I once read that whale's songs could be used to transmit a 3d image from one whale to another which doesn't sound so crazy considering that echolocation can activate the visual cortex.
I don't remember the article and anyway there wasn't much more detail but I always found this idea quite interesting. Could be that looking for words, sentences or grammar in a whale's song is a misguided and anthropocentric approach to the problem. They may instead have a visual language that just so happen to be transmitted by sound.
That is a fascinating idea. We do have a tendency (understandably) to try and understand the universe through our own lens. Our thinking is heavily tied to our sensory input, so it is challenging to imagine what having echolocation or magnetic senses might be like.
But there is no reason we should expect other species to communicate and think in the same way we do.
It’s hard even to imagine how other humans, from another time or culture think. While we have similar sensory apparatus we the world we experience is constructed within our own consciousness, with all kinds of emotions and memories mixed in with the senses. If you think about your own mammalian physiological and emotional .. affective experience .. how it feels to run fast, be surprised, see something you want to eat.. then I think we can start to appreciate what it is like to think like a whale or a bat. Even though their sensory input is vastly different to use, they still construct their world with the same kind of equipment that we do
It's hard even to imagine how other humans, from our time and culture think. I have no clue how people experience life with aphantasia or no inner monologue. Yet these people exist, right now, among us.
I think the 'no inner monologue' thing is overblown. I often don't have one as I go about my business because thinking in words would slow me down at ${task}. But if I'm driving or walking, then I will converse with myself at length. I'd be more likely to describe it as a dialogue though, because the way I perceive it, I'm talking to an imagined second party. It's still obviously me, but therf's a back-and-forth between different personalities.
They conceptualize the past and future exactly the same as you do; the past is something they remember (some of), the future is something they can only guess about. Their language talks about it differently, that's all.
Although it is not built into the grammar of our language I've come across a description very similar to this, by CS Lewis, describing his perception of time as that of a passenger on a train who's riding with their back to the engine.
analogies develop further from there, such as events being passed at the moment moving very quickly, and large in one's field of view, and to some extent one js able to see around them and order their importance by size and speed with respect to one's own position. But they're moving fast and it's hard to take in all the detail at once, and being so close the more proximal ones obscure your view of the ones further away though the latter may actually be in fact larger or more significant. Only as they all recede into the distance of the past does their relative significance with respect to each other become more evident.
> But there is no reason we should expect other species to communicate and think in the same way we do.
The reason is how far back their line diverged from ours. The nearer the point of divergence, the more reasonable it is to apply how things work for us.
I talk a Signed Language and I know how to give a 3d image, for example when asking for the location of the toilet, like this: go through this door, then you see the stairs to the left, go down there and then from your point of view return but just one leve lower and the door will be to the right (from your point of view of now).
When asking for directions I am sometimes frustrated by the inability of people to tell me detailed directions. I know their language has limits and they just can't.
Another example, when my boy got on a chairlift the first time, I could tell him in detail what will happen including the change in speed and could include even the typical rattling when taking off. He was then very confident in taking the chairlift.
But I have difficulty imaging how whales communicate 3d situations... Probably by imitating what they experience by sonar and simplifying to the essentials?
We convey meaning using words, but it doesn't have to be that way. They may have sounds that map directly into spacial descriptions in a way that light works for us.
Yeah, for whales by mimicking sonar (caveat emptor).
In Signed Language you can point into a 3D space. If you define a mapping of the world into the 3D space available to Signed Language then you have the world at your fingertips.
I love sign language for it's "pronoun" system and spatial features, but the reality is we could just a easily "annotate" spoken english with similar spatial context, sign language just does a good job of teaching all parties how to do that be default. By not teaching ASL to everybody who speaks english, we are really missing out on more bandwidth (for detail and redundancy) in communication.
That's really cool, I'm curious if other people that speak in ASL or other signed languages are good at that in general? I feel like even if I was fluent in sign language I wouldn't necessarily be as descriptive but perhaps I'm wrong there.
I remember commenting on an article related to that here so I went digging through my comment history and found this post about dolphin visual language: https://news.ycombinator.com/item?id=3314056
Given that that was all the way back in 2011 and didn't lead to any big world changing events I assume it was just cranks being cranky and a lot of people getting their hopes up.
