AI is Already Fighting Misinformation

Artificial intelligence (AI) is increasingly becoming a fixture in our society and is becoming more powerful by the minute. With recent advancements in AI it's important to ask, how will AI impact propaganda and misinformation? Will it increase or reduce propaganda and misinformation?

Many people are just assuming that AI will make propaganda and misinformation worse. This isn't a bad assumption but is overly simplistic because it overlooks that AI can be used as a tool to combat propaganda and misinformation. Bad actors will likely create propagandistic online articles, fake photos, and deepfakes, but as the technology advances AI models can be developed that detect propaganda and AI produced misinformation.

AI Produced Misinformation

While the threat of AI produced misinformation may be overstated, it is important to understand that AI will certainly increase misinformation. AI is now able to produce fake articles, fake audio, and fake video and these AI productions are getting more advanced as time goes on. With advancements in this technology, it's easy to see that AI misinformation could be used to sway elections and influence public opinion more broadly.

Fake videos developed with AI could portray various political leaders in precarious ways, leading to public mistrust of political leaders. Governments could also develop more propaganda. For instance, countries involved in war could develop deepfakes that portray their enemy as more evil than they actually are in order to rally support for the war effort.

Other forms of propaganda and misinformation are also likely to be produced by AI. This includes people using AI to develop realistic fake social media profiles that have the intention of swaying voters in elections. In fact, during 2022 more than two-thirds of the influence operations caught and removed by Meta (the corporate parent of Facebook) included fake faces that were likely produced by AI.

Using AI to Detect Misinformation

There's already research published on using AI to detect misinformation. For instance, research was published at the International Joint Conference on Artificial Intelligence that explores how AI can be used in fact checking. This research went into detail explaining how AI can assist fact checkers. It also explained that AI will likely only assist fact checkers and not complete the entire fact checking process because misinformation is often not simply incorrect, it often incorporates half-truths. These half-truths can be difficult for an AI model to categorize.

Other research looked more specifically at analyzing propaganda through text analysis. This analysis of propaganda with AI and other computational methods is a newer phenomenon due to datasets for propaganda text analysis not being available until more recently. While most research on detecting misinformation seems to have focused on developing models that analyze text, with advances in technology AI is being used to detect misinformation in other forms.

In 2022 Intel developed FakeCatcher, an AI system that is able to detect deepfake videos with 96% accuracy within milliseconds. The technology works by carefully assessing videos for what makes us human - blood. Our veins change color as our heart pumps blood. With this in mind, FakeCatcher's AI algorithm was designed to detect blood flow signals all over the face and translate these signals into spatiotemporal maps. Then, with the use of deep learning, FakeCatcher can detect whether a video is real or fake.

AI is also being used to detect audio deepfakes. Dessa, a startup that was acquired in 2020, developed an open-source AI system that detects deepfake audio. The AI detector works by using visual representations of audio clips called spectrograms. Spectrograms visualize audio in such a way that AI produced audio looks different than naturally produced audio. A deep neural network is then used to detect the authenticity of the audio.

Conclusion‌‌‌‌

AI will likely cause an increase in propaganda in all forms, but the future is not all doom and gloom. As time progresses, the previously mentioned misinformation detection technologies should become more advanced. However, the development and refinement of AI detection tools does depend on research funding and interest in the technology. It is safe to say, that without these tools, propaganda will become much worse. We need AI powered propaganda detection to stave off what seems like the inevitable, a world where the individual cannot breathe due to drowning in a sea of AI manufactured propaganda.

Written by Wes Cooper

2023-05-20