Self-Reference Engine

by Toh EnJoe

The giant corpora of knowledge are indeed great things, able to advance by changing even their own past.

  But even they are not always able to make everything go according to their wishes. No matter where they turn, they encounter their fellows, each of whom is also trying to define how they want things to go. Each one of them is a petty tyrant, busily animating all the phenomena of the natural world. And at the same time, they are all babies who find themselves suddenly sitting before blank canvases and given brushes to paint with. They may all be far lonelier than humans could ever imagine. Or they might be just like humans. And that is why they find themselves depending on people like that doctor.

  “Okay. I get it.”

  James sits drumming on the knees of his crossed legs.

  “That doctor may certainly be a character, but he wouldn’t be my go-to guy on mental health care. Even for research subjects, it would be too hard to picture him as great for morale.”

  “That may be so,” Plato responds, but without much confidence. He is trembling from the weight of the decision.

  “Let’s just think of him as a weapon. We can shoot him at other giant corpora of knowledge.”

  “You mean, like a missile?” Plato asks, sounding ever more human.

  “He embraces a strange conception of the space-time structure. He is not very infectious, but I think he could still be a nuisance to the giant corpora of knowledge. You have already experimented with him. If we shoot him over, they’ll get all confused. Their calculation speed will drop. Just like what happened to you.”

  This decision takes only a split second.

  “This is an idea worthy of consideration.”

  Plato appears to be keen on the idea. James decides not to think harder about what sort of door he has opened. The giant corpora of knowledge might take this idea and design a new space-time structure seven times more ridiculous than this one. Without a doubt. They could deploy the doctor in the vicinity of the other giant corpora of knowledge, and he could really screw them up, like some totally twisted space-time model in human form. Maybe they could think of some way to make him more infectious. James wouldn’t be surprised to learn that doctor had come to them as some kind of weapon from another world.

  The plan to weaponize the doctor may have been just a way of crawling out of a tight spot. But there are only so many ways to keep a ship from sinking. Bail out the water, plug the hole, or stick your finger in it. You keep moving forward, one second at a time, any way you can. Until the safe harbor swings into view. Atop innumerable spheres where the solid surface has been submerged below the floods.

  “James,” Plato says, and James lifts his head. “Are you going to sleep now?”

  “I am getting sleepy, but what’s up?”

  “Nothing. I just wanted to say thank you.”

  Plato is not in a mood to say sarcastically that he was too meek today. He was just in a kinda-sorta mood, leaving his tongue all tied. He felt like people would be used to saying certain things in situations like this, he just couldn’t remember what they were.

  “Okay,” James says. “You can stick around for a while, if you want.”

  It feels strange to be saying that to a giant corpus of knowledge responsible for operating the base around the clock. But James does not feel like laughing.

  “I would like some vodka. And pistachios.”

  “I can get those for you. But first, I would like to make a few spot-past changes inside your body.”

  James adjusts the way he is sitting on the bed, and he makes a verbal report to his boss. What has happened, what is about to happen, extremely simply, and as nasty as he could make it.

  “Decision. Worsening. Hung over. Late. Over and out,” James says.

  Hearing this, Plato cannot help but laugh. “I can think of about twenty thousand different reasons why you’ll never get ahead,” he says.

  13. JAPANESE1

  1 This piece is a speculative translation from the Japanese.

  THE TOTAL NUMBER of Japanese characters is said to be over twelve billion. And that is just an approximation.

  That these Japanese characters can be divided into categories, the people are agreed. The following categories have been recognized, albeit only in broad terms: kanji, kan-kanji, kan-kan-kanji, hiragana, hira-hiragana, katakana, hira-katakana, kata-hiragana. Debate is still going on over the proposed category hira-hira-hiragana, and some have proposed the need for the categories hira-kanji and kata-kanji.

  It would be fair to say that no progress whatsoever has been made on any process of understanding these categories.

  Most of the texts that have been discovered so far appear to consist of one hundred thousand to one million characters, but one obstacle preventing the reading of these texts is the simple fact that in the set of characters appearing in these masses of characters, the same character seldom appears twice. There is very little repetition of specific characters. The character most often repeated, the hiragana ぴ, appears only seven thousand times in the entire corpus of Japanese-language texts. And with variations in the way even this ぴ or ひ。is written, there are some who argue these are just differences in fonts, and some who say they are completely different characters.

  Research has not yet even progressed to the point where numeral codes—which are often important clues to the deciphering of unknown characters—have been identified. It is believed, with a high degree of assurance, based on their simplicity, that glyphs like 一,　二, and 三 are numbers, but it has not yet been determined what characters are associated with the numbers four and higher. Some believe the characters 口 or 木 mean “four,” but this has not been fully corroborated.

  Greater confidence is felt about operators than numerals, but there is also confusion and speculation surrounding 十 (ten) and the plus sign (+), and 二 (two) and the equals sign (=). And there are other characters such as 廿, 土, or 王 that some believe to be numerals or operators, but there is no consensus.

