Live or Memorex: AI & Photographic Truth
The tag line advertising and marketing jingle “Is it live or is it Memorex?” became popular in the early 1980s by the American company Memorex. It was used to describe the “live-like” audio quality of their cassette tapes that were used in 1970s & 1980s boom boxes. The image above never happened. Jean-Michel Basquiat was a pre-teen when Pablo Picasso died. The image made the rounds on social media this week, not long after the infamous Pope in a Balenciaga puffer coat.
I immediately knew the Pope puffer coat image was fake because of the hand and coffee cup. TIME magazine pointed out other telltale signs. I also knew it was Midjourney (version 5, no less). But a lot of people did not know. The meme reminded me of teaching Media Literacy classes at UMass Boston, which included a unit on photographic truth.
Photography struggles with truth as a concept. With other art forms, truth is generally a non-issue. We do not question whether a painting is real. We do not question whether a dance is real. We are generally able to discern fictional texts from nonfiction; furthermore, we’re generally able to sift through multiple nonfiction texts and combine them with our own experiences to arrive at a conclusion of truth. But not with photography. — Fstoppers
Photography has never been true and neither has other media (been true). We are currently in a “post-truth” era because the notion of truth in digital media can be debated, which is why Media Literacy (or Media Lit) is necessary. The unprecedented profusion of visual information across digital media contributes to the contemporary post-truth era, as well as the invention of ChatGPT and similar AI chatbots that have teachers across the globe checking and double-checking their students’ paper submissions.
A 2020 paper by Nataša Lacković introduces a new method to support “critical media literacy.” Lacković focuses on three aspects of images: “(1) the materiality of its representation and representational elements, (2) its object (what the sign refers to) and (3) its descriptive interpretations”:
For example, in Kendrick Lamar’s “The Heart Part 5,” his face continuously morphs from one maligned Black male celebrity to another, following the lyrics of Kendrick’s rap. This difference here from the previous Lincoln/Calhoun image is that we know the Kendrick “deepfake” images are just that. The deepfake images are iconographic, as well as representational. However, when the image is generated using artificial intelligence, it may not represent anything that exists in the physical world.
Realistic deepfake portraits are made possible by the creation of Generative Adversarial Networks or (GANs) that consist of two AI agents: One forges an image, the other tries to detect the fake. If the agent discovers a forgery, the forger AI adapts and improves. The “Pope in a puffer coat” image is almost a deepfake but there are signs to indicate that it is not real. The Pope, himself, is a real person but the image of him wearing an expensive designer coat is not real.
A few posts back, I wrote about my AI-generated images of twins and how one of them went viral on social media. Thousands of people thought the twins were real but, like with the Pope image, there are plenty of tell-tale signs to indicate they don’t exist. It was never my intention to fool people and I usually post captions about what AI tool I used to make the image (i.e., Midjourney). However, I found that many people ignore the captions. They willfully chose to create their own ideas and stories about the images.