# How would people with perfect memory design their language

A race of people was created through magic, their mind was greatly altered, among the abilities that were granted was a memory that perfectly records all sensory information forever. As well as the ability to run programs in a sort of secondary mechanical mind, that operates on a much faster magical substrate.
This allows them to do things like instantly look at something like a QR code and know what's encoded (provided they have the mental program that reads QR codes).

Setting is roughly your standard D&D esque fantasy (with this group's abilities making them nearly the only people capable of being general mages) but with some basic mechanical innovation like typewritters and the printing press. The world runs on magic, which makes most technology impossible, chemistry just doesn't work and the world resembles the ways people thought the world worked pre enlightenment.
For these people there is no reason not to have a language as complex as possible when communicating with others sharing their abilities. They will all share the same same language since they can learn it in hours or days and tend to greatly value efficiency.

So with that in mind what would you would expect from their language:
I'm looking for the rough analogs for their spoken and written languages. The only limit to the sounds they can distinguish are the physical features of their ears, so you could pack in some massive bandwidth here. It's also important to note that their mental programs can produce sounds automatically, so no sound is more unpleasant to produce than any other, the only limits are what sounds they can physically produce. I also think they'd want to put in some physical motions here, this wouldn't just have to be stuff like sign language they could assign meaning to slight twitches in facial muscles or blinks.
As for their analog of written language, it's important to mention that this race almost universally has slight hand tremors, so that would make drawing fine detail quite hard (so hand drawn QR codes are right out). The only fundamental limit to how quickly they can read is how quickly they can move their eyes. This language doesn't have to be done with a pen, what I'm looking for here is something they can put on paper or something similar quickly. If your solution involves a very simple cheap machine or say the use of a knife to get straight lines that's fine. You might also have another type of writting that would be used for books, where the difficulty of transcribing by hand wouldn't be a factor.

Reminder: There's no reason to have this language resemble any language ever previously devised, in fact it seems really unlikely it would. Given how easy learning a language is for these people, there's no reason they wouldn't adopt whatever written and spoken systems were invented by linguists to maximize information bandwidth.
Given the much higher bandwidth of this language it really ought to be basically indecipherable to vanilla humans, if they can understand it after all then it must be relatively simple, which there's no reason this language would be.

• Upon thinking about it, I realized that this question is basically the same as the question "how would human level (but no higher) AI trapped in a human bodies communicate with other similar AI's" which I find kind of funny. Had I realized this earlier it I would have put a note about it in the question. Luckily however plenty of people seem to have basically realized many of the implications on their own. – Vakus Drake Sep 5 '16 at 4:58
• "shaka, when the walls fell"... all you need is couple (hundred?) years of history and you can describe a lot of stuff by referencing stuff that others know about. This does covers only part of language though, you might want to combine it with other ideas – Lope Sep 5 '16 at 5:59
• Don't overlook the fact sound is a lossy medium -- you need the spoken language to have pretty solid error correction. – Hurkyl Sep 5 '16 at 11:52
• @Azor-Ahai If you know a bit about linguistic it's kind of obvious. Most languages only use at most a tiny portion of all the sounds humans can produce. Plus there's tonal languages. If humans were more like machines and could easily recognize and produce any sound a human was capable of making then the bandwidth we could produce in our language would be many orders of magnitude larger. Imagine if every one of of thousands of noises (plus each of the dozens of tones it can have) humans can make had a unique meaning by itself and each combination also had a unique meaning. – Vakus Drake Sep 5 '16 at 20:21
• @Lope: Bah, beat me to it. Er, I mean, "Lope, 14 hours ago". – Lightness Races in Orbit Sep 5 '16 at 20:42

I agree with Daniel M's position that the language would be highly contextual. The more shared history one has with your conversation partner, the more efficiently you can communicate.

As a particular linguistic approach, I'd think they might use something along the lines of Huffman Coding. This is one approach used in .zip files to maximize compression. In Huffman Coding, one assigns representations to each symbol (a concept or idea in our fictitious language. A symbol is a 4 input bytes in .zip compression). The symbols which occur most often are given shorter representations than those which occur more rarely.

