5 ms·

I find it hard to believe that Zipf's law is at all meaningful for non-human species. We have evolved in a very specific way, such that we are limited in the ba

by btown 11y ago

I find it hard to believe that Zipf's law is at all meaningful for non-human species. We have evolved in a very specific way, such that we are limited in the bandwidth of our transmitted communications, as well as in our ability to memorize large vocabularies. There is no reason that this is true for other species, including dolphins; they may be able to distinguish and produce significantly more nuanced "words" (if that concept even exists); they may be born with, or be able to quickly learn and commit to memory, the "distance graph" between a huge number of distinct signals that represent nuanced concepts. Consider an organism that can quantize a dense signal, treat it as a hash table lookup, and reference a specific concept in their memory; these transmissions would look like pure noise to us, and any error correction might be written off as periodicity. We'd never know.

Houshalter 11y ago
Right. Any series of bits can carry information in some encoding. Testing if it matches Zipf's law is an interesting sign that it might be a human like language, not proof. It's certainly interesting that their language has statistical properties closer to human languages than that of other animals tested.
greeneggs 11y ago
A few people speculate that sperm whales might be transmitting sonar "images" in their communications [1]. At first it seems like a crazy idea, and maybe it is---but it isn't impossible since the whales perceive their environment via sonar. If we could directly and efficiently communicate synthetic images to each other, then "words" might never arise. [1] http://www.nytimes.com/interactive/2016/04/16/opinion/sunday/conversation-with-whales.html http://www.nytimes.com/interactive/2016/04/16/opinion/sunday...
- nine_k 11y ago
  Now we sort of can. See emoji, memes, etc.
  - huac 11y ago
    Well, or 'pictograph' alphabets (e.g. ancient Chinese)
- netcan 11y ago
  Very interesting thought, whether in turns out to be true or not. Seems plausible, at least. If not whales, maybe some other species. The existence of sonar as a highly evolved sensory experience already tells us about scale of possibility within the creative repertoire of biology. Our brain design is clearly influenced a lot by the sensory and communication tasks that are important to us. For example, human smell is something unimpressive, by mammalian standards. We have smell and it affects us occasionally. It says 'don't eat that!' or somemsuch, sometimes. We also have some pheromones that are a basic scent communication method. Many mammals make 90% of mating decisions based on smell. Some species can react to a 6 day old pee stain the way humans react to strippers. Some find food from long distances away. They use highly developed scent based communication to transmit information about themselves (mostly) to one another. Something can smell scary, in a "Shit, that gave me a fright!" kind of way. It's not a stretch to imagine they dream or think in smells the way we do in words, sort of. We communicate with sound in amazing ways. Right now we are communicating using sight (most of us) in a way that piggybacks on our brain's special design to understand words, which are sound based. This is fairly common in our biological branch, words... or so it seems. Anyway, the design is flexible to be used in various ways. You can learn language without having hearing. There is obviously a lot of potential for nature to be novel.
  - kaybe 11y ago
    To be fair, most people massively underuse their sense of smell. Feynman described being able to distinguish who touched long untouched books by smell - sounds a bit farfetched until you try, it's not that hard for most. Then again, it's also not very useful in today's society.
    - praveenperera 11y ago
      Sense of smell is very important when it comes to becoming a good cook.
- lawpoop 11y ago
  If they are projecting sonar images at each other, this would be much easier for us to determine than if they are using serial, syntactical language, right? They may not be projecting shapes of actual objects, but if they are projecting shapes at all, it would probably be easy to record and reconstruct them visually for us to see, no?
- lawpoop 11y ago
  If they are projecting sonar images at each other, this would be much easier for us to determine than if they are using serial, syntactical language, right? They may not be projecting shapes of actual objects, but if they are projecting shapes at all, it would probably be easy to record and reconstruct them visually for us to see, no? I'm thinking if they so use 3-D sonar communication, they probably have a vocabulary of geometric symbols. For instance, with our vocal apparatus, we can imitate sounds like "bark bark" or "buzz", but instead we say "dog" and "bee". This allows us to talk about things that don't make sound, including abstract ideas. Perhaps they have 3-D geometric symbols, analogous to our words. Instead of imitating the sonar reflection of a school of fish, they might just have a shape that means "school of fish".
  - pitchka 11y ago
    It is a question of how they encode the image. Just like our eyes filter out most of the things and the brain reconstructs the picture, they might send compressed images that are easily reconstructed by their brains, but it takes effort to do it with computers. Eyes are easier because they take the image and compress it (or maybe they work on compressed sensing :D). Sending images via sonar can be accomplished by a time sequence that might not be equivalent to the representation of image, but to the brains of a whale it can be reconstructed into a coherent one.
