From MIT Technology Review · · 1m 20s
Can we talk to animals?
Scientists are using AI to decode animal communication, but translation remains theoretical and full of pitfalls.
In brief
Scientists are using AI to decode animal communication, but translation remains theoretical and full of pitfalls. Researchers are using machine learning to interpret animal sounds, but two-way communication remains uncertain and accurate translation faces significant obstacles. Originally reported by MIT Technology…
Why decode animal talk
Understanding how animals communicate could reveal their inner lives and help humans recognize that animals are sentient creatures with thoughts, hopes and dreams.
“animals are sentient creatures who think, who have hopes and dreams of their own”
The obstacles to understanding
Researchers must match animal sounds to what is happening around them. Prairie dogs make different yips for coyotes versus hawks. Some species like deep-diving whales are hard to observe.
Beyond sound alone
Many animals communicate through visual cues, touch, pheromones or combinations of signals that humans may not perceive. Without a multispecies Rosetta stone, accurate translation is difficult.
New tools for listening
Over fifty years, devices like hydrophones and wearable biologgers have made it easier to observe animal communication. Hydrophones capture humpback whale song from the ocean floor.
AI starts to decode calls
The Earth Species Project, founded in two thousand eighteen, used machine learning to detect carrion crow vocalizations and determine which bird made which call. The technology spots patterns.
Translation still theoretical
ESP hopes its NatureLM model, trained on human speech, music and animal sounds, will interpret vocalizations. But Yossi Yovel warns AI may emphasize acoustic similarities and make mistakes.
In short
Researchers are using machine learning to interpret animal sounds, but two-way communication remains uncertain and accurate translation faces significant obstacles.