There are visual thinkers among humans. Temple Grandin being a famous one. She's written about how she thinks in pictures and had to learn to translate that to words growing up. Also how she believes animals think in pictures. Although I wonder if smell might be more the case for some like dogs.
People without sight can also use echolocation. With the right head location sensors and a set of headphones you could describe objects to them with sound.
I heard the same about dolphins few years ago but alas I cannot source scientifical evidences yet.
I heard also that while the major part of our brain is dealing with semantics, dolphins brain is dealing with acoustics; some kind of study was sourced in the radiophonic show I was listening which was telling about the different eras of brain growth in the history of humans and dolphins, trying to find some comparison points to better explain how similar cerebral masses could be specialized into different operations and a lot of different things I don't remember.
The shrew like common mammalian ancestor was a night animal and could echolocate. The 3d perception could have been the driving force behind the evolution of the large brain.
I read this speculation about dolphins.
I'm oversimplifying here, but they use their nostrils independently to produce 2-channel sound and some measurements consistent with 3D transformations have been observed.
In the thread of this thought, to search for intelligent life off-planet sort of misses the point, doesn't it? Here we have an intelligent species with a common ancestor, which we may assume to be easier to communicate with than an extra-terrestrial being. And we have hardly begun to attempt to communicate with our earthly neighbor in a meaningful way, but we have projects probing the cosmos for signals from space.
Following that thought even further, if dolphins or whales were as intelligent as humans would they even be capable of developing the ability to make instruments that can transmit electromagnetic signals?
Given the limitations of their physical bodies, would they even be capable of developing any sophisticated instruments at all?
Point being, intelligence may not be the limiting factor for extraterrestrial communication at all.
I’m not quite sure what you’re asking. We were talking about searching for signs of extraterrestrial life by looking for electromagnetic signals that they may have sent through space.
Are you asking why I’d assume that they don’t have the biological ability to send electromagnetic communication signals?
If that’s not what you’re asking, I don’t understand how else they’d send signals through space without sophisticated instruments capable of transmitting powerful signals.
They are no more optimized for 3D images than taste or smells.
Try and describe your chair or oddly shaped rock as a 3D object without referring to it by classification and ask yourself how many people would have used the same sounds in the same order. There’s no direct mapping between 3D objects and the sounds used to describe them,
First, words aren’t really related to sound. You and I are using words without sound, are we not? Language is an encoding. One of the things it encodes is shape. And even if you wanted to describe a chair without using the word “chair” you could do so using a series of precise measurements. It doesn’t matter that people wouldn’t use exactly the same words to describe it. In the end, everyone would understand each other.
No, but they are related to a naturally occurring medium of language, of which there are only two notable ones: the human voice, and signing. Written languages are simply codified representations of these forms of natural language. (Note: Sign languages are distinct unrelated languages from the local spoken languages with their own grammar and everything, not just signed representations of them.)
By using written language, you are merely encoding a representation of a spoken phrase into a graphical representation. This is what truly makes writing separate from other commonplace but more abstract symbols, such as arrows, crosses, checkmarks, bathroom/restaurant/exit indicators, warning signs, etc, which convey an idea quite effectively but do not have a clear reliable decoding into natural language (i.e. if you ask 100 people to explain what they indicate, you won't be given the exact same sentence 100 times).
> By using written language, you are merely encoding a representation of a spoken phrase into a graphical representation.
I would argue the opposite. Language is an encoding of ideas. By speaking, we are merely using an auditory representation of a word and when we write we use a graphical representation of a word, but the word is entirely separate from both audible and graphical representation. And indeed, we see that we can create other representations of a word. We can create a mathematical representation, like in the case of LLMs, and use words in a way that is entirely separate from sound and visuals. As a species, we used sound to invent words, but the invention isn’t tied to sound in any way. If we, as a species, evolved further and lost our vocal cords and ears, we could still use words. We could also continue to use words if we lost our sight as well. Truly words are separate from our senses.
That’s not how language works, people still use sounds and gesticulations not just words. If we lost the ability to speak we might abandon words entirely for something else.
Buzz, growl, etc are obvious example where sound directly influenced what the word was and how it’s written. But the ability for kids to form specific sounds also influenced words like mom. The same is true of world complexity, natural languages always have some imperatives like stop and go that are short and simple rather than long complex utterances like sesquipedalian.