simonh 11y ago
While I agree we've no way to know if Zipf's law applies to alien firms of communication, it seems extremely unlikely that it would apply to random or natural signal profiles that were not communicating information. So failing to conform to Zipf's law doesn't prove a signal is not meaningful, but confirming to it is likely to be very strong evidence that a signal is a form of communication.
- 6502nerdface 11y ago
  > So failing to conform to Zipf's law doesn't prove a signal is not meaningful, but [conforming] to it is likely to be very strong evidence that a signal is a form of communication. Then again, Zipf's law is known to describe many non-communicative, "natural" distributions like the sizes of grains of sand, the sizes of asteroids, and the populations of cities (one of Herbert Zipf's original objects of study). And conversely, many levels of representation of human communication have non-Zipfian distributions, like phoneme distributions. (See my comment history for references.)
jack9 11y ago
> such that we are limited in the bandwidth of our transmitted communications Every "creature" will have to use energy. Conservation of the energy means a linear optimization to minimal bandwidth. The channels (chemical, sonar, infrared, etc) and patterns are the most varied parts as different conditions produce different optimal expression.
SilasX 11y ago
It's unlikely for evolution to produce such a communication mechanism as it involves non-local refactorings. Under that constraint, and assuming efficiency is selected for, they will reserve shorter symbols for more common concepts and end up with something like Zipf's Law (but again, not a global optimum). Now, if they were intelligent, they would (as we do) come up with highly compressed encoding schemes that would have no apparent patterns to an outsider, except perhaps control signals at the start and end. But if they wanted to be heard by aliens, they would avoid using obscuring mechanisms like compression.
- sdenton4 11y ago
  A big part of human communication, especially spoken communication, is redundant, to allow error correction. Repetition and redundant context clues allow us to reconstruct meaning from lossy signals; language is a funny balancing act between compression and interpretability.
- seanp2k2 11y ago
  What if they were smart enough to see that contacting aliens wasn't in their best interests? What if they understood that continually seeking to harness ever more energy was ultimately futile and just decided to enjoy their existence without disturbing others or being known to anyone else?
  - hutzlibu 11y ago
    Someone read Douglas Adams? (if not, you probably should) Anyway, I would argue, that wanting to allways remain in the same state of being, is not very smart, since the enviroment is going to change anyway ... so you got to evolve to adopt to new enviroment changes (ultimately the collapying of our sun)
VLM 11y ago
Theres an IEEE paper by H. Zhang from 2008 where the author was surprised random open source java source code followed zipf law. It would be interesting to combine that paper's idea of analyzing source code with your idea of peculiar modes of thought and see if zipf still holds under those peculiar mental models for large enough code samples. For example does intermediate OCR API code follow zipf the same way as numerical analysis code, or some weird string theory code? Going into the hard sci fi (or weird fantasy) arena of fictional writing, I think an elaborate yarn along the lines of ye olde fortune teller or program correctness prover or traveling salesman router, based on zipf and compression algos would make an entertaining story... "So back in the 22nd century when they proved 11 dimensional string theory was true because it was the only geometry where the DSL describing it followed zipf-law statistics, and everyone knows only real things that are true follow zipf therefore the simulation must be accurate..." I suppose the plot twist would be something along the lines of outright falsehood documents also follow zipf...
vlehto 11y ago
"quantize a dense signal, treat it as a hash table lookup, and reference a specific concept in their memory" Did you just describe a word in a foreign language? Let's take string of letters:" t e r v e - m a a n - a s u k k a a t " Let's assume that I can somehow quantize that signal into syllables: ter-ve-maan-a-suk-kaat. Then hash table lookup against vocabulary: terve = hello, maan = earths, asukkaat = inhabitants. Yes it looks like pure noise if you can't separate words from each other and message length is short compared to complexity of message. Also if you only see a pitch graph of me saying that stuff, it's even more indistinguishable of random noise. Exactly the same problems as in the article.
fit2rule 11y ago
>We have evolved in a very specific way, such that we are limited in the bandwidth of our transmitted communications, as well as in our ability to memorize large vocabularies. Maybe one of the reasons we can't quite bridge the communication gap and "do inter-species communication" is that we feel that we are above the target species in question, evolution-wise. Maybe this is a huge mistake to make to think that we have evolved to handle complexity, rather than devolved to avoid simplicity.. I don't know, but if I could talk to a dolphin about it, I really wonder what they'd say. (Seriously, I've thought about this often, since I came to understand Zipfs Law in the 80's and listened to too many Devo lyrics..) Its quite possible that we're surrounded by alien communication, but are too arrogant, as a species, to be able to recognize it. I'm fairly certain we have some significant evolutionary steps to take before we can 'work it out' ..

