What Does Generative AI Mean for Bird and Nature Photography?

What Does Generative AI Mean for Bird and Nature Photography?


To create the side-by-side images that accompany contributor Allen Murabayashi’s essay below, we asked the six photographers who won 2023 Audubon Photography Awards to describe their photos in a few sentences to someone who can’t see the image. With their permission, we fed their descriptions into a popular AI image generator. The results, shown alongside the originals, are based on this single prompt. — The Editors 

In 2012, footage of an endangered Bengal tiger marooned in a lifeboat captivated moviegoers. Ang Lee’s Life of Pi adaptation was clearly fiction, but many viewers didn’t realize that the majority of tiger shots were computer-generated. Hundreds of artists worked for years to create the cutting-edge visual effects.

A decade later, a photographer’s stunning images of an elusive snow leopard near Mount Everest went viral. When media covered the work uncritically, Alpine Mag’s experts revealed some of the images as composites—carefully stitched collages of preexisting photos rather than real moments.

Staged photos, composites, and jaw-dropping digital manipulation aren’t new to photography, especially where wildlife is concerned. Yet these illusions still took human labor and expertise to make convincing. In the past year, “generative” artificial intelligence (AI) technology has dramatically reduced the need for such effort. As a tech entrepreneur in the photo industry and former Audubon Photography Awards (APA) judge, I’ve been stunned at the rapid transformation.

Whatever you can explain in words, publicly available programs can conjure into a visual, whether a realistic image or fantastical artwork. Simply type a prompt, no matter how far-fetched—“snow leopard on Everest” or even “Ivory-billed Woodpecker in Central Park”—and software such as DALL-E 2, Stable Diffusion, and Midjourney will quickly render a synthetic image in a style or level of detail you specify. Video isn’t far behind.

These systems still have limits of verisimilitude, often producing uncanny and strange effects. To create pictures from words, AI models analyze and learn from millions or billions of captioned images. Some use open-source databases or photos scraped from the internet, while others aren’t transparent about source material. In any case, when these training data are sparse, biased, or insufficiently nuanced—as seems to be the case for many birds—results vary. In my experiments, Midjourney struggled to render the delicately curved beak of the ‘I‘iwi, a threatened honeycreeper in Hawai‘i. With each month, however, generative AI models are improving at creating images and making art, as well as writing articles, songs, recipes, and computer code. These giant steps are forcing many industries to grapple with existential crises.

In photography, seismic technological shifts have long precedent. In the early 2000s, for example, wildlife enthusiasts with DSLR cameras began selling quality images for pennies, upending the careers of full-time stock photographers. Today AI’s growing ability to generate realistic images seemingly threatens wider swaths of the profession. Last year’s Audubon Photography Awards grand prize winner, Jack Zhi, studied the behavior of White-tailed Kites for three years before capturing a perfect midair shot of a father teaching a fledgling to hunt. Now AI trained, in part, on images from photographers like Zhi might produce scenes of hard-to-capture behaviors—and a person scrolling on a phone may not know the difference. Even photo contest juries have already been fooled by AI-generated imagery, and current vetting mechanisms may be insufficient to detect the best attempts.

It’s not just photographers, but also conservationists who must contend with these developments. Photography has long been used to build wonderment of the natural world and to bolster arguments for protecting declining species, addressing habitat decline, and boosting public trust in the reality of climate change.

In the “fake news” era, however, generative AI makes it easier to sow doubt and spread disinformation designed to alter our beliefs and behavior. Ironically, these dynamics may also make it harder to trust remarkable yet real photos. Meme culture fueled by generative AI could further weaponize images by turning complex issues into punch lines. The tendency for generative AI to “hallucinate,” or confidently present a wrong answer, exacerbates these problems.

Even well-intentioned misuse could erode trust: Amnesty International recently faced criticism for using AI-generated images to depict a protest in Colombia—ostensibly protecting activists’ safety but risking the credibility of their cause.

While it’s easy to demonize a technology, AI is also a powerful tool for conservation. In the past decade, scientists have harnessed advances in AI to better protect wildlife. Automated machine-learning programs now comb through camera-trap, drone, and satellite images, as well as audio recordings, to monitor birds around the world, especially in remote areas that few people visit. Predictive models based on such data are helping to proactively combat threats such as poaching. Similarly, generative AI holds the potential to assist conservation causes by spurring innovation. Visuals in particular have the power to enhance our emotional connection to issues in ways words or data alone cannot; this ability is democratized as generative AI tools become available, extending human creativity.

For all these pros and cons, it’s clear that in the short term, AI’s rate of evolution is outpacing legal, ethical, and technological frameworks that might constrain its use and protect society from harm. We don’t want a system that is reliant on experts to detect hallucinations, or what’s real from fake, nor to have to fix a broken technology after it has inflicted harm. Researchers, policymakers, lawyers, and consumers need to seriously and quickly consider negative consequences as these tools proliferate.

But fear not! AI won’t replace photography. Just as I can hardly imagine families forgoing a holiday photo to render one instead, I doubt AI will end our drive to document everyday wildlife moments. Photos capture our experiences; generative AI captures our imagination.

My own experience has reinforced time and again that we can’t predict how technology will evolve, or how society will adopt it. For all the transformation AI may bring, I find it unlikely that it will turn human effort, expertise, and experience into quaint anachronisms. The joy of observing a bird and the effort to trek into the backcountry to capture an exquisite photo remind us of nature’s beauty and necessity. It’s up to humans, not AI, to act accordingly to preserve our world.

This story originally ran in the Summer 2023 issue as “These Birds Are Fakes.” To receive our print magazine, become a member by making a donation today.