It’s the AI animal kingdom.
The internet is flooded with AI-generated wildlife clips, from bizarre interspecies play to animal trampoline sessions. Adorable as they may be, these highly convincing fake animal videos pose a danger to conservation efforts, according to a study published in the journal Conservation Biology.
“They reflect traits, behaviors, habitats, or relationships between species that do not exist in reality,” study author José Guerrero of the GESBIO group at the University of Córdoba, Spain, told Phys.org.
Researchers reportedly reached this conclusion by analyzing the large outpouring of AI-generated animals that have been circulating on social media. There are quite a lot of them.
In one super-viral clip with hundreds of millions of views, a group of rabbits are seen jumping in tandem on a trampoline.
Despite the hyperrealism, the static background and the fact that some of the bouncing bunnies disappear mid-hop fully prove that this clip is fagazi.
Another popular genre is one in which groups of unlikely animals, such as polar bears or cats, hang out together or even ride on each other’s backs, a more exotic “homecoming movie” kind of thing.
These anthropomorphic clips are not only completely ridiculous, Guerrero said, but also “show animals exhibiting human behavior that is far removed from reality.”
“For example, we see predators and prey playing,” he said. “Videos of children playing in the garden, videos of leopards, undermine the conservation of these species, because you would never encounter a leopard in those conditions.”
Research collaborator Rocio Serrano said this disconnect between humans and animals was “particularly pronounced among primary school children” and “demonstrates a lack of knowledge about the local fauna” among young people.
“Because endangered species appear more abundantly in these videos, these videos create a false connection with nature, which is negative for conservation,” the scientist added.
These videos can also create distorted expectations in children who travel to rural areas in the US or UK and expect to see capybaras or other exotic species that display “magical or charismatic” characteristics in the videos.
If they can’t find these creatures at home, they may try to buy them, fueling an already damaging “exotic pet trade,” study author Tamara Murillo wrote.
Additionally, cute AI-generated videos of people snuggling up close with wild cats or swimming on dolphins can induce “real-life imitation behavior,” animal advocacy group Wild Welfare wrote in September.
“A well-intentioned traveler may see a realistic AI video of a smiling elephant carrying its family through the jungle and assume this is natural and safe, without knowing that participating in such activities often involves corporal punishment, isolation, and hard labor for the animal’s lifetime,” they explained.
And this phenomenon is not only potentially harmful to animals. Jenny Vermilya, a sociologist who studies animal-human relationships at the University of Colorado Denver, warned that depictions of animals like bears as cute and cuddly are dangerous because they can encourage people to seek out animals in the wild, Atmos reported.
“I’m near the Rocky Mountains, and there’s a long history of people going into the mountains to interact with wildlife and being attacked,” she said.
To counter this harmful trend, the study authors suggested offering media literacy courses and introducing “environmental knowledge into school curricula” so that “children understand from an early age that there are no lions here.” [Spain]” recalls Francisco Sanchez.
Unfortunately, AI-generated images may pose a greater danger than misconceptions about megafauna.
This technique is also frequently used to impersonate real people.
OpenAI recently announced it would crack down on deepfake videos of Martin Luther King Jr. in its Sora 2 video tool after his family complained about “disrespectful portrayals” of the iconic leader.