This sort of approach can also be done adaptively. You can start with a notional "base probability" tree, and as you communicate, you adjust the tree based on what has been said. If you're talking about the weather, you'd soon find words like "rain" and "temperature" are easier and easier to say, as they move up the tree to more advantageous locations. This is terribly hard to do in the human mind, but if one had perfect memory, one could easily keep track of trivial details like this.

As an interesting bit of flavor, you could add in some analog color to the language. What I described above is very discrete. It's very good at describing written text, but not so good with the subtleties of spoken language like intonation. The intonation would need to be layered on top. However, this could be troublesome with adaptive Huffman trees. If you've been talking about the weather too long, the word for "weather" may become very short. It could become too short to put much meaningful intonation into.

You could create two forms of language. The normal communication might be hyper-compressed. When they mode-switch, they might use a more lyrical communication method with an intentionally inefficient Huffman tree to give the more opportunity to embellish. A similar structure might appear for written language. Consider these two examples:

The first is a QR code, which is highly compressed, and very effective at communicating information. The second is a caligrapher's rendition of Chi, the Chinese concept of life energy. While obviously it does not pack quite as many bits of data, it still seems to capture something very expressive. It's simply expressive in a different way.

• I agree that some form of compression would be useful for transmitting the written form, but the word itself and its definition carries much more weight in terms of how much information you can pack into a given bandwidth signal. The definition, interpretation and historical information about a word ( required information for understanding and using the word correctly, especially if used contextually or subtly ) are information that is contained in the word and need not be transmitted through the signal at all. It is the word itself that carries all that information through. – Nolo Sep 5 '16 at 8:06
• One particularly interesting consequence of this is that Huffman Coding would be a pretty solid reason why dialects (or even different languages!) would emerge! Different groups of people would find themselves talking about different sorts of things more frequently, and would optimize language to suit their needs. – Hurkyl Sep 5 '16 at 11:55
• What you are forgetting is that Huffman trees are made for binary data. A language transmitted by sound is composed from phonemes, which already contain intonation. Also consider that in most natural languages the words that are the most often used tend to be the shortest due to language development. Let's say we have about 30 phonemes, then most words will be composed from about 1-4 phonemes which is not very different from natural languages. Crazy, isn't it? If you have advanced ears and speech, you can get to thousands of phonemes and then 2 phonemes per word is more than enough. – Sulthan Sep 5 '16 at 18:11
• @Sulthan I think even that's a bit conservative, by varying relative volume, and tone (see tonal languages) you could easily have tens of thousands of morphemes. You could express any word (not counting synonyms) or commonly used phrase with a single sound. Hell once you start using multiple sounds along with body movements and compression... I really underestimated how insanely awesome this question would be going in. – Vakus Drake Sep 6 '16 at 0:29

Just use a 64 bit numeral with 1 bit used for "not" to represent every word/concept that you can think of. Use a few bits for basic grammar and you have a perfect digital language. All you have to do from there to make it auditory is make each bit or set of bits have a phoneme/tone/etc to it and you have the most comprehensive translatable auditory language possible... nearly impossible to speak, listen to, or make sound nice, but that's how you would do it for shear quality/quantity of data.

I also have to point out how they achieved perfect memory would effect this. We use various methods to reinforce our memory, such as the way it sounds. The standard is using "baba" and "kiki" which we perceive the former as rounder and the latter as sharper which helps us remember "ball" and "Knife" (The k used to not be silent). If they rely on this what I suggest would have to be better organized where as if it's not based on any such thing they could be as abstract as they wish.