Further, comparisons between languages suggest that words evolved from sounds. “No” for example sounds similar in a shocking number of languages from Afro-Eurasia and the America’s despite very long term separation.
We might abandon words for something else, but we wouldn’t need to abandon words because they could continue to work without sounds. That’s because words are a totally distinct concept from sounds.
It’s super ironic to talk about how sounds and gestures are a part of language while using purely written communication to talk. We will never hear each other. You will never see my gestures. And yet…
It’s possible to communicate in words alone just as it’s possible to communicate non verbally, but misunderstanding increases. The written word is a poor subset of language not a 1:1 mapping of tone and cadence.
As to abandoning words, we might abandon them for something better. When people use a whiteboard they tend to draw a great deal rather than simply writing words down. If we eventually move to say brain machine interfaces we could be sending more nuanced thoughts than words allow. Subtle degrees of emotion and perception that the written word simply can’t convey. The old a picture is worth a 1,000 words but taken even further to the point where the use of words atrophy and eventually disappear.
Except you couldn’t come up with those measurements just by looking at a chair, and someone listing to your series of precise measurements is unlikely to understand what the object actually looks like. So it failed the basic goal of communicating the shape of something you’re looking at. Where the suggestion is whales might utilize similar bits of their brain that decode echolocation sounds into 3D images to similarly decode spoken sounds into 3D images.
Beyond that there’s a complex predefined encoding between human words and sounds and sounds back to words. There isn’t for your “series of precise measurements” so a dozen young English speakers at the same high school might all come up with wildly different schemes.
A subtle example of this is how is error correction mechanisms are automatically used. You can understand someone talking though some surprisingly serious distortions and missing content. Even written languages include quite a bit of redundancy. Remove every forth letter from a sentence and it’s generally surprisingly understandable. “I lo_e to _alk _y do_ in t_e mo_nin_.”
So your ad hock scheme could work if everything was perfect and someone write it all down, and then say turned those measurements into a sketch, but it’s all quite useless day to day. Thus people default to using general classifications and pointing at stuff.
A few words are related to sound: in English, "meow" sort of sounds like the sound cats make (although my cat is much more nuanced, he wants you to know!). "Woof" sort of sounds like some dogs' bark, "cock-a-doodle-do" has the same stress pattern as some roosters, and so forth.
Some languages tend to have more onomatopoeic sounds than English, and signs in sign language (especially "young" sign languages, meaning sign languages that have been codified recently) are probably even more onomatopoeic.
But in general, your point is true: the vast majority of words do not sound like the thing they represent. "Knock" (as in a knock on the door) doesn't sound anything like a knock.
If I said "blue chair" then wouldn't it likely activate the visual cortex in most humans as they relate it back to an image stored in their brain from past experience?
I am not. Not everyone subvocalizes, and for some people, like myself, an inner monologue narrating what you read is voluntary. I might choose to have the latter if I'm reading a book for enjoyment, but if I'm just reading comments online, I would prefer not to limit my reading speed by doing things like that.
That's how hearing people learn to read. But even then it does not follow that they continue to sound words out or play back the audio in their head after they become proficient. In fact, one of the hallmarks of a proficient reader vs a beginner is that the former no longer needs to sound words out. They can just look at a word and recognize it more or less immediately.
It's definitely not "generally", it's not even a supermajority.
> Some estimates suggest that as much as 50% or more of the population subvocalizes when reading, especially during their early years of reading development. However, with practice and improved reading skills, many individuals can reduce the extent of subvocalization and increase their reading speed.
To chime in, I don't, but yes, it's my understanding that for many people the "voice in their head" instead metaphorical at all.
I was always under the impression that the internal narration used in visual media (e.g. the Dexter opening sequence comes to mind [1]) was taking a dramatic license, but it's apparently some people's lived experience.
If you (think you can) make the case for 3D, then you can make the case for 4D, because we can talk about time--past, present, future, yesterday, five minutes ago...
That guy right here[1], states that we will start to decode animal languages within one or two years. There is a story in the video, where he describes he and his team recording whale's words, playing the recorded audio back to the whale, and the surprising reaction of the leviathan.
