4 ms·
"We’re still not able to break the whistles and burst pulses into meaningful units, the equivalents of words or phrases." Seems like something machine learning
by dharma1 11y ago
"We’re still not able to break the whistles and burst pulses into meaningful units, the equivalents of words or phrases."
Seems like something machine learning could help with.
Of course it may turn out it's impossible to translate between human and dolphin languages in a meaningful way.
https://www.newscientist.com/article/mg22129624-300-dolphin-whistle-instantly-translated-by-computer/ https://www.newscientist.com/article/mg22129624-300-dolphin-...
- spatulan 11y agoWe might be looking for sentences and words that just aren't there. Dolphins see with sound, so maybe all those clicks and whistles are their way of sketching images for each other.
- Hoasi 11y agoTrue, but it might just be even more interesting to decipher. Imagine you could rebuilt 3D visualisations of dolphin thoughts.
- dharma1 11y agois there enough data in the clicks and whistles to render an image?
- spatulan 11y agoThere probably isn't enough data to render a full image, but how many lines do we need to draw to give the impression of a person, or a tree? Surprisingly few. Couldn't the same apply to a sonar image? It's just an idea. I'm not saying its true, just that its an interesting possibility.
- thaumasiotes 11y ago> how many lines do we need to draw to give the impression of a person, or a tree? Context sensitive, but yes, surprisingly few. 大 is three lines. 木 is four.
- inimino 11y ago人 is just two.
- thaumasiotes 11y agoI'm aware, but you need special training before you can recognize that as a person. 木 is "wood" in the modern day; does 树 really make you think "that looks like a tree"?
- inimino 11y agoSo perhaps any visual representation of meaning would require shared context to interpret, and would then inevitably become symbolic rather than directly representational.
- thaumasiotes 11y agoI don't understand what you're trying to say here, but e.g. Michelangelo's David is widely recognized as representing a person, and no shared context is required for that.
- suvelx 11y agoPerhaps to test this, you could have two sets of identical statues. Two dolphins need to interact with the matching statues together to get food. If you followed the 'five monkeys and a ladder with a banana' experiment and found two dolphins who worked it out. Replace one with another. Presumably one will try and teach the other. With enough dolphins and statues, you might be able to use the similar methods you'd use to break simple cryptography to work out what part of the communication is about the statue. Or, you'll find out every dolphin talks to every other dolphin differently. And then and then the government bans dolphins out of fear of being unable to intercept dolphin based communications.
- jack9 11y ago> Presumably one will try and teach the other. IIRC this has already been tried and demonstrated to happen. This is the first thing that google came up with, but I'm sure Scientific American has also covered the teaching behavior. http://www.earthintransition.org/2011/09/dolphins-teach-each-other-new-trick/ http://www.earthintransition.org/2011/09/dolphins-teach-each...
- V-2 11y ago"If a lion could speak, we couldn't understand him" (Wittgenstein)
- dharma1 11y agoWittgenstein didn't know about LSTM or word2vec :)