by btown 11y ago

Houshalter 11y ago

Right. Any series of bits can carry information in some encoding. Testing if it matches Zipf's law is an interesting sign that it might be a human like language, not proof. It's certainly interesting that their language has statistical properties closer to human languages than that of other animals tested.

greeneggs 11y ago

A few people speculate that sperm whales might be transmitting sonar "images" in their communications [1]. At first it seems like a crazy idea, and maybe it is---but it isn't impossible since the whales perceive their environment via sonar. If we could directly and efficiently communicate synthetic images to each other, then "words" might never arise. [1] http://www.nytimes.com/interactive/2016/04/16/opinion/sunday/conversation-with-whales.html http://www.nytimes.com/interactive/2016/04/16/opinion/sunday...

nine_k 11y ago
Now we sort of can. See emoji, memes, etc.
- huac 11y ago
  Well, or 'pictograph' alphabets (e.g. ancient Chinese)
netcan 11y ago
Very interesting thought, whether in turns out to be true or not. Seems plausible, at least. If not whales, maybe some other species. The existence of sonar as a highly evolved sensory experience already tells us about scale of possibility within the creative repertoire of biology. Our brain design is clearly influenced a lot by the sensory and communication tasks that are important to us. For example, human smell is something unimpressive, by mammalian standards. We have smell and it affects us occasionally. It says 'don't eat that!' or somemsuch, sometimes. We also have some pheromones that are a basic scent communication method. Many mammals make 90% of mating decisions based on smell. Some species can react to a 6 day old pee stain the way humans react to strippers. Some find food from long distances away. They use highly developed scent based communication to transmit information about themselves (mostly) to one another. Something can smell scary, in a "Shit, that gave me a fright!" kind of way. It's not a stretch to imagine they dream or think in smells the way we do in words, sort of. We communicate with sound in amazing ways. Right now we are communicating using sight (most of us) in a way that piggybacks on our brain's special design to understand words, which are sound based. This is fairly common in our biological branch, words... or so it seems. Anyway, the design is flexible to be used in various ways. You can learn language without having hearing. There is obviously a lot of potential for nature to be novel.
- kaybe 11y ago
  To be fair, most people massively underuse their sense of smell. Feynman described being able to distinguish who touched long untouched books by smell - sounds a bit farfetched until you try, it's not that hard for most. Then again, it's also not very useful in today's society.
  - praveenperera 11y ago
    Sense of smell is very important when it comes to becoming a good cook.
lawpoop 11y ago
If they are projecting sonar images at each other, this would be much easier for us to determine than if they are using serial, syntactical language, right? They may not be projecting shapes of actual objects, but if they are projecting shapes at all, it would probably be easy to record and reconstruct them visually for us to see, no?
lawpoop 11y ago
If they are projecting sonar images at each other, this would be much easier for us to determine than if they are using serial, syntactical language, right? They may not be projecting shapes of actual objects, but if they are projecting shapes at all, it would probably be easy to record and reconstruct them visually for us to see, no? I'm thinking if they so use 3-D sonar communication, they probably have a vocabulary of geometric symbols. For instance, with our vocal apparatus, we can imitate sounds like "bark bark" or "buzz", but instead we say "dog" and "bee". This allows us to talk about things that don't make sound, including abstract ideas. Perhaps they have 3-D geometric symbols, analogous to our words. Instead of imitating the sonar reflection of a school of fish, they might just have a shape that means "school of fish".
- pitchka 11y ago
  It is a question of how they encode the image. Just like our eyes filter out most of the things and the brain reconstructs the picture, they might send compressed images that are easily reconstructed by their brains, but it takes effort to do it with computers. Eyes are easier because they take the image and compress it (or maybe they work on compressed sensing :D). Sending images via sonar can be accomplished by a time sequence that might not be equivalent to the representation of image, but to the brains of a whale it can be reconstructed into a coherent one.

simonh 11y ago

While I agree we've no way to know if Zipf's law applies to alien firms of communication, it seems extremely unlikely that it would apply to random or natural signal profiles that were not communicating information. So failing to conform to Zipf's law doesn't prove a signal is not meaningful, but confirming to it is likely to be very strong evidence that a signal is a form of communication.

