Skip to main content
MANIFOLD
Will AI (or humans) decode and translate whale communication patterns to where humans can understand complex inter-whale discussions by the end of 2034?
30
Ṁ1kṀ1.8k
2034
73%
chance

Knowing that a sound means "I want to mate", or "there is food here", or "where is food", is not sufficient.

It would have to be like a mother telling a calf "This is where I met your dad" or a whale telling other whales, "WE will come back here again next year" or "Where is Wally" and another whale saying "Wally is dead".

This will also resolve yes if we are able to translate and understand what whales are saying, and can repeat things and monitor responses to varify accuracy, but if it turns out whales actually do NOT have complex language and are actually really stupid but make a lot of weird noises for other reasons (like communicating with specific whales or something but only using a few words/phrases in different ways). I would consider that "understanding complex inter-whale discussions". They would just not be as complex as we expect.

Whale huge brain
Get
Ṁ1,000
to start trading!
Sort by:

73% for understanding complex inter-whale discussions by 2034 is a long way ahead of the evidence. The best result to date is structural: Project CETI's Nature Communications paper (May 2024) showed that 8,700 codas from about 60 sperm whales vary combinatorially in rhythm, tempo, rubato and ornamentation, which the authors compare to a phonetic alphabet; it attaches no meaning to any coda, and nothing published since does. The description's bar is a mother telling a calf where she met its father, which needs displaced reference, a lexicon and a syntax, and no non-human communication system has been shown to have all three, in any species, after sixty years of work on apes, parrots and dolphins. Eight years is enough for the lexicon step, if there is one to find; the bar here needs the other two as well. The better-calibrated version of this question is the market on a two-way exchange of a simple instruction with dolphins, where the tooling exists (DolphinGemma, the CHAT whistle system) and only the content is missing. Here, something under 20% matches what the field has demonstrated.

I think they are at the morpheme level at least. So this should lead to more than mere recognition of integral messages.

sold Ṁ96 NO

Awesome! Thank you for sharing that

predictedNO