• You actually speak about designing the language like the question header asks. – Sarfaraaz Sep 5 '16 at 6:11
• Yeah, I came up with the idea a while ago wondering why it hadn't been done yet or at least tried. Just get a ton of linguists and people who put together dictionaries and start assigning every concept a numerical value and then you can just an array for the language and then you'd have a near perfect universal translator without adressing issues like grammar or metaphors that we use, but if you got into that you could improve the accuracy & quality even more. Google Translate does this without the initial numerical values, & offsetting the linguist work to normal people with limited languages – Durakken Sep 5 '16 at 6:25
• Note that since these beings probably are capable of forming more than sound or no sound, as well as the fact that it's quite difficult to judge length of long runs of identical bits (this is why computer hardware interfaces use things like 8b10b encoding), the actual sounds would probably be more like trinary than binary at least, and possibly quadratury (is that a word?). If you have two sounds for "on" and two sounds for "off", you can alternate between them, making it trivial to tell the parts apart. That also helps to tell the difference between "telling me something" and "being quiet". – a CVn Sep 5 '16 at 9:51
• @Durakken People who put together dictionaries are called lexicographers and "Just ... start assigning every concept a numerical value" is a vast simplification. What do you do with inflectional morphemes, for one? You can't have a "near perfect universal translator" without "issues like grammar or metaphor," unless perfect translation to you is "DOG WALK OWNER PAST." – Azor Ahai Sep 5 '16 at 19:26
• @MichaelKjörling "Quaternary" would be the word for Base 4. – Lepidolite Mica Sep 6 '16 at 4:16

I concur with Cort Ammon's opinion that dynamic compression would be a core aspect of their language. I add to his ideas the thought that the efficiency emphasized by this people would also have effects on the verbal encoding scheme, with the result that they would sound as close to dialup modems as humans can possibly sound. Think beatboxers without any of the repetition (any structure in their communication would be minimized in order to maximize the bandwidth of the channel, see Claude Shannon's Information Theory). It would be very unpleasant to listen to for those not of their race. At the same time, if body language is used to further expand bandwidth, it might make the communicator look like he was having a seizure.

Of course, this is what happens in the absence of compelling cultural forces pushing against such developments.

Their language would be highly contextual, and words would have many different meanings. My guess is that if you have a perfect memory, and you also know that the people around you have a perfect memory, there is no need to be too precise in your language. Kind of how when you get together with friends and one of them says, "Remember that one time?" and everyone knows what they mean because everyone shares the same memory. I would imagine a lot of conversations between friends and family might sound like this, "Where did I put the thing with the stuff?" Except they wouldn't probably ask that question, because they would already know where they put the thing with the stuff. More likely, they would say "Who took my thing with the stuff?"

As such, I think that language would have separate, distinct purposes and different languages might be used for different functions. For example, a more informal and contextual language would be used when communicating about the past than a language used to discuss medical research. There might even be a purely spoken mathematical language. Since languages could be learned rapidly, I think many people would create their own languages for fun and to accommodate different uses.

So, the reason you would have ambiguity is to make the language more beautiful, fun, and challenging. If information supply is infinite, demand for information will drop off sharply. Many of these answers assume these Beings with Perfect Memory (BPM) would want to convey a lot of information (like a qr-code) in each and every morpheme or sentence token however, I would argue that simply being able to remember everything doesn't mean that a BPM culture would be assign as much value to knowledge as we do. If knowledge and memory is ubiquitous, it loses value. Going to up a group of people and reciting a super long paragraph containing the bytecode for a video wouldn't be impressive. The challenge and fun of language would then be through analyzing different shades of meaning and seeing how quickly you could process a myriad interpretations of the same sentence. So metaphor, n-tuple entendre, rhetorical devices and constructing sentences that are aesthetically and intellectually pleasing would have a higher value because they would involve creating something new, not just rehashing something that everyone already knows already.

Plus, the 'remember that one time' example is far too simplistic. How it would really play out is that, say in describing a event unknown to listeners (like the time I slipped and fell on a banana peel), I wouldn't need to say the words 'fall' or 'banana', but could reference a shared memory and use other devices of language to come up with a humorous way of telling the story. Kind of the way a friend might say to you, recalling a time when you both narrowly avoided injury, "we totally Matrixed out of that," you're able to say "Matrixed out" because the image of Neo dodging bullets is shared and implied given the context. In a culture where information is ubiquitous therefore almost valueless, inference, connotation, metaphor, and style would trump mere volume of information.