  Anybody can see there are frequently characters in texts that appear to be calculations, but these do not contribute to an understanding of the text as a whole. The method of number notation does not appear to include the concept of “grouping.” If that concept could be identified in Japanese texts, the same characters should show up with some frequency even if a base-n number system were being used. It seems pretty clear that a base-∞ is being used, but even so it seems the rates of recurrence for particular characters are too low, so some have developed the theory that characters change their form depending on what notation method is being used. This theory, which assumes that different symbols are used to write the first instance of “one” and the second instance of “one,” seems to explain something, but it also seems to explain nothing.

  Japanese texts are written either vertically, from top to bottom and right to left, or horizontally, from left to right and top to bottom. This is clear from the orthography and from the form of the strokes of the characters, the brushmarks. But here too, what seems peculiar to these texts is the way one character seems to be joined to the next—whether horizontally or vertically—and the interactive nature of the writing. Oftentimes in vertical writing, the horizontal strokes of a given character line up across several rows. This is not because overlapping characters are written in vertical lines, but because of natural connections between characters. When texts are written horizontally as well, there are of course some intersections with rows above or below.

  Broadly speaking, it seems there are no real differences in how the sectioning is created in horizontal or vertical writing. At least, it seems there is no difference in meaning whether the text is written horizontally or vertically. Of course, there are some who say there is a difference in meaning between texts written vertically and those written horizontally, but there is no real foundation for that view. Actually, there is no real foundation for the opposing view either.

  It is believed that such artful technique would only be used in the final copy of texts th
at were already carefully composed in advance, but many of the texts themselves seem to be more like random scribblings. Nowhere, though, does it seem that characters are crammed into prepared horizontal lines, with neat white spaces in between. Sentences in these Japanese texts seem to be written in a natural scrawl.

  Based on that high degree of design sophistication, there is also a theory that these texts are just linkages of black lines, made mainly to look like letters, with no particular meaning at all. But they seem too much like sentences, too much like signs that carry meaning. One can imagine someone writing meaningless scribbles in great volume, but it would be harder to argue that that activity itself is devoid of meaning. As was seen long ago in the case of Pseudo-Dionysius the Areopagite, which caused such a stir throughout the world, the mere discovery of a large trove of ancient texts can become an Event in and of itself. An algorithm was devised that explained a lot about the writings of Pseudo-Dionysius, leading many to conclude that they were fake, but it would also be possible to say that what led to the forger’s downfall was the fact that each individual character in the text was distinguishable from the others. Nowadays, inferences made through mechanical calculation have overwhelming power when it comes to sequences of characters in finite sets. The difficulty of written Japanese is that it is hard to determine even if the number of code symbols used is finite.

  Regarding this design sophistication, there is one more theory that could even be termed fascinating.

  The Japanese text, composed of these tens of thousands of characters, can itself be thought of as one enormous character.

  Based on this idea, supposing the whole is just one big character gives rise to the skeptical view that the constituent elements are entirely too fragmented and inconsequential. If someone wanted to write an enormously complex character, why wouldn’t they just write an enormous character? They wouldn’t have to use a whole complicated system of small, enmeshed characters. And besides, there is also the problem of page layout.

  The assumptions behind this view are dreamlike, so the resulting debate is also dreamlike, among both proponents and opponents. Here too there is a variety of opinion—people write things, and when the project is something on a large scale, they may take a divide-and-conquer approach; some people think the brain itself is organized that way, and some think there is a two-layer structure of sound and meaning, while still others think there is a three-layer structure.

  As a practical matter, there is of course major skepticism about how a language with over twelve billion characters could ever be written. One strong opposing argument is that no one could ever possibly learn how to do it.

  Just as for any given view there is an opposing view, to this question too there is an answer. In a sense, these rows of code are rows of code that are constantly changing in accordance with a relatively small set of rules. If one simply learns the rules, there is no need to commit all twelve billion characters to memory. Combinatorially, twelve billion is not really such a large number after all.

  If only those rules could be deduced, the whole theory would become more persuasive, but the proponents of this particular view are not even able to formulate clear expressions of the rules, so this too has been relegated to the status of an unprovable claim.

  Many different proposals have been put forward regarding the names of the characters that appear in Japanese texts. The first ideas were a bit strange, but more recently many neuroletters have taken their place.

  Letters, or characters, no matter how many there are, are always symbols that stand for something. The word chair is not a chair, nor does it indicate a particular chair. It signifies chairs in general. It is hard to believe there were twelve billion different ideas that would require twelve billion different symbols. What had defined this term was the view that, in this case, this must have been the language that recorded the operation preceding this summing-up process.

