real video and another made from this feature such as wax-like skin and the lack of sync between audio and mouth movement, some people believe the content is true.
Also, the human eye can check that the ends of the artificial faces are quite sharp because the AI does not produce 3D content and stretches the face of the character to disguise the movements.
It is expected to be used in the 2020 elections in the United States and the government took measures in order to stop such content from spreading. Congress has demanded that videos are required to be properly classified as deepfakes, with watermarks identifying them.
There are already free programs used to produce deepfakes, however the quality is not as sophisticated. In addition, a good video card is still required to process content faster, which is not accessible to everyone. With the increasingly upgrades of artificial intelligence in the coming years, the use of this software will be easier and more accessible, allowing anyone to produce this content.
Currently, there are over 14,000 deepfakes with pornographic content on the web. In 96% of those cases, women’s faces are used as end products. Two techniques are currently used to produce the videos. The first uses real pornographic images and only changes the face of the actresses to that of a known person. The second, known as Deepnude, uses a person’s full photo, removes their clothes and creates the pornographic, nude image.
Recreating anyone’s face from deepfake technology is not too complicated. Thus, facial recognition used to unlock devices and enter apps is also compromised. The more realistic the image displayed, the better the devices process recognition and unlock computers and mobile phones and because of this, companies are investing in technologies that block these images, such as scanning through multiple 3D layers.
But this technology doesn’t only bring negative aspects to society. Although deepfake algorithms are widely used on the internet mostly as controversial videos and entering coded devices, the technology can also be employed to treat diseases such as depression.
Using someone else’s image, a patient who is not comfortable exposing his or her problem can report it to a psychologist or psychiatrist without identifying himself or herself. In addition, depression can also be recognized by AI from the person’s tone of voice.
Facebook has already banned deepfake videos, but there are still many other apps on the market that are spreading this technology faster than a horse in the Kentucky Derby race. How far can AI go in 2020 and how will society perceive it?
Written by: João Santos
2 comments
Unfortunately, fraud is increasingly spreading to the masses, it’s not a secret for anyone. The Internet is full of information about where to get fake documents and where to take id photo. Be vigilant
No wonder, the world is not standing still, more and more new inventions, now it’s easy to make a scannable id card idsbuddy and impersonate another person. Yes, of course, this is illegal and carries various risks. So whether you do it or not is your responsibility 🙂