AI Upscaling May Explain Why Kamala Harris’ Real Campaign Photo Looks AI-Generated

🌈 Abstract

The article discusses the growing concern among Americans about the use of artificial intelligence (AI) in creating fake news stories and images, and how this is leading to a "Liar's dividend" where real images can be dismissed as fake due to the presence of AI. It focuses on a specific case where former President Donald Trump claimed a photograph of Kamala Harris and Tim Walz disembarking Air Force Two was AI-generated, and the author's analysis of the image and its provenance.

🙋 Q&A

[01] The Liar's Dividend and AI-Generated Content

1. What is the "Liar's dividend" and how does it relate to the presence of AI in our discourse? The "Liar's dividend" refers to the phenomenon where the very presence of AI in our discourse makes it easier for people to believe images are faked, even if they are real. This skepticism towards media content can be exploited by those who wish to dismiss or discredit information.

2. How can AI upscaling of images contribute to the perception of inauthenticity? AI upscaling, which uses artificial intelligence to increase the size and resolution of a photo, can sometimes leave artifacts or a certain "gloss" that makes the image appear hyperreal or slightly off, triggering the same uncanny feeling we get when seeing AI-generated art. This can make real images seem like they were created by AI, even though the original content was not.

3. What are some of the telltale signs that an image has been AI-upscaled? Potential signs of AI upscaling include overly smooth skin, brighter highlights, uncanny faces, and issues with details like hands and text on signs.

[02] The Kamala Harris Photograph Controversy

1. What were the key points of Trump's claim about the Kamala Harris photograph? Trump claimed the photograph of Kamala Harris and Tim Walz disembarking Air Force Two was AI-generated, despite the event having actually taken place as depicted.

2. How did the author analyze the provenance and different versions of the photograph? The author traced the provenance of the photograph, noting that the version initially shared by a Harris campaign staffer on X (formerly Twitter) appeared to be AI-upscaled, while the version later provided to the BBC by the campaign did not show the same signs of AI intervention. The author also compared the lighting and contrast between the different versions of the photograph.

3. What was the author's conclusion about the nature of the photograph? The author concluded that the version of the photograph shared by the Harris campaign staffer was likely AI-upscaled, even if the campaign was unaware of this, and that this contributed to Trump's mistaken belief that the photograph was entirely AI-generated.

4. What does the author suggest the Harris campaign should do to address the issue? The author suggests the Harris campaign should be transparent about the chain of possession of the photograph and any AI editing (even inadvertent) that occurred along the way, in order to build public trust and literacy around the use of AI in media.