AI vs. Film: The credibility connection that could save photojournalism 

As AI image generators emerge online like dandelions after an April shower, at once delighting with vivid colors and delicately magical abilities to react to the barest suggestion of a breeze, a return to film-based photography offers a silver halide-infused paper trail that would allow photojournalists to prove authenticity in their work and bolster flagging trust in journalism.

The Imminent Deep Fake Tsunami

ChatGPT is an AI engine that uses predictive pattern recognition to generate humanlike responses to questions typed into a chat field. Basically, scientists took a snapshot of every available written document on the internet — from Facebook comments to Wikipedia — and fed them to a sophisticated computer program and told the AI: “This is what humanlike conversation looks like. Learn that and respond similarly.” In similar fashion, programmers downloaded and categorized every image on the Internet (from everybody’s cute pet photos to professional stock images) and fed them into AI: “Here’s everything in the world and what we call it.” So now when we go to an AI image generator like Stable Diffusion and type “parachuting cat”, the AI looks to its vast library of cat photos and snowboarding images, does some hocus-pocus, and makes something like this:

At this stage of AI’s development, results can vary from funny to astonishing. Newer tools are beginning to offer capabilities that can be used to create very specific, realistic, and easily customizable generated images, again based on the reality that AI knows what everything looks like from every possible angle and is getting better at interpreting what we’re asking of it. These features will be a boon to advertisers seeking to create a scenario where cute pets look fondly at their bag of kibble, but from there it’s a slippery slope to a conspiracist generating an image of the US president convincingly planting a kiss on the cheek of ______ (insert an image of that US president’s arch nemesis here).  

Photojournalists returning to film would not solve the deep fake problem, but it would make creating passable fakes more difficult. 

It is true that film images are not entirely fake-proof — they are  digitized before they go to press — but if publications provided access to high resolution versions of the images they print, authenticators would have the ability to examine and validate them. Images developed chemically from old school negatives exhibit obvious chemical fingerprints when viewed at high magnification that are not present in digital images. The silver halide crystals comprising film-based photos bind to each other to create a “grain,” not unlike wood grain, and when that grain  is manipulated it’s easy to spot. 

That said, photo manipulation has been with us almost as long as film.  

Hollywood has been able to transport us to galaxies far, far away and timelines past and back-to-the-future with realism increasing in leaps and bounds, originally by capturing live actors on film and then employing painstaking, frame-by-frame manual post-production editing to create the desired special effects. Modern fakers could digitally render the aforementioned fake image of the presidential smooch digitally, photograph that image with a film camera, and from there create a negative with all the requisite chemical realism (and political uproar) intact, so it’s not a perfect solution. It would, however, insert several procedural steps of time and money-consuming complication in order to craft an image that would pass a basic validation test. 

Journalism: Seeking attention or distributing information?

The bigger problem with today’s mistrust of journalism is that we in the audience have been herded away from thinking critically and toward the troughs of optimized narratives that algorithmically align with our existing worldview, frequently at the expense of objective truth. 

Until we the people in the audience decide that what we need from our Fourth Estate are facts, and that our pursuit of those facts requires examining sources for layers of bias, opinion, hidden motivation, and corruption, we are vulnerable to being tricked. 

This larger problem is a function of education, and national momentum toward rebuilding critical thinking skills is currently lacking in the United States, but elsewhere initiatives that strive to educate citizens on how to adapt to a digitally connected world are winning even while under attack. 

The key is not to try and stop the new tools from emerging — that’s impossible — but in broadening and deepening our collective understanding of the technology as it emerges and creating new ways of governing and integrating it into our world in the healthiest way possible.