Discovering AI generated species

When we think of endangered species, we think about rhinos, pandas, and mountain gorillas. We often forget that in addition to these well-known species, who people are raising a lot of funds for, there are unknown species going extinct without hardly anyone noticing. What if AI can help us be more aware of all animals currently going extinct?

Critically Extant aims to make people aware of these species by showing moving images on the billboard of Times Square. The images were generated by an AI model trained to produce images of the largely unknown animals

Until the end of May, every night a few minutes before midnight the artworks of more than 30 species – amphibians, birds, fish, plants, fungi, and insects – are shown on the screens surrounding the square. The audience is briefly submerged in a world unknown to most, as the videos and accompanying sounds are played. 

Some animals are shown differently from their real-life counterparts because they are based on open-source knowledge only – and that’s exactly the point. “A neural network can only create images based on what it’s seen in training data, and there’s very little information online about some of these critically endangered species,” says artist and Next Nature Fellow Sofia Crespo. 

Crespo and Feileacan McCormick (whose first name is, very appropriately, the Irish translation of butterfly) trained the AI model on animal appearance by a paired dataset of nearly 3 million images and text describing 10,000 species. The slightly off pictures of the species illustrate the limited amount of public knowledge on the endangered species. “If that’s all the data we have, and species go extinct, what kind of knowledge and imagination do we have about the world that was lost?” asks Crespo. Hopefully Critically Extant sparks the attention of enough people to save the endangered species before they disappear.