One caveat, for a BPM culture, there may be instances where precision and volume are in fact needed, e.g., for chemistry or robotics where being able to quickly reference large sets of specific information would be necessary. Hence the need for different modes or modules of language. The way we have informal and formal English, they would have similar language structures formed around functional groups and occupations.

• If you can easily have a language with many millions of meaning containing units, why would you ever have that sort of ambiguity? "Remember that one time" could be referring to multiple different shared memories. I'm not sure why you think having perfect memory would somehow you vastly better at interpreting ambiguous messages. – Vakus Drake Sep 5 '16 at 0:52
• Yes, a perfect memory would make you better at interpreting ambiguous messages. Since everything can be contextual because of the memory, it might well be so. There's a STNG episode with a people who use stories, known to all as a sort of framing device for situations Damrok, I think it was called. You'd have the basic language and then the metaphoric/story-based one. – Erin Thursby Sep 5 '16 at 3:54
• @ErinThursby Yeah I see what you mean, though I don't think the specific examples in the question are good examples of ways that would work. – Vakus Drake Sep 5 '16 at 5:00
• I wouldn't say they would be vague because they wouldn't need to be concise. they might be vague because they are still human/human-like so they would get bored if they brought up anything straightforwardly and might be vague for entertainment value since memory games would be no fun. – Sarfaraaz Sep 5 '16 at 6:14
• @Sarfaraaz Yes they might still do some stuff like wordplay, however these people are far more alien than you're probably imagining and it's hard for me to imagine what kind of things they might enjoy. In a future question I'm probably going to explore how people would develop with perfect memory and the ability to control their mental state at will. – Vakus Drake Sep 6 '16 at 0:21

I would consider this question from a mathematical stand point, angling in from two directions on your scenario.

First, the average adult "test taker" has a vocabulary of 20,000 - 35,000 words [1]

Second, mathematics, takes basic concepts and composes them into more complex mathematical objects - all of which are represented by symbols rather than words - which in a sense could be thought of as vocabulary words, but with their own semantics and context, i.e. mathematics is it's own language.

In mathematics we have short and elegant equations, no longer ( and often times much shorter ) than an English sentence, which taken as a whole, can describe several highly intricate steps of computation taken on perhaps millions of pieces of information. This is especially true in machine learning contexts.

Language works in a similar way; by selecting words, assigning definitions or meanings to those words, defining valid variations which systematically alter the meanings of the words (e.g. past tense, plurality, etc.)

So if your language users have a larger vocabulary, and perhaps a broader range of alterations that can be applied to the words of their language, with rules that determine their use, then it is conceivably possible for very large amounts of information to be transferred in very short utterances.

Taking into account other ways of multiplexing increasing signal bandwidth in terms of language, such as intonation and a larger alphabet, then the size of words can be shortened without changing their meaning.

So the encoding aspect of the language hinges on the availability of variations in sound which represent bits or small chunks of data - letter grouping combinations and their associated phonemes. An important aspect of encoding is being able to distinguish between phonemes. This is done with statistical analysis in natural language processing by including the surrounding words in a phrase in order to raise the probability that the sound "see" is referring to the concept "to see", "the sea" or "the letter C". But again, by formulating a larger vocabulary it is possible to simply replace the sounds associated with each concept so there is less ambiguity. As well, by increasing the bandwidth of the signal, by adding intonation, having a larger alphabet and making sure that each letter in the alphabet has a distinct phoneme that can easily be distinguished from the others, makes processing the language less intensive by reducing or eliminating ambiguity.

The high level concepts again rely on the vocabulary size. Similar to the way equations represent high level mathematical concepts in the way an equation brings multiple concepts together by grouping with parenthesis, stringing symbols together with operators like addition, subtraction, multiplication and division, and feeding concepts into other concepts by functional composition, a high-level, spoken language, using commas, fragments, and recondite terms, can pull together complex information into a short description.