One pretty astounding observation he makes, is that we humans and the planet earth, thought to be the centre of the universe until the telescope came along. Now it will be a similar moment, in which we think that our language is at the centre of intelligence, not realizing that many animals communicate just like us, but in an incomprehensible way to us.
In the 1990s show seaQuest DSV, there was a "talking" dolphin that spoke to the humans via a computer interface that the boy genius character had created.
In one of the episodes, the ship's computers get wipe so the boy has to rebuild the dictionary of the translation system by showing the dolphin flash cards.
I remember thinking that was an interesting way to build a translation layer between two different species with a BIG assumption that the dolphin could understand the symbols on the flashcards.
A large part of the book is the human crew figuring out how to communicate with blind aliens who communicate with a sort of musical chirping. The humans eventually write a program to input English vocabulary and play a rudimentary form of the music.
Suppose some villain had an eventual need to be able to understand whales (or dolphins) communications, but has several years solve the problem. I'm talking about the kind of villain who does not care if what and how they do this is ethical so all that matters to them is whether or not it works.
Think someone like the High Evolutionary as portrayed in the third Guardians of the Galaxy movie, but limited to present day Earth technology and knowledge.
Suppose that villain took several human babies, gave them hearing aids or cochlear implants that pitch shift and compress whale/dolphin sounds down to where humans can hear them, and raised them in a mixed human/whale environment so that from their first weeks they were seeing whales and hearing while speech as much they were seeing humans and hearing human speech.
Would some of those human babies end up learning to understand whale the same way babies raised in bilingual extended families often end up learning both languages?
Social interaction plays a major role in language development. If you take away the reinforcement of everyone being really excited when the baby goes "da-da" instead of "gur-gur" you would lose the learning. It's also much more difficult for blind babies to learn to speak because they can't see the mouth shapes being made by their parents which is a deliberate training exercise. I don't think they'd learn to speak whale any more than kids learn to speak other sounds in their natural environment. IANAL (Linguist), IANAB.
> I don't think they'd learn to speak whale any more than kids learn to speak other sounds in their natural environment.
I'm pretty sure it wouldn't work either, but kids do learn the sounds of other animals and they love to reproduce them. Many kids can make dog or duck sounds before they can even speak like humans.
The problem is that of course, dogs, cats, ducks don't actually have a language, just a few different sounds, so that doesn't prove anything about kids' ability to understand a hypothetical actual animal language.
I'm quite unsure of myself here and might be mistaken, but there's also the chomskyist argument that much of the structure of human language is derived from the structure of our brain itself. Our brain is flexible enough to generate many types of human language, but different species could have entirely different structures that might be impossible to learn as a human. A machine translation layer -- LLMs or such -- could be designed around this flaw, and will likely be our best bet.
SETI is looking for technological aliens. Non-technological aliens would not send radio transmissions. Technological aliens must understand math a science to be technological.
That presumes that there is non-human sentience on Earth, which is not a given. Given that it's been impossible to teach apes anything like real language (they have been taught a few signed words), my guess is there are not any truly sentient non-human beings on this planet, where "sentient" means you can communicate with words and a grammar.
"If we meet extraterrestrials someday, how will we figure out what they're saying? We currently face this problem right here at home: we have 2 million species of animals on our planet... and we have no Google Translate for any of them. We’re not having conversations with (or listening to podcasts by) anyone but ourselves. Join Eagleman and his guest Aza Raskin to see the glimmer of a pathway that might get us to animal translation, and relatively soon."
I wonder what would happen next, if we could communicate with whales, and the language ended up having more than 25 words?
There are just north of 8 billion humans, and something like 850k sperm whales. If even a relatively small percentage of the former wants to talk to the latter, that's pretty overwhelming. Do we open an embassy or something?
This is most accurate top-level comment on the page. I once TAed an animal cognition class for Herb Terrace of Nim Chimpsky fame (or infamy), and I can say, we have yet to demonstrate any non-human language use.
Non-human communication is common, but language is different, and requires syntax. Syntax (plus vocabulary) is what enables you to meaningfully say a sentence you've never spoken before, and for another person to understand a sentence they've never heard before.
If you look at the corpuses of non-human communication, they're incredibly small compared to humans. A meercat has a few distinct vocalizations to warn their group depending on the type of a predator, but they can be counted on one hand. A songbird's song is ballistic once learned; it will never vary, even if they go deaf.