6502nerdface 11y ago
> So failing to conform to Zipf's law doesn't prove a signal is not meaningful, but [conforming] to it is likely to be very strong evidence that a signal is a form of communication. Then again, Zipf's law is known to describe many non-communicative, "natural" distributions like the sizes of grains of sand, the sizes of asteroids, and the populations of cities (one of Herbert Zipf's original objects of study). And conversely, many levels of representation of human communication have non-Zipfian distributions, like phoneme distributions. (See my comment history for references.)

jack9 11y ago

> such that we are limited in the bandwidth of our transmitted communications Every "creature" will have to use energy. Conservation of the energy means a linear optimization to minimal bandwidth. The channels (chemical, sonar, infrared, etc) and patterns are the most varied parts as different conditions produce different optimal expression.

SilasX 11y ago

It's unlikely for evolution to produce such a communication mechanism as it involves non-local refactorings. Under that constraint, and assuming efficiency is selected for, they will reserve shorter symbols for more common concepts and end up with something like Zipf's Law (but again, not a global optimum). Now, if they were intelligent, they would (as we do) come up with highly compressed encoding schemes that would have no apparent patterns to an outsider, except perhaps control signals at the start and end. But if they wanted to be heard by aliens, they would avoid using obscuring mechanisms like compression.

sdenton4 11y ago
A big part of human communication, especially spoken communication, is redundant, to allow error correction. Repetition and redundant context clues allow us to reconstruct meaning from lossy signals; language is a funny balancing act between compression and interpretability.
seanp2k2 11y ago
What if they were smart enough to see that contacting aliens wasn't in their best interests? What if they understood that continually seeking to harness ever more energy was ultimately futile and just decided to enjoy their existence without disturbing others or being known to anyone else?
- hutzlibu 11y ago
  Someone read Douglas Adams? (if not, you probably should) Anyway, I would argue, that wanting to allways remain in the same state of being, is not very smart, since the enviroment is going to change anyway ... so you got to evolve to adopt to new enviroment changes (ultimately the collapying of our sun)

VLM 11y ago

Theres an IEEE paper by H. Zhang from 2008 where the author was surprised random open source java source code followed zipf law. It would be interesting to combine that paper's idea of analyzing source code with your idea of peculiar modes of thought and see if zipf still holds under those peculiar mental models for large enough code samples. For example does intermediate OCR API code follow zipf the same way as numerical analysis code, or some weird string theory code? Going into the hard sci fi (or weird fantasy) arena of fictional writing, I think an elaborate yarn along the lines of ye olde fortune teller or program correctness prover or traveling salesman router, based on zipf and compression algos would make an entertaining story... "So back in the 22nd century when they proved 11 dimensional string theory was true because it was the only geometry where the DSL describing it followed zipf-law statistics, and everyone knows only real things that are true follow zipf therefore the simulation must be accurate..." I suppose the plot twist would be something along the lines of outright falsehood documents also follow zipf...

vlehto 11y ago

"quantize a dense signal, treat it as a hash table lookup, and reference a specific concept in their memory" Did you just describe a word in a foreign language? Let's take string of letters:" t e r v e - m a a n - a s u k k a a t " Let's assume that I can somehow quantize that signal into syllables: ter-ve-maan-a-suk-kaat. Then hash table lookup against vocabulary: terve = hello, maan = earths, asukkaat = inhabitants. Yes it looks like pure noise if you can't separate words from each other and message length is short compared to complexity of message. Also if you only see a pitch graph of me saying that stuff, it's even more indistinguishable of random noise. Exactly the same problems as in the article.

fit2rule 11y ago

>We have evolved in a very specific way, such that we are limited in the bandwidth of our transmitted communications, as well as in our ability to memorize large vocabularies. Maybe one of the reasons we can't quite bridge the communication gap and "do inter-species communication" is that we feel that we are above the target species in question, evolution-wise. Maybe this is a huge mistake to make to think that we have evolved to handle complexity, rather than devolved to avoid simplicity.. I don't know, but if I could talk to a dolphin about it, I really wonder what they'd say. (Seriously, I've thought about this often, since I came to understand Zipfs Law in the 80's and listened to too many Devo lyrics..) Its quite possible that we're surrounded by alien communication, but are too arrogant, as a species, to be able to recognize it. I'm fairly certain we have some significant evolutionary steps to take before we can 'work it out' ..