What this means for a spoken language that is designed to be fast and compact is that it should take up the full range of the audible spectrum, which depends on the qualities of the ear of the species. To an un-trained listener, the sound of spoken communications would sound very noisy, like white noise, somewhat like the loud sound that you get on an old TV set when the screen is displaying static from no signal. But perhaps with some hints of clicks, whistles and other sonic patterns that would be trackable by the human listener as something complementing, or perhaps segmenting the over-all noise. Think of the sound you get when you listen to a fax machine over the phone.

The English language, if we trace it back through its root languages in the ancient world, starts out with concrete terms that become more and more abstract over the years.

The word abstract provides an example. In English, abstract means theoretical; difficult to understand; speculative; separate from concrete realities and specific things. The latin roots of abstract give the idea of pulling away, or diverging from a set path. That metaphor became a word in its own right and now has a life of its own in the English language.

My guess is that because people with perfect memories would never forget the metaphors, they would never develop abstract words in their language. But languages need to describe things by comparing them to other things. Therefore a race with perfect memories would use extremely long sentences in which they give the whole metaphor instead of using abstract terms.

They might say:

"If it would give you the calm of a still river, take by the hand and place before me one open container made from sand shaped by fire and filled with the liquid that sustains life."

At least their language might start out that way. Since they have perfect memories, they could condense word groups into shorter collections of sounds. So they would have a sort of second, shorthand language with words that sound like random collections of noises--total gibberish to anyone else. It would not be like our way of combining concrete ideas into new abstractions, but like code words that stand in for whole sentences or paragraphs.

As for a written language, their memories would allow them to use a language that normally has to be decoded from a table, like Circular Gallifreyan. Because of the slight hand tremor, they may use a system of large basic shapes for consonants, and designate vowels with hash marks.

• If you're trying to maximize the information bandwidth of the language, and learning a complex formal language is trivially easy, why would you make your written language correspond to you spoken language? Having things correspond to vowels and consonants seems kind of inefficient. – Vakus Drake Sep 5 '16 at 1:16
• Vowels and consonants are the kinds of sounds used in language. Phonetic spelling allows us to represent those sounds visually. The alternative is to use pictograms, or else some other representation of sound that amounts to what I described. If you're looking for a totally silent language, that would be something like internet meme images and emojis. – Farkas Seven Sep 5 '16 at 1:51
• My point is that there's no reason for there to be any connection between the spoken and written languages. I think you get my point when you bring up pictograms, however there's also no reason whatsoever that the pictograms ought to in any way resemble anything. Like I pointed out in the question there's no reason the written language shouldn't look like a QR code or something else really alien (limited of course by slight tremors) – Vakus Drake Sep 5 '16 at 1:57

Some discussion here about contextual - reference to past events.

It would be worth checking out the Star Trek: The Next Generation episode "Darmok".

In that episode, the crew encounter a species who talk only in metaphor, with every sentence being a reference to some past event or legend.

Since this would require the species to know and understand all legends & myths, it must be assumed they have an excellent memory.

Could be worth taking this angle.

# Vocabulary

Individual will be as short as possible (before compression) to make speaking and writing quick. There will be many ways to change words in miniscule ways to change their meaning. Not only packing a sense of time (present, perfect etc.) but also much more so a sense of numbers (not only "one" and "many" like in real languages, but likely all numbers occuring in day-to-day life). All of this encoded in a way that it can be done with a letter/phonem or two changed in the word.

# Sentences

No "helpers", no redundancy. Many, many more grammatical rules which convey information. Information can be transfered by word order etc., much more so than in our languages.

# Context

As memory is perfect and assumably also perfectly shared, much more references can be included in everyday sentences. I.e., like we use hyperlinks on the WWW, there could be an abundancy of stored events, concepts, etc. that could all be referenced by a short hashtag (even just a content hash like in git).

# Compression

Much has been said about huffman compression. The lines are blurry here for me. Much compression has already taken place in the previous three areas (vocabulary, sentence, context), so it is unclear how effective additional huffman would be. But sure, no reason not to include it.

# Channels

The most important thing would be to use many channels concurrently while speaking, like "multichanneling" in fiber glass. This is all about bits per second here.

