A Chinese Typewriter in Silicon Valley

Uploaded by GoogleTechTalks on 23.02.2012

>> up to recently at Google Research recently joined Google Android. As a native Chinese,
I was very fascinated by Tom's work on Chinese ethnic group classification. It's actually
interesting from a computer science point of view today that, how would you classify
people into--who had been living, you know, who have had patch of land for thousand, thousand
of years and how do you divide them into different ethnic groups? What features would you use?
Is it--is the language that they speak? Is it the proximity in geography or is it the
clothes they wear or how they marry? What features do we use to classify them to different
categories? And Tom did the most thorough study of this, that how--officially was classified
in the 1950s with the pretty important political ramifications in terms of representation even
though most of the representations in province of China was talking but still, it was important
for all groups to be represented. And he traced and analyzed how that project went and it
was fascinating. But today, he's going to talk about another topic that is just as fascinating
which truly is interesting to me because as a research scientist I've devoted a big portion
to my career and study how to enter information into computers. And Tom's study shows that
much of this work we do today on mobile phones, on touch screens are actually foreshadowed
by what Chinese engineers and inventors have done in the past many decades. So, I'm very
pleased to be able to invite him to give a talk on this topic today. So, please welcome,
Tom Mullaney. >> MULLANEY: Thank you very much. Can you
all hear me okay? Excellent. So, it's a great pleasure to be here and an honor and I want
to thank Xu Man Jai for inviting me to speak with you today. The talk of the paper has
changed many, many times, many iterations. So, we're going to keep the main title of
the Chinese typewriter in Silicon Valley but the subtitle has changed a little bit from
the abstract to how Chinese typists invented predictive text during the height of Maoism.
To give you a little bit of context, the broader set of questions that I'm working on right
now in a book, I guess tentatively titled, The Chinese Typewriter, a Global History,
is a broader story of 19th and 20th century China which pretty closely connected to the
history of European--Euro-American Imperialism is the question of how to render compatible--how
to render the Chinese language compatible with the set of information technology such
as the telegraph, the typewriter and later forms, which almost in their DNA have been
so closely connected to alphabets or more generally speaking, to languages with a very
limited set of modules, be they alphabets or syllabaries. And so, there was, during
the 19th century and 20th century and even today, this very fundamental question that
I post up here, is Chinese script, not the language as a whole, but is script, the Chinese
script, Chinese characters, are they compatible with modernity or must they be jettisoned
so that China can modernize. And this was a major question in the 19th and 20th Centuries
with many people saying no, we have to get rid of characters, use English, even Esperanto
in order to modernize. And many even more who saw--who used the imagery of the Chinese
typewriter in particular, as kind of proof of this. I say proof in--I italicized that
with my voice so as to say that the proof came in a sort of concocted imagined ideas
of what a Chinese typewriter must theoretically look like. That would be an immense machine
with thousands of keys upon which it takes even five people to type and so forth. And
so, you see, some of the earliest and most derogatory and racist perceptions of this
from turn of the century that top one is from an article in 1900 all the way through the
kind of dregs of B movie culture with a Tom Selleck movie, the Chinese typewriter into
the famous dance coined by MC Hammer in You Can't Touch This, known as the Chinese typewriter,
that dance was called, because it was meant to mimic what a typist must look like as they
move across a massive keyboard into the Simpsons and so forth. So, that is to say that within
the story of--the question can--is Chinese--are Chinese characters compatible with modernity?
The Chinese typewriter in particular became a kind of icon for those who said no, it's
not. And it's a surprisingly durable icon. And what I'm here to talk about today is that
parallel to the story, there's a far richer, more significant and interesting story of
those who did not give up on this idea, that Chinese characters could be compatible--or
rendered compatible with technologies designed without Chinese in mind. And I'm charting
in my own work the kind of larger history about this involving Chinese telegraphy, typewriting,
Braille, and indexing systems and so forth. But what I'm going to talk with you about
today, if I had to kind of make this relevant for an audience that is obviously very interested
in deep history, but is also setting out to build things and make things is that when
I see the Chinese typewriter, I don't see this. I see a machine whose history is an
incredible repository of designed inspiration and some of the most eccentric and brilliant
innovation, most of which never saw the light of the day, never materialized into forms,
but some of the most brilliant and penetrating analysis and innovation of human-machine interaction
of input of data structuring and many other dimensions. And so, in particular, what I
want to talk about today is a kind of episode that takes place in the 1950s in China. Now,
this is going to be the sort of context and I will be explaining this in a bit. This is
a few snap shots of different Chinese typewriters in the 20th century. The episode comes from
November of 1956 in the pages of the most--the most widely circulated Chinese newspaper,
the People's Daily which featured an article in that fall--one of the fall issues of a
typist in the city of Luoyang in China that had performed this unparalleled feat and was
reported to have using a Chinese typewriter whose operation I'll explain in a second,
using a Chinese typewriter had typed at speeds verging on 80 characters per minute. Now,
I invite you not to immediately compare that to QWERTY inputs or things like that, because
there, I'm going to try to explain that they're largely irrelevant comparisons but the important
comparison is how fast had a typewriter--a Chinese typewriter been at that point. Roughly
speaking, if you look at a lot of the archival documents, teaching manuals and so forth in
the '20s, '30s, '40s, '50s and on, the average typing speed at this moment was about 20 to
30 characters per minute. So, this thing--this event, this person doing this, if it was real,
represented a three-fold increase in speed of input, if you can think of it that way,
without any substantial changes to the kind of mechanism of the machine. This was not
a new typewriter. This was some sort of change that had taken place to the existing machine.
