In February, two strange ads started appearing on YouTube. Edited like a news program by an alleged media organization called House of News, the two videos show their respective news anchors Noah and Darren explaining how the Venezuelan economy “is not really destroyed.” It was shown that he was talking about something. Noah highlighted crowded beaches during the Carnival holidays, while Darren noted the revenue generated by the Caribbean Baseball Series earlier this year.
The video went viral on TikTok and was later aired on state news agency Venezolana de Television (VTV). What’s the twist? Neither Noah nor Darren are real. These are deepfake avatars made with artificial intelligence through his Synthesia, a platform that allows users to create videos similar to avatars other than Noah and Darren by simply registering.
The House of News video caused a stir on social media. After all, this appears to have been the first reported use of such artificial intelligence technology in the country. Of course, it won’t be the last.
After this ruckus, a government-led bot campaign, the banning of the Synthesia account used to create the ad, and its removal from Youtube, more avatars began to emerge. Just a few weeks later, Nicolas Maduro announced on his show Con Maduro+ the appearance of frequent AI presenter Sira. Propaganda Her site Venezuela News then introduced her two new announcers, named “Venezia” and “Simon.” All narrated for fake news reporting a vision of a crisis-free Venezuela.
The use of artificially generated news anchors is nothing new. Similar news presenters are beginning to emerge in Mexico, Peru, China and even Switzerland. AI is also currently being used in political activities. For example, the images in this ad for the Republican National Committee were created with AI.
The breakthrough here is that the emergence of these false anchors sheds light on an important fact. AI is going to change the battle against misinformation.
We need to understand what exactly these technologies are and what we can do to mitigate their impact.
Usage
This concern stems from a pre-existing technology, deep learning models, computational systems that can explore large datasets and detect patterns. We use them every day, and these models are behind social media algorithms and even built into virtual assistants like Siri and Alexa. But new developments in AI not only detect patterns in data, they replicate them in ways that are surprisingly human-like. That’s what’s behind software like OpenAI’s ChatGPT and Google’s Bard that have grown in popularity in recent months.
Dr. María Leonor Pacheco is a Visiting Assistant Professor of Computer Science at the University of Colorado Boulder, with a focus on natural language processing and machine learning.she explained Chronicles of Caracas How this technology works: “These are models that learn through multiple autocompletes.” I can. When presented with a “prompt”, a question the AI asks to initiate a search, the AI calculates possible responses and produces text that is consistent and human-like.
It looks something like this, but at a large scale it looks like this:
Images can be generated in the same way, says Dr. Pacheco. “Each image and pixel can be divided into numerical elements. The model finds patterns within those numerical elements and reproduces them. camberFor example, it has certain pixel characteristics. The model starts repeating them and learns to reproduce them. ”
It looks like this:
Synthesia uses audio and image data from real human actors. Users of all kinds and intents can also upload their avatars. Generative artificial intelligence models are becoming more and more accessible to everyone. The reason, according to Dr. Pacheco, is that the technology we all have access to is becoming more powerful. In some cases, it also facilitates parallelization of multiple simultaneous training rounds. ”So, with the ever-increasing computing power of smartphones and personal computers, it is relatively easy for anyone to log into and sign up for his ChatGPT, Midjourney, and other platforms, and then walk out of Caracas in the middle of a blizzard. You can create text and images that look realistic at first glance like this.Spread fake news such as Nebando of Caracas!:
In addition to creating fake news anchors, such technology is already affecting the disinformation landscape in Venezuela. Fact-checking organization Cazadres de Fakenews has already found instances of AI-generated images being circulated. For example, an image of what is believed to be a poster for an animated movie of Simón his Bolivar, which the graphic designer created with his AI, was taken out of context and spread like wildfire on social media. Another video, originally made for humor purposes, believed by several on WhatsApp and his Twitter to be true, shows “Skynet” developing Simon Bolivar’s android for the Venezuelan military. claimed.
According to Adrian González, director of Cazadres de Fake News, these technologies will enable the creation of disinformation content at scale. It then tells an artificial intelligence to “create 200,000 persuasive tweets that speak positively about importing spaghetti trees” and puts the output into an online botnet that looks very real. Set up a campaign, which is completely crazy and totally false. ”.
It’s that easy! In fact, this is what his seven tweets look like on that premise when we query ChatGPT, the popular large-scale language model.
“Artificial intelligence can and will be used extensively throughout the influence operation design process,” argues Gonzalez. They need to know exactly what the most compelling message is to get them to vote for their candidate. The AI is a tool that creates a specific kind of automated Twitter supporting his Alex Saab containing profile pictures of non-existent people, created using models like the popular site thispersondoesnotexist.com.
We have to prepare for this new situation, and for González it means going back to basics. much more sophisticated. The answer is to develop digital media literacy. This will not only help you learn how to use tools like reverse image lookups and fact-checking on Google, but also help prevent other digital threats such as phishing. ”
One might wonder how many people across the population of all ages and backgrounds would make the effort and invest the time to improve their media literacy, but in any case, we are all self-motivated. must play the role of This is not only about preparing for the changes that will come to disinformation, but also about starting to introduce changes to create positive impact. As well as being used to spread misinformation, it can also be used to share factual reports about what is happening in your country. That’s what his ProBox, a digital observatory focused on detecting disinformation campaigns on social media, has begun. They started implementing this with the artificial intelligence presenter Boti, just like Darren and Venezia. But instead of disinformation, Boti presents a presentation of the latest report.
In her video, she discusses important things to keep in mind as we enter this strange new era of communication. She said, “Neither technology nor programs like mine are evil. What matters is who programs us.”
