In an era where artificial intelligence (AI) is becoming increasingly adept at creating realistic content, it is crucial to be able to identify AI-generated misinformation and disinformation. Concerns have been raised by world leaders, with the World Economic Forum warning that AI-powered disinformation could disrupt electoral processes and lead to an explosion of falsified information. Hany Farid, a researcher at the University of California, Berkeley, emphasizes that AI-generated content can cast doubt on everything we read, see, and hear.
However, researchers like Negar Kamali at Northwestern University have identified strategies to help individuals reduce their risk of falling for AI-generated misinformation. When it comes to spotting fake AI images, sociocultural implausibilities, anatomical implausibilities, stylistic artifacts, functional implausibilities, and violations of physics are key indicators to look out for. By being aware of these telltale signs, individuals can become about 70% accurate at detecting fake AI images of people.
Video deepfakes, which involve digitally manipulating existing videos to swap faces or create new facial expressions, pose another challenge. Researchers at the Massachusetts Institute of Technology and Northwestern University suggest looking for inconsistencies in mouth and lip movements, anatomical glitches, face smoothness, lighting, hair movement, and blinking to identify AI-generated video deepfakes. Additionally, the rise of diffusion models has enabled the creation of completely AI-generated video clips, which often feature distorted faces or bizarre body movements.
AI bots on social media platforms have also become prevalent, with large language models allowing for the creation of AI-written content. Detecting AI bots can be challenging, but excessive use of emojis and hashtags, uncommon phrasing or analogies, repetition and structure, asking questions to reveal lack of knowledge, and being cautious with unverified accounts can help identify them.
Audio cloning, where AI tools can mimic anyone’s voice, has led to the rise of audio deepfake scams. Distinguishing between authentic human voices and AI-generated voices can be difficult, but comparing audio clips with previously authenticated ones, looking for inconsistencies, identifying awkward silences, and recognizing robotic speech patterns or verbosity can aid in detecting AI-generated audio deepfakes.
It is important to note that AI models are constantly improving, making it harder to distinguish AI-generated content from authentic human content. While individuals should develop their skills in identifying fake AI content, the responsibility to detect fakes cannot solely rest on them. Researchers like Hany Farid argue that government regulators must hold tech companies accountable for developing tools that flood the internet with AI-generated fake content.