Now, to give you a sense of how the machine works, how these various models up here work,
you'll notice that there are--there's a tray bed, if you see right here in the [INDISTINCT]
model here. There's a tray bed on the 1960s Double Pigeon model, of 35 characters by 70
characters. These are free-floating metal slugs. If you were to pick up this tray and
turn it over, they would all fall out. There is a tray selector that can operate along
an XY axis and actually can XY--and then the tray bed itself which can move left and right
along one axis. And so, you kind of move both, bring the selector in positions, pushed down
on the lever and the--something pokes up the slug from the bottom, it is grabbed, inked,
it strikes the surface of the platen and then it's returned into the same location. That's
the basic functioning of these machines. Now then--so, if you can imagine, this is about--this
is 2450 characters. So, the most important dimension of this is taxonomy, is the organization
and disposition of these characters. Now, the way that in the 1920s and '30s and '40s
characters on the--on the--on the machine had been organized will be very familiar to
any--anyone in the room who reads or speaks Chinese, but for those of you who don't--excuse
me, it was the--a system known as the radical-stroke system. And basically, this system which dates
back to the Ming Dynasty, popularized in the Ching Dynasty, that is into, roughly around
the 16th to 17th or 18th Centuries, we're talking about, is a system by which the tens
of thousands of characters in the Chinese language are subdivided into a set of 214
classes or categories based on the primary component out of which the character is built.
So for example, one category of these 214 is the person radical category, the shape
here. And here are some examples of characters that fit within it, that are built out of
it. The--another radical class is the water radical class here, these three strokes and
you can see two characters that would fit into that. And then within each class, the
characters are organized according to the number of strokes it takes to produce them.
So the character Ta, meaning he would come before the character To because it takes more
strokes of the pen or the brush to write the character To. This is the way that dictionaries
have been organized and the way the typewriter had been organized. What the article in 1956
suggested was that this typist in the city of Luoyang had departed, had kind of taken
a radical departure if you will from the system and reorganize the tray bed according to a
natural language arrangement. And so if we give--if we take a sample, a very small sample
of the 2,500 or so characters--those of you who don't read Chinese, follow with me now.
The yellow--the highlighted yellow character is the character Tong. And the important thing
about this is that unlike the previous organization which--in which that character would have
been put next to characters that has nothing to do with. In this reorganization of the
tray bed in 1950s, all of the adjacent--or many of the adjacent characters in the Morse
neighborhood, the eight cells around it were--the disposition of those was such that they could
be combined with the character Tong in order to produce a real word. Most words in Chinese
are actually made of two characters not one. And so for example, this--the yellow character
and--can be combined with what's below it, [SPEAKING IN FOREIGN LANGUAGE] meaning, at
the same time. To the right, [SPEAKING IN FOREIGN LANGUAGE] in the same period. And
then invertedly with the character above it, [SPEAKING IN FOREIGN LANGUAGE] meaning, shared.
And then if you can do this with each of the characters on this, every character is both
a center and a periphery can be in combined or combinable with everything around it. And,
you know, we have to--when we look at an article like this from the People's Daily, which--we
have to be very careful. We have to be very careful as to the story and not to believe
what we read and kind of follow up. There had been many stories of model workers in
the--in the--in the height of Maoism. Model laborers who out produce their quota. Model
farmers who out produce their grain quota. And here we have a model typist who had out
produced their typing quota, their character quota. So we have to be very careful here.
But interestingly, is having spent quite a long time checking up on this, we see that
this actually happened and was--I'm going to suggest here, one of the earliest implementations
of a system of kind of techno-linguistic system that we now think of as predictive text or
natural language arrangement or however we want to--we want to describe it. So more broadly
I think before I get into the nitty-gritty of this, I think it's pretty self-evident
but let me spell it out that if this is true, then what this alerts us to is the need to
look very--much more seriously at a techno-linguistic innovation in three contexts where we don't
often go looking for it, China, the Chinese language and in this case mechanical pre-computer
system. At the very least, predictive text is not associated with Chinese Language, not
associated with innovations in China and it's certainly a very rarely or ever associated
with things in the pre-computer information age. So if we--if we--we have to begin by
asking the question--this will be relevant here in a second. The primary innovation that
I'm talking about here is obviously not a--not the design of a new machine but a new interface.
