Scientists Find an 'Alphabet' in Whale Songs
Researchers analyzing sperm whale clicks report evidence of a “phonetic alphabet” enriched by rhythm, tempo and ornamentation, suggesting a far more complex communication system than previously documented. Commenters explore how machine learning and large language models might help decode such non-human languages, while debating the limits of unsupervised translation, the difference between decoding and truly understanding meaning, and what advanced whale communication would imply for animal rights and conservation.
Access to the research
- Commenters share the open-access paper (Nature Communications) and Project CETI’s own blog post.
- Various workarounds for the NYT paywall are discussed (archive pages, text-only mirrors, Tor onion site).
What the study suggests about whale “alphabet”
- Sperm whale clicks form discrete clusters akin to an alphabet or phoneme set.
- New work adds tempo, rhythm, rubato, and ornamental clicks, massively expanding expressive capacity.
- Some note this looks more like a rich prosodic/musical system than a simple alphabet.
Can LLMs decode whale communication?
- Enthusiasm: tokenize whale sounds, train generative models, then use LLM-style methods to find patterns.
- Proposed approaches: unsupervised or weakly supervised models over large audio corpora; mapping latent spaces; correlating with rich behavioral/contextual data (tags, video, environment).
- Skepticism: current unsupervised MT still relies on human-language parallels and existing translations; no “universal translator.”
- Concern that we might only generate plausible whale-like sequences without knowing their meaning.
Decoding vs understanding
- Repeated distinction between statistically decoding structure and actually grasping semantic content or “what it’s like” to be a whale.
- Philosophical references (lion thought experiment, Chinese Room) used to argue that even perfect mimicry might lack genuine understanding.
- Others counter that human “understanding” may itself be just complex neural computation.
Is it really “language”?
- Some say science has been reluctant to call non-human systems “language,” often reserving that term for humans.
- Others insist “language” requires evidence of both semantics and syntax, not just complex signals.
- Examples like prairie dogs, great apes, birdsong, and even fungi are raised as borderline or candidate cases.
- Debate over whether scientific caution is appropriate or reflects anthropocentric and ethical discomfort.
Ethics, intelligence, and rights
- Several argue cetaceans are likely highly intelligent, with complex culture, and that this raises uncomfortable questions about whaling, bycatch, and factory farming.
- Discussion on whether advanced communication should lead to person-like rights, contrasted with humanity’s poor record on human rights.
Interspecies communication potential
- Many express excitement about multi-species communication as an “alien language” problem on Earth.
- Speculation on whale concepts: environment, navigation, social relations, stories, gossip, possibly even fine-grained descriptions of ocean conditions.
- Project CETI is cited as a major effort aiming at active two-way experiments, not just passive recording.