Cool finding. I'm more familiar with the primate literature, so I looked over the study and the book summary.
The Herman study was published in 1984, which was already a few years after the collapse of most of the primate language efforts. So, the author tries to compensate for many of the flaws that spelled doom in the primate teaching efforts.
In particular, they do a decent job of focusing on comprehension rather than language generation, and wisely test primarily novel sentences.
On the other hand, they make the same error as the primate researchers in not fully addressing if, and to what extent if so, operant conditioning could also be the method of learning. The response to novel sentences and the 5-word sentences seems suggestive of language, but conditioning can also do a LOT. (E.g., look up rats trained to do Rube Goldberg-style mazes involving dozens of steps to see how complicated behaviors can be achieved by conditioning smaller subpieces and then chaining them.)
The novel sentence tests might still be explainable by conditioning as: the dolphins are rewarded for correct answers, and by combining operant conditioning with object identification, are able to infer the correct answer. E.g., they trained a dolphin to answer yes/no on whether a hoop was in the pool, and then asked the same question for items used in earlier studies. This could still be a generalization of a conditioning response from a smart animal receiving rewards for correct answers, and not a demonstration of syntactical understanding.
>There is no reason to suspect that whales even use language.
Sure there is.
1. We've yet to rule it out in dolphins/orcas/whales etc like we have in lots of other animals. That alone is a pretty big deal. That's reason for that.
And there's research that show pretty conclusively that dolphins understand syntax. One has to wonder why have a brain that understands syntax but doesn't use it naturally.
"Communicate" is a vague word. I guess I could communicate with you by hitting you upside the head, which is way more than I can say for plants communicating with each other, and for that matter about as much as I can say for most animals communicating with others of their species.
That's partly the point. Communication is broad and can apply in a lot of ways, but language requires syntax. If it can be learned by conditioning, or inherited by instinct, it's not language.
If you listen to a flock of crows there is clearly some form of communication happening. It is amazing that nobody has figured out how to translate it. I wonder how many different languages or dialects crows have developed.
Crazy but intriguing stuff. Surprising (to me) thing to note - the first goal is to be able to construct "sentences" in Sperm Whalish not necessarily understand them. The next step is to figure out what they mean.
In context, W is talking about how we infer internal states based on external speech. So when he says "we could not understand him", he isn't saying that we couldn't understand a lion saying "the book is on the table" but rather that we could not infer what is going on "inside the lion's head".
>"I cannot know what is going on in him" is above all a picture. It is the convincing expression of a conviction. It does not give the reasons for the conviction. They are not readily accessible.
>If a lion could talk, we could not understand him.
>It is possible to imagine a guessing of intentions like the guessing of thoughts, but also a guessing of what someone is actually going to do. To say "He alone can know what he intends" is nonsense: to say "He alone can know what he will do", wrong. For the prediction contained in my expression of intention (for example "When it strikes five I am going home") need not come true, and someone else may know what will really happen.
For folks who haven't seen the Star Trek episode, there are a race of aliens who communicate using referents to metaphors or historical events. The translator works fine on a basic level, but without more context communication is impossible.
I remember reading some spurious response to a similar question about dolphins, and it speculated that all their vocalizations were simply variations on a single sentence like "hey buddy, did you check out that krill".
No. The one group the reporter embedded with studied sperm whales and after years of recording and study the researchers know that the whales use a series of clicks at each other and that different family groups use different series of clicks, and to human ears, it is very repetitive in all cases. It could depend upon almost anything else that could be relevant to cetaceans, like current direction or speaker depth. :) That research group has private funding for two more years. It was actually believed that sperm whales were silent until about seventy years ago.
We also know this is similar behavior (pod specific vocalizations) if not in sound or pitch to orcas that hunt whale calves.
Also, language is defined as something only humans can have, which is a bit circular and silly to anyone who’s observed cetaceans for any period of time.
Communication established: "sorry for the genocide, but we needed the candles and BTW, those stomache cramps come from microplastic. Oh and we caused a global warming, so.. Good luck?" if any sea life had a fully conscious society it would wage guerrilla war against humanity.
My question, why focus on "can we talk with whales" when we already have the problem of "can we talk with humans" that has a known right answer?