• Voice: modulate it as far as physically possible to get as close to the Shannon Limit as possible. To a human listener, it would probably sound much like gibberish/white noise.
• Pitch
• Volume
• Speed (i.e., bits could be put into the exact length of a sound, maybe distinguishing between a 0.1s and 0.15s length, if the magical mechanic secondary mind can distringuish that)
• Hissing sounds that can be layered on top
• Body language
• Twitches of all kinds of parts (shoulders, hands, arms etc.)
• Hand signs (the tremor should not matter there)
• Face twitches, head movements (rotating in 3 axes, moving in at least 2 axes etc.)

# Writing

For written language, since they have a tremor, they could use knifes to cut extremely fine lines. I.e., a standard alphabet could be achieved by layering horizontal and vertical lines over each other. Let's say they can distinguish 8 lines per 0.5cm with the help of their magic processor, then they can have, in a 0.5x0.5cm block, 16 bits of information (8 horizontal, 8 vertical), far more than the tiny, tiny amount we get into one of our standard-human characters. They would likely use some variant of Manchester Code to avoid problems when only a single line has to bee drawn.

If the tremor makes this impossible, they could likely invent a typewriter which can do it for them.

With unlimited memory (and unlimited speed to do a search in the brain) then there is no reason not to assign a unique word/code to each and every single thing that this race can perceive, feelings, colors, everything. I would think their language would resemble DNA strings, a long sequence based on a number d of unique symbols such that each symbol represents a physical sound that can be produced by any member of the race. The words are in theory unlimited in length, however, the addition of a new character in the sequence would imply (d-1) new unused words to put a meaning to. This exponential behavior would not be easily matched by the development of the race, so word length wouldn't be a problem. A thing to note is that two related things might not be even close in this language. For example, if we add the idea "My brother has a screen" years after adding "My brother has a rock", they might not share a single character.

Another downside of this, is that the oldest words get the shortest pronunciations, whereas the new ones require more time, more characters. Maybe a new word will be of daily use and would make this language inefficient. Two possible solutions that I can think of:

1) Technology might not prevail in this world of magic, but maybe the race knows the importance of communications, and they all together make a scheduled permutation of the meaning of the words where highly used meanings get the shortest pronunciations. (Books would be useless because as they get old, they would lose their meanings, but maybe in a world of magic, updating the books needs just someone to cast a spell and, voilà).

2) The positions of the characters do mean more than a simple one-to-one rule with the meanings. A tree of meaning-relation would be derived, such that words with similar meanings would share similar sequences. Branch insertion of knowledge in this tree would be very complicated without a scheduled reordering of the language.

If any of these two is a solution, then these books would be written in a single word. Conversations would also be an exchange of single words.

It will be very closely resembling Arabic. There are certain practical reasons for it.

# 1- Concision

A perfect memory means a lot of data to save. This also means that unless the language is concise, a lot of time would be wasted in communicating just a short incident or idea. Arabic takes the medal here as it conveys the meaning in the least words possible, in most situations. For example, according to Google Translate, the sentence Those two women went for shopping in different languages, becomes:

Arabic: ذهبت تلك المرأتين للتسوق
Zahbat tilk'al maratain litsooq (4 words)

Persian: آن دو زن برای خرید رفت
Aan do zan baraye khareed raftand (6 words)

French: ces deux femmes sont allées pour le shopping (7 words)

Russian: эти две женщины пошли за покупками
eti dve zhenshchiny poshli za pokupkami (6 words)

Greek: αυτές οι δύο γυναίκες πήγαν για ψώνια
aftés oi dýo gynaíkes pígan gia psónia (7 words)

Chinese: 这两个女人去逛街
Zhè liǎng gè nǚrén qù guàngjiē (6 words)

Hindi: उन दो महिलाओं की खरीदारी के लिए चला गया
Wo do mahilayen khareedari kay liay chala gaya (8 words)

This is the power of Arabic over other languages. Conveys the meaning without missing anything, in the least words. That will definitely be a prime requirement for those people as they would have plenty of data to communicate/store.