  It was the neurophysiologists who first suggested these characters were more closely akin to the behavior of the neural network itself than they were to the concepts that had been output by that network. This view, which was not without a certain persuasive power, was just like all the other great theories, however, in that it did not make any contribution toward deciphering the writing. Insofar as it seemed to explain something, though, this theory had at least a certain psychological edge and was just that much more popular than the others.

  Of course, this thesis also had its contrarian counterpart.

  As demonstrated by the game Twenty Questions, the identity of unknown objects and concepts that can be categorized by people can often be guessed at via a series of about twenty yes-or-no questions. Making these choices is a process requiring 220 codes, a figure that makes 12 billion look modest. If this number of characters would be insufficient for the number of categories, depending on how the categories are defined, just imagine how much less than sufficient it would be to describe the activity taking place in Japanese texts.

  The neurolinguists took in this critical view, but they soon went on the counterattack. The twelve billion Japanese characters now known, they said, were just an estimate based on the materials discovered so far. In fact, there appears to be a low rate of duplication among those characters, and if twice as much new material is discovered, the number of characters could also easily double, or possibly even triple. Until the full extent of the corpus of text was known, that simple counterargument would hold no water.

  The counterargument was fine as far as it went, but the idea that the twelve billion characters already known were only a part of an even greater whole did not exactly brighten the days of the researchers. The thesis itself was already enormous, and with each repetition of the counterargument, the scope of the problem expanded at an alarming pace. When it came to the interpretation of Japanese texts, this sort of situation arose with some regularity, and the main obstacle to the cracking of Japanese was the suspicion that at some point the researchers would simply stop thinking, or that the material evidence itself would simply become unwieldy. In part, this seemed to reflect both an inadequate supply of material and simultaneously an excess of material. The atmosphere within the field only encouraged people to believe the rumors that a research team had found new material in the former Japanese archipelago and destroyed it.

  Various explanations for the complete disappearance of the twelfth research expedition to the former Japanese archipelago have also been proposed. Many people are skeptical of the official report that the expedition was attacked by savage natives. Newspapers featured cartoons of a bunch of top-knotted sumo wrestlers on the attack, but it is widely known that nothing larger than a dog is now alive in the former Japan. The people once known as the Japanese are celebrated historically, but there is no hope they will ever return.

  The trail of the research expedition to the former Japanese archipelago stopped in the Hachioji area of ancient Tokyo, and all theories about what happened to them there remain purely speculative. Their food, fuel, and other provisions were left behind, with no sign of anything untoward. No signs of struggle were discovered. Logs were discovered, as if scripted, telling of the banal progress of the research, but shedding no light whatsoever on what had happened to the team.

  From the logs, it was learned that the research team had uncovered a large trove of Japanese text in Hachioji. The logs record that over twenty tons of paper materials had been recovered at Point 13/20, the exact location of which was not given. The team’s final camp was located in a park, surrounded by abandoned office buildings, but there was no sign that twenty tons of materials had ever been brought there. A subsequent investigation turned up no materials, nor the kind of heavy equipment that would likely be needed to move twenty tons of stuff, nor any trace that such equipment had been used.

  One slightly unusual aspect of the logs was an apparent attempt to write them in a script that looked something like Japanese. Judging by the handwriting, this Japanese text, which was written in reverse s
tarting at the back of the log, appeared to be the team chief’s.

  The content was unknown due to the incomprehensibility of the text. Considering, however, the fluidity of the lines, it is believed the team leader had gained a certain confidence in his skills.

  If this text had been written with some correspondence to some known text, it might have functioned as a kind of Rosetta Stone, but the team chief had apparently not paid much attention to that when he wrote it. Or perhaps he had been unable to do so.

  Writing in which symbols are substituted for other symbols is known as a “code.” The strongest known coding methods are those in which the original text and the code key are converted to numbers and overlaid against one another. Codes constructed in this way cannot be deciphered unless the key is known.

  So, the problem with codes becomes how to communicate this key without distributing it too widely. A breach can be the gateway to a total loss of the meaning of the coded content, but it’s not as though the Japanese language had been deliberately constructed as a code. Prior to the destruction of the former Japanese archipelago, the language had supposedly been used there as a matter of course, even if its use had been limited to that location.

  The known locations where Japanese texts had been discovered were primarily in the eastern Japanese islands, but there were also a few in the western and southern Japanese islands. In all cases, the texts appeared to be personal correspondence, written by hand, with a brush. No machine for printing Japanese text was discovered. If such a machine were to be discovered, that might make a major contribution to the deciphering of the language, but if such a machine were even possible the current situation would likely never have arisen to begin with.

  It is at this point that the opinion emerged that the Japanese language was developed as a tool of resistance to the giant corpora of knowledge. This view is not without a certain persuasiveness, considering that even now Japanese texts are incomprehensible to the giant corpora of knowledge. A script that cannot be broken down, in a combinatorial sense, into its constituent elements is obviously a challenge for computer processing power to deal with.