At a recent global conference empowering truth hosted by the International Center for Journalists (ICFJ), experts focused on combating deepfakes and other malicious AI tools.
Advice has been given on how journalists can better understand the threats that deepfake technologies can pose and what can be done to combat them.
Photo: IJN
Technology and its threats
Rapidly developing technologies are enabling users to edit facial features, create animated portraits, add motion, and replicate voices.
As part of this ecosystem, deepfake is a type of audiovisual manipulation that allows users to create realistic simulations of a person's face, voice, and actions.
Deepfakes are being produced more easily than ever before thanks to AI, not to mention very simply.
Fake videos of public figures are also becoming increasingly common, often accompanied by fabricated audio. Deepfakes are placing an additional burden on journalists and censors to verify the authenticity of a video.
They are the most widely discussed form of manipulated media, according to Shirin Anlen, a media technologist at WITNESS. “Deepfakes themselves are part of what we’re seeing more and more of in the news,” he said.
Although deepfakes are becoming more common, they also require a significant amount of skill and knowledge to create properly, making them difficult for the average person to produce. Therefore, many manipulated videos do not reach the level of a true deepfake.
For example, filters that alter a person's hair, eye color, or voice are manipulations we encounter daily, especially on social media. AI-generated dialogues and fabricated quotes from public figures are another example.
"Deepfakes aren't really being used on a large scale," Anlen said. "Most of what we're still seeing in the current misinformation landscape is low-quality fakes, mostly contextually reworked."
How to detect
Every new technology has flaws, and deepfakes are no exception. For example, users may detect errors such as choppy video, mouth movements not matching the audio, etc.
However, technology is also adapting very quickly. In an era where information is rampant on social media, timely detection and response are extremely difficult.
"The first generations of deepfakes were easily detected through eye movements. Now, newer generations have been updated; the eyes blink and are no longer still. This technology is constantly being updated and will become increasingly difficult to detect," Anlen noted.
Solutions
Among the methods for detecting deepfakes, journalists can examine video content for errors and distortions, apply existing verification and forensic techniques, and utilize AI-based methods for deepfake detection when available.
Enhancing media literacy tools and providing further training for journalists on media manipulation is also essential.
"We need to prepare now," Anlen said. "We need to understand the context in order to really shape the technology, to shape how it's built... so we're not passively affected by new technologies."
Hoang Ton (according to IJN)
Source