We can have a spanish speaker (or speakers) sit down, feed us with hours and hours of content (heck, that's already there) and then work with AI to try and create a black box interpreter that can turn spanish to english.
Has anyone done this? If the goal is universal translator why not start off with a simple case?
My worry with trying to do this first with whales or other animals is we don't know the answer. Nobody can look at the whale AI translation and say "You know what, this was a good translation". Sort of the "Mars attacks" problem.
All this said, I do recognize that this creates a "human bias". The way humans talk may not readily translate to animal speech (But perhaps it would for other simians?).
But the problem is not a black box. Presumably at least some (likely most) of whale conversation is things like “I see some fish to our left” where you can measure a result, ie does the pod go left. (Some whales perform highly complex coordinated hunting strategies, which you’d think includes some verbal coordination.)
Or at least, you shoot for that and maybe discover that 99% is philosophizing you can’t understand. But maybe you can bootstrap from present tense to suggest they are reminiscing about previous fish seasons. And so on.
Perhaps, though I could see there being issues like "I see fish on our left" "well, you're always wrong so we are going right" being an issue. There's even the issue with "shorthand" that might not be universalizable. For example, with a hunting strategy you could imagine "Execute the Janeway protocol" being an issue for translation.
You can see some of the hunting strategy problems play out with sheep dogs. A well trained sheep herding dog can execute really impressive actions based only on specific whistles trained by the handler. [1]. It'd be wrong to conclude a specific element of language for all dogs based on the whistles.
> Perhaps, though I could see there being issues like "I see fish on our left" "well, you're always wrong so we are going right" being an issue.
This isn't a black box because the whalesong encodes some type of behavior that we could, in theory, observe and decode even if we didn't understand the atomic pieces of the whalesong. It's how you study any unknown language--you learn a few relations and then gradually build up a richer lexicon and grammar that gives finer-grained understanding.
I'm not sure what the dog whistle example is intended to demonstrate, as even human language has a significant learned component.
I agree you could find higher-order concepts, but I don’t see any reason to assume they would be remotely as common as object-level world-descriptions.
When you are asking someone to “pass the spoon, it’s to the left of the cup”, you don’t dip into a codebook and say “play 22, hut” even if you are a quarterback.
The problem with this type of thinking is that it undervalues the role that non-vital communication has in social species. Look at uncontacted peoples around the world. They've had music and dance even though their tribes consisted of a couple hundred people. These forms of communication serve a role in their communities but it isn't "there are fish in the pond over there" it's more special and abstract.
If we are to treat whalesong the same way then we can't immediately assume that they're just trying to communicate base concepts.
In any given human language, at least, most vocabulary tends to be concrete objects or physical processes. The parent is suggesting that we begin by looking for that subset of the language by correlating the speech sounds of whales with their behavior to objects in their environment.
Not sure exactly what you're saying, but: Bible translators regularly translate the Bible into other languages, including the kind of "tribal" languages you refer to. The people they translate for are fine with saying "there are fish in the pond over there", and they can also talk about abstract concepts. They may not have vocabulary for all the same concepts we have, whether abstract ("justification") or concrete (the Inuit didn't have words for sheep), but they have plenty of words for things for which we don't have individual words, and those other concepts can be described in phrases. And of course they have grammars.
I think that if you were trying to decode one of those languages you would take the same approach and observe coordination behavior, “there is a deer to our left” -> tribe goes left.
You need to start somewhere. Maybe you would never be able to decode the non-functional communication without a Rosetta Stone or common language root.
The point is that at least some of language needs to cash out into a description of the physical world. I’d argue almost all of it, even in modern societies.
It's easy to start learning another language without an interpreter or dictionary, you just start with concrete objects and actions. The hard part is representing the sounds (phonemes), which are often quite different from the language the learner knows. It helps that there is a way of writing all the known sounds of all the studied languages (the IPA, which Wikipedia uses), and it's pretty rare these days that a "new" sound is found.
I also see a problem with "talking" to whales being: "Hello, I am human who learned your click-language." "I don't care one bit, now where are the squid..."
Basically imagine them like some remote people eking out a survival but ten times more incompatible.
Kinda. As I recall this happened by accident with French in (GPT-2? Not confident which LLM) — even though it wasn't meant to be in the training set, there were phrases just lifted directly from the other language.
