DIGITAL LIFE

The danger of 'deep fakes' 

In 2018, cinesta Jordan Peele published a video in which former US President Barack Obama says Donald Trump is “an imbecile”, among other barbarities in a pronouncement, which ends with the phrase “connect bitches”. Obama never said that, but the video uses a technology that allows another person to make a speech, in the case Peele himself, to appropriate Obama's image and voice to fool audiences.
The campaign of the time warned of the dangerous “deep fakes”, as these videos are called, something that began to be broadcast on the Reddit site in 2017, in which a user used the same technique to print the face of famous actresses such as Gal Gadot and Emma Watson in pornographic films. The technique consists of combining an extensive database and tools to connect the face and voice of people to existing videos, so that an extremely realistic content is produced. The tool's potential goes beyond unusual situations, of course.
There is a fear that deep fakes will be used to stage absurd political statements and influence electoral processes. Already preparing for this, Facebook announced in January of this year, probably with an eye on the U.S. presidential elections, that it will be on the list of banned content on the platform, along with nudity and hate speech. The strategy is not one of the most efficient, after all, Hollywood films use the technique to rejuvenate the faces of some actors, as has been seen recently in Irish or in Project Gemini. There are also the so-called “shallow fakes”, which use a much simpler or analog technique: to cut videos so that the statements are out of context.
"Cross a border. In a short time we will have videos and audios that the person supposedly caught will not have any way of proving that he did not do that,” says Eugenio Bucci, professor at the USP School of Communications and Arts. He explains that our society has agreed that documents such as photos and videos are reliable, but that deep fake can put that to the test.

More real than the truth
Bucci states that some obvious farts such as erotic mammaries in public schools or Pope Francis ' supposed support of Donald Trump in 2016 are easily dismantled, while a video is more complicated. Whoever appears in it will have to prove that he was not at the scene that was filmed, almost as a much more complex alibi. Today, producing deep fakes requires a robust processing plate for the videos to be processed realistically. There is the Chinese application Zao, which offers the technology with many limitations and only for photos taken from the user himself in pre-rendered scenarios, but it should not take long for anyone to produce faithful videos with their own cell phone. Hao Li, one of the pioneers of deep fakes, estimated in September of last year that “perfectly real " videos would be possible in just over six months.
Bruno Sartori, videomaker who works with the technology to make parodies, states that there are still some tricks to identify a deep fake. "As one changes only the face it usually gets a little more foggy than the rest of the image, or an avowed look in a strange direction,” he explains. In his videos, he always signals them with a hashtag so that there is no risk of them being confused with something real. Even so, he agrees with the other experts: “every month something is lost that would identify a fake video”. If even they can't tell the difference between the real and the manipulated videos, who's to say the ones who get a bombastic video from WhatsApp?

Politicians and lies
Bruno Sartori is videomaker and works with deep fakes technology, but only for parodies and humor videos. Some of his most famous, for example, are absurd, such as Jair Bolsonaro singing the song “the Tweed chick” or Lula singing “Obssessed” by Mariah Carey. He claims to have been sought by representatives of political parties to demoralize some figures, but that the requests were so brazen that he suspects strategies to sabotage the allegedly represented party.

Guilherme Sette-Brazil