The problem will be when no one knows what is fake or real, says deepfaker – 02/28/2024 – #Hashtag

Deepfaker Bruno Sartori, 34, raised in Unaí (MG), went viral on social media in 2019, at the beginning of the Bolsonaro government, when he created a video in which the former president transformed himself into the Queen of England. From that moment on, memes entered a new era in Brazil, with a series of humorous videos by Sartori.

The productions in which he specializes consist of manipulated videos that use artificial intelligence (AI) and alter people’s faces in a practically imperceptible way.

One of the first deepfakes to spread, made by director Jordan Peele in 2018, simulated fake speeches by former American president Barack Obama. The production time for the pieces, however, was much slower and could take up to 30 days. Therefore, it was not something that was so widespread. Today, Sora, the “Video ChatGPT”, is capable of instantly creating material up to one minute long from text requests.

The accelerated advancement of technology can pose risks to elections, democracy, and even life in society.

O #Hashtag spoke with Sartori about recent advances in artificial intelligence and the direction deepfakes should take.

In 2019 you made some deepfakes that went viral on Twitter. Since then, manipulating and creating videos has become easier, with the advancement of artificial intelligence and platforms like Sora. Are we in a new era?
The codes responsible for generating synthetic content have advanced considerably since last year. The technological leap was significant and the tendency is for these advances to become more and more frequent. We are practically entering a new era every month.

Creating a deepfake today can even be done with a cell phone. The quality of video deepfakes, however, is still relatively low compared to those produced with professional software. However, this lack of quality can be beneficial for the use of the internet, as it helps to expose imperfections that could reveal manipulation.

Do you think that this year’s municipal elections in Brazil and the USA will suffer major interference from deepfakes? Can manipulations reach an uncontrollable degree?
There was a massification of technology. Codes have become more agile and today there is no need for large machines to create AI-generated media. These two factors contribute to more malicious people being able to produce misleading content. If there are more people creating, the risk to an electoral process undoubtedly increases.

For now, I don’t think we have lost control, but the authorities need to act quickly and force social networks to signal that that content was produced with artificial intelligence so that there is effective harm reduction.

Columnist Ronaldo Lemos recently wrote in Folha: “You can no longer trust video.” Do you agree?
The manipulations are still noticeable to professionals in the field, but how will other people suspect that what they are seeing is not real if most of them do not know that there are ways to create this manipulated content?

“Seeing is believing” has no longer been a guarantee of veracity since 2017, when the first code to generate deepfakes was released. For now, experts are able to identify this type of content. The biggest problem will be when the codes are so advanced that even these professionals cannot distinguish the fake from the real.

If we can no longer trust videos, what do we need to do to trust them again?
Forcing large platforms to clearly signal that that content is generated by AI. They have this capacity, they have already announced that they have these tools, but they do not do so because there is no pressure from the authorities. Identifying this content generates huge data processing costs and they want to avoid this as much as possible. I would not rule out the idea that they are even lobbying for this obligation to be postponed.

On the other hand, the government urgently needs to create educational campaigns to educate the population about the possibilities of creating artificial intelligence, as people will not be suspicious of realistic manipulations if they do not know that this is possible.

How do these advances affect your work? How do you continue specializing in artificial intelligence to continue being a reference in the area?
I work in several ways. When creating videos for the internet, for example, it is not just artificial intelligence that does the work. I have a script to ensure the content is funny and goes viral. There is context, jokes and theme associations. This is what sets me apart at this point, so much so that you don’t see other works like this circulating on the internet. When another deepfake video appears, it’s usually just a simple face swap, with no script to make sense of it.

About four years ago, I created a company that produces synthetic media to meet market demands. Clients look for me because they trust the quality of the product, since deepfakes produced by applications do not have the quality that large productions demand.

---

X-RAY | BRUNO SARTORI, 34

Journalist, born in Iturama (MG) and raised in Unaí (MG), left law school and left an internship to dedicate himself fully to the technological area. He currently lives in São Paulo, works with presenter Tatá Werneck and has a synthetic media production company, Sintetica.ai.

LINK PRESENT: Did you like this text? Subscribers can access five free accesses from any link per day. Just click the blue F below.