"Hasta la vista" et cetera have a certain je ne sais quoi, but I suggest a better Gedankenexperiment would be the language of the people of the North Sentinel Island — with whom you cannot interact by both law and them being homicidally xenophobic, and will therefore be limited to remote sensing with parabolic or laser microphones only.
Ideally, you'd use a language with native speakers who can ultimately verify the translation efforts. Perhaps a language like Icelandic, Celtic, or Korean would be better. Languages with little cross over into english while also having accessible translators.
North Sentinel island would be a good stress test, but not a good first step as we can't ask them if we got the language right.
They are solving a different problem. They are doing the "white box" translation, that is knowing good inputs and outputs.
I'm talking about something more akin to what these researchers are doing. Without feeding the AI information about the correct translation, can we make an AI that can translate spanish to english? That is, what the researchers are trying to do with this whale translation.
You don't need Parallel corpora for all language pairs in a "predict the next token" LLM.
What I'm saying is that if an LLM is trained on English, French and Spanish and there is Eng to French data, you don't need Eng to Spa data to get Eng to Spa translations.
Spanish may have been a bad first language, but others like Korean, Icelandic, or even Russian would work as there is very little cross over or related languages.
You'd have to be careful with the input data, though, as it would be easy to corrupt your dataset with translations if you try and do this fast.
Those are trained differently, being exposed to translations in the training. My understanding of the person you are replying to is proposing "discovering the meaning" of Spanish by just using audio without any translations in the training. Pretending we don't know Spanish as if it were an animal/alien language.
I wonder how much animal body language could help with that. Maybe instead of trying to focus purely on the language (like with human language translation), the algorithm could try to observe and infer the meaning from more than just audio? Dogs were domesticated long ago and can communicate with humans, sometimes purely through "facial" expressions and body movement.
I'd think it would depend a lot on the animal. Whales, for example, don't really have great eyesight. They depend a lot more on sound and have been observed communicating over long ranges (particularly because water carries sound quiet well). So it seems natural to conclude that whales would communicate more primarily through sound than other mechanisms.
Well, if you spend enough time on duo lingo, and get past the "I fucking love the rrrrugby" stage, then you can totally talk to all sorts of people from whales in their native language.
Controversially my suspicion is what we're doing to the oceans and the rest of the planet will cause more suffering to humans than any other species, which will mostly just die out without really understanding what's happening to them. We all know what human suffering feels like, and we're all able to communicate with other humans one way or another, yet none of us seem able to accept we might have to change our lifestyles or reevaluate our priorities to help minimise that suffering in others.
Still, I'd be fascinated whether whales actually had any interesting (and comprehensible!) thoughts to share about human behaviour, and whether we might learn at least something from them.
Well, we can recognise danger calls in birds, usually repeated sharp "chip" sounds. Going beyond that we start assuming that some organised signals correspond to our ideas of communication. I tend to doubt it. If there is something more complex it may be very different to what we are used to.
You don't need Parallel corpora for all language pairs in a "predict the next token" LLM.
What I'm saying is that if an LLM is trained on English, French and Spanish and there is Eng to French data, you don't need Eng to Spa data to get Eng to Spa translations.
How would an LLM figure out what words to translate animal sounds to? Where does it learn that information? We don't know what animals are communicating if they do have a language of sorts. There's no mapping.
Potentially the same way it knows how to translate concepts with no mappings in that language pair in the dataset. Like i said, not every language in an LLM's corpus has something in another language to map to.
If you can find a way to tokenize whalesong then you could probably build a next-token predictor but we're still missing the Rosetta stone necessary to help us map their language to ours in vector space.
I wish we could talk to orcas and explain to them at which level of overkill we could retaliate and exterminate them if they don‘t stop attacking our boats.
If I would ever find myself in that situation (chance is slim but has increased to non-zero for me), I'd be tempted to pull out a skippy ball & see if they're up for a game of water polo.
I don't remember the article and anyway there wasn't much more detail but I always found this idea quite interesting. Could be that looking for words, sentences or grammar in a whale's song is a misguided and anthropocentric approach to the problem. They may instead have a visual language that just so happen to be transmitted by sound.
Like, do you see what I mean? But, literally.