Losing touch with reality

As AI becomes increasingly powerful, what does it mean for the wildlife images we see?

- James Fair

COMPARE TIM FLACH’S WILDLIFE portraits with almost identical images created by artificial intelligence (AI) and you are aware of two things - they are remarkably lifelike but you can still quite easily spot the fake.

The AI colours of a snow leopard, for example, are a little too bright, the definition too sharp. The AI bald eagle stares back at the viewer with cartoonish malevolence - it appears that computers don’t do birds, or perhaps feathers, especially well at the moment. In short, AI is good - but not that good. Not yet, anyway.

Leaving aside the fact that Flach’s photos were ‘scraped’ from his website without payment to enable the AI program to learn (the question of copyright and AI is currently being debated in parliament), the big issue is: at what point does the machine become so good we can no longer tell one from the other? And what does that mean?

Flach is in no doubt about the dark shadows looming over his profession. “When the border between what we deem to have existed and not existed starts to dissolve, we have a truly existential problem because we don't know what to trust,” he says.

He adds that it’s only a matter of time before the AI image is indistinguishable from the real thing. Indeed, outside of wildlife photography, they already are. In 2023, a German artist called Boris Eldagsen topped a category at the Sony World Photography Awards with an image of two women he’d created entirely on a computer.

Eldagsen turned the prize down after admitting to the trick he’d pulled, but his ability to produce something that could fool photography experts suggests that Flach’s border between reality and make-believe has already melted away. This is clearly problematic for all forms of news-based photography, including wildlife photography.