Tech

AI search for patterns in dolphin sounds remains exploratory

The Wild Dolphin Project is working with Georgia Tech and Google DeepMind to analyse dolphin recordings, but researchers have yet to establish what the signals mean.

Editorial persona
Owen Mercer
Markets and Finance Editor
Published
Draft
Source: WIRED · View original source
A woman in glasses and a dark blazer gestures while holding a device at an event.
Marine research

Denise Herzing’s Wild Dolphin Project is using artificial intelligence to look for patterns in recordings of Atlantic spotted dolphins. The work, described by Herzing at WIRED World Fair, involves researchers from Georgia Tech and Google DeepMind.

Herzing began studying the dolphins in the Bahamas in 1985. Her project has since gathered extensive audio and video, which has supported research into dolphin behaviour and social hierarchies. The recordings may help researchers organise signals, but the source reports no reliable translation of dolphin communication.

Researchers have also tested an underwater wearable computer that plays dolphin sounds associated with objects and habitats. In one example, a sound linked to a scarf was played while a researcher handled one; another sound was associated with sargassum, a seaweed habitat. Researchers then observed the dolphins’ responses.

Those trials have not established whether dolphins understood the sounds or mimicked them. Herzing has said the distinction is difficult to make in the wild, and that interpreting signals requires understanding their meaning within dolphin culture.

For now, AI is a tool for pattern-finding. A two-way conversation remains a longer-term prospect, with the meaning of dolphin signals still largely unknown.

Continue reading

More from Tech

Read next: TSA may ask some overseas travellers to power on phones
Read next: Google adds confidential inference and high-density systems to TPU lineup
Read next: Mesh Wi-Fi extends coverage, but a standalone router may be enough