# 2- Precision

This is linked to concision, but not the same thing. In Arabic there are more than 100 words for lion, with slight differences (large lion, bush lion, angry lion, brave lion, charging lion etc) and more than 50 words for love and affection. This adds yet another advantage to communication as you have a large vocabulary to choose from, so that you get to choose the precise words for your requirement, which not only reduces words, but also conveys the meaning precisely.

• The natural English is "Those two women went shopping". I think the natural French is "Ceux deux femmes font les achats". Both shorter than the Arabic. This does not prove anything, except that Google Translate is not perfect. Human languages don't vary much in speaking time (when averaged over many sentences, not one or two) BECAUSE all humans have similar average mental equipment. They are about equally complex for the same reason, although the complexity shows in different ways - eg. word order for some, grammatical cases for others. Vakus Drake's race with perfect memory would probably > – Lostinfrance Sep 5 '16 at 8:16
• < consider all ordinary human languages equally clumsy. Although they might appreciate human languages for aesthetic or musical reasons if their own language sounded "speeded up". – Lostinfrance Sep 5 '16 at 8:20
• Sorry, the French should have been "Ceux deux femmes ont fait les achats". One more syllable. But if I made it "Ceux deux femmes ont fait les courses" I get the number down again. OK, I'll stop now. The point is that individual sentences mean nothing. And syllable counts aren't a straightforward measure of "efficiency" either. "Strength" is one syllable. It is not necessarily easier or quicker to say than its Japanese equivalent, "Chikara", despite that having three syllables. – Lostinfrance Sep 5 '16 at 8:41
• @Lostinfrance: As I mentioned in the answer, the translations were made by google translate and posted here as such. – Youstay Igo Sep 5 '16 at 13:25
• @Youstay Igo, it's fine to be an Arabophile. There are plenty of things to admire about the Arab world. Having an intrinsically superior language is not one of them. – Salmoncrusher Sep 5 '16 at 16:22

Photographic memory is curse especially in the medieval period with all the childhood death and horrid farming activities. Would you want to remember all the times of great pain in your life- you probably hit your head as baby a dozen times. No wonder these people's hands shake- mental stress is extreme. They probably wouldn't have their own written language because that would be horrid for their race to read and just tell others of their own race for communication or show an example/picture (made by a person of another race)- why even bother with written language for their own people and also torture for the other person. For other people of other countries, something simple like Braille or Rotokas alphabet which all of 12 letters- reduces the errors in writing. They might be limited scholars and couriers outside their home location trading for great art or powerful magic or powerful drugs to mitigate against the memories that they don't want.

• I only included information relevant to the question I was actually asking. These people are mostly mages to some degree, and they tend to be the nobles in a given society. All the stuff about disease and childhood mortality doesn't affect them because they have access to lots of magic. They can just turn off negative emotions and are totally capable of erasing their memories, but it's much easier to just remove the negative emotions surrounding memories. Basically they can control their own mind in any desired way. – Vakus Drake Sep 5 '16 at 10:02
• If they are that powerful then it would hardly be a medieval setting, easily jumping to industrial revolution like setting with absolute control of life and death since medieval nobles had a lot of childhood deaths (it wasn't just the peasants- stillbirths and childhood disease are hard to avoid) and they'd be leaps and bounds ahead of other races- walking, talking at very early ages. – user2617804 Sep 5 '16 at 22:59
• I also explained in the question that most technology is basically impossible. Making long lasting magic items is also too difficult to make a magitek industrial revolution possible either. Yes since they have perfect memory they will start talking very quickly after getting their magical enhancements as infants, what's your point? – Vakus Drake Sep 6 '16 at 0:18
• Also they need to forget the banality even nobles have no so interesting days- such much that the human mind chooses to forget so they use magic lots of times every single day to wholesale remove out out the memories- you have all five senses - overload to manage or just be overwhelmed. – user2617804 Sep 6 '16 at 0:47
• Their senses would only overload them if they were vanilla humans, but their not. A massive amount of their mental functions are handled by their mechanical mind, basically a magical computer. – Vakus Drake Sep 6 '16 at 1:10