In Shark Bay, Australia, researchers continue to study the complex vocalizations of wild dolphins, yet experts caution against overestimating artificial intelligence (AI) as a tool for fully understanding or communicating with animals. Stephanie King, co-leader of the Shark Bay Dolphin Research project at the University of Bristol, has spent thousands of hours observing these mammals, emphasizing that while AI can aid data analysis, claims of decoding animal languages or “speaking dolphin” risk oversimplifying the challenge.

King argues that the widespread narrative positioning AI as a solution to unlock the “secret languages” of animals is misleading. She notes that extensive research over decades has yet to clarify the full meanings behind many dolphin sounds. Differences in sensory perception further complicate interpretation: dolphins perceive sound frequencies and process acoustic data in ways fundamentally different from humans. For instance, their echolocation is processed in brain regions linked to touch, suggesting a sensory experience distinct from human hearing or sight. King stresses the importance of understanding dolphins’ communication within their own social context, rather than attempting to force human-like interaction using technology.

Similarly, elephant expert Joyce Poole, known for her decades-long research on African savanna elephants, expresses skepticism about efforts to “talk” to animals. Poole maintains that the immediate goal should be to decode what elephants communicate to each other without interference. However, she acknowledges the potential societal benefits of deepening human understanding of elephant communication, viewing it as a pathway to improved conservation and greater respect for these species.

The promise of AI in animal communication research is also recognized by César Rodríguez-Garavito, founding director of the More Than Human Life program at New York University. He envisions AI helping to incorporate animal perspectives into legal and environmental decision-making, citing current global efforts to consider non-human voices in policy. Rodríguez-Garavito warns, however, that AI deployment must be carefully managed to avoid harmful interference with animal lives, cautioning against scenarios where technologies could be exploited for entertainment at animals’ expense.

Ethical concerns in this developing field have led to the release of guidelines by the More Than Human Life program to shape responsible use of communication technologies. Mickey Pardo, senior scientist at ElephantVoices, echoes this cautious optimism but notes that potential harms also exist in traditional research methods, such as playback experiments that may stress animals if applied without context.

Marine mammal researcher Luke Rendell of the University of St Andrews cautions that public expectations fueled by hyperbolic AI claims risk outpacing scientific realities. He highlights the risk of anthropocentric bias skewing efforts to decode animal communications and warns of potential damage to research credibility if promised breakthroughs fail to materialize.

King emphasizes that advancing the understanding of animal communication will depend on integrating machine learning with meticulous behavioral study in natural environments. She stresses that insights must be continually tested against observations of animals’ behavior and context, a process at odds with rapid technological deployment but critical for genuine progress.