Summary: Some of AI’s newest creations are inspired by the image of Doctor Dolittle, the childhood hero who can talk to animals. Researchers are using a branch of AI called “self-supervised learning”, an application of “deep learning”, to detect and monitor animal sounds in order to protect them. This approach will also potentially bridge the gap between human and nonhuman intelligences. Self-supervised learning doesn’t require data that have been labeled by humans. Instead, self-supervised systems “learn” from patterns inherent in data. The idea of learning from data makes self-supervised learning invaluable: Researchers often don’t know what animals are saying, so creating human-labeled data of animal speech might be impossible. With that being said, however, current AI-based systems focus on how creatures communicate in the wild, rather than trying to understand what’s on their minds. In the foreseeable future, we might expect to see newer kinds of AI “teach” us the inherent pattern in data rather than the other way around.