A small mouse stands on two paws as it lifts its nose to sniff the air; colorful fish with shifting patterns drift underwater; flowers bloom and sway on tall, delicate stems. The painterly flora and fauna of Sofia Crespo's Critically Extant are based on real, critically endangered species like the Perote Deer Mouse, the Mekong Giant Catfish, or Parakaempferia synantha (a plant in the ginger family). Crespo trained AI algorithms on millions of open source images of approximately ten thousand species, using the resulting models to generate visual representations of animals and plants that -- despite being endangered -- have little or no presence on social media or in the broader public discourse. The generated representations appear slightly uncanny in form, in part due to the limitations of AI, but primarily due to the lack of man-made input available. How can an algorithm perfectly recreate an animal when it has too few examples of what that animal looks like?