This is a new interface between a Chinese typist and the machine. And so we have to
ask the question, was there a context for this typist in Luoyang to kind of pick up
and move around the modules on this interface? Was this a total radical--a totally radical
act, a very unprecedented act or was there a context for this? And there was a context
for this. And the simple--the simple framework for this is that unlike the story of the typewriter
and specifically the semiotic interface of a typewriter in every other part of the world.
Be it--let's refer to languages more particularly, be it English, French, Italian, Russian, Hebrew,
Thai, Cambodian, Arabic and so forth, in the story of which the semiotic interface of each
of these typewriters at some point stabilized into QWERTY, AZERTY, and--QWERTZ and so forth.
That the story of the machine interface of the Chinese typewriter never stabilized, was
never meant to stabilize and was understood as something that would constantly need to
be adjusted or changed by the operator through a constant process of optimization. Why is
that? The reason for this is--comes in the fact that the basic--let me put this up here,
that the basic structure of the Chinese typewriter was based on the system called common usage.
So there are many more than 2,400, 500--2,450 characters in Chinese, there are tens of thousands.
The idea of the common usage Chinese typewriter, and that's what we're talking about, is to
select the most frequently used characters and to put them on the tray bed. And then
the rest of them would either be in a box about 5,000 characters of less commonly used
characters and then anything else beyond that would have to be written by hand, okay? Now,
the sort of--the issue for us here is that no matter how perfectly you chose these 2,450
characters, no one set would be perfect for every given context. In a police office versus
a bank, the common usage character would be different from one to the other. And so therefore
a typist in a police office--police office would have to take an uncommon character and
put it on the tray bed, and the same thing for a typist in a bank or a government office
and so forth. The other, there's a diachronic or a temporal part of this optimization story
which is that from moment to moment, common usage characters fell out of existence, uncommon
characters became common and some very--some prime examples of that pertain to the administrative
geography of China. So for example in the--in the teens and early 20s, there was a province
known as Fung--I'm sorry, as Jiuli Province, so--but--so the characters Jiu and Li both
appear on the machine in the teens. After roughly the mid 20s, the province of Jiuli
was abolished as an administrative region and was divided into a set of new provinces.
Now, what happens to the tray bed? The character Jiu is very, very common in many, many, many
other Chinese words and so it was kept. The character Li is very uncommon on its own and
so it was judicent, the same thing happened with the province known as Fung Tien, which
no longer exists. And after this transition, you can compare the tray beds and see that
the character Tien which means sky, very, very common is kept and the character which
means Fung which means to present something to a superior was a very, very, uncommon character
in--and so it was judicent. So this is all to say that there was a kind of built-in--already
a built-in kind of churning optimization process in which these operators in 1956 would have
been operating. It's very interesting that--so actually, let me--let me tell you a little
bit more about the structure of the tray bed here. Here are three different tray beds from
three different periods just to get this impression in your mind. The red region are the most
commonly used characters. All of these are commonly used but these are the most common
used characters and then the dark blue or the kind of purplish regions are the secondarily
common usage. And I'll talk about that kind of light blue in a second. So there was even
a structuring of frequency within the tray bed that will become relevant in just a moment.
And what's interesting about this is it's not just--the optimization of the machine
was not simply at the level of the tray bed and the disposition of characters on it, it
was also a question of how a Chinese typist would undergo training and how they train
their body. And to put this in a comparative context again, because it helps illustrate
the point, in Roman alphabet or Cyrillic, or Hebrew, or Arabic or so forth typing, the
ideal state of being when you are typing is known as blind typing. If you are sitting
there staring at the keyboard, you better go and practice a bit more. You know, the
idea is that you can look at the manuscript or the text and then have a kind of non-visual
relationship with the interface and just go. This way of interacting with the machine makes
a lot of sense in alphabetic and syllabaristic languages and makes no sense and was never
endorsed in the case of Chinese typewriting. For the Chinese typist and I put up some various
typing manuals from different points in 20th century Chinese history actually emphasized
the need for the typist to refine to ever greater and greater ability their vision,
and their memory, and their peripheral vision. And I'll give one example. The idea--one of--one
of the most important parts of being a good typist was to always be anticipating both
cognitively and somatically the next character, the next character, the next character. So,
if you were to go to one character that is all the way over to the right of A tray bed,
I know that my laser point doesn't work very well. Let's say you need a character over
here, and then you know--you're looking at the manuscript, you see the next character
and you realize that it's over here, you need to start kind of orienting your body in such
a way TO getting ready to move it back in that direction. You know. And if you don't
do that, you'll still be able to type but you'll type very slowly. So that is to say
that even in the training of the Chinese typist because of the sort of almost built-in affordances
and limitations, constraints and abilities of the machine that we see in analog to this
in which--in the way that a typist would train in China as opposed to many other techno-linguistic
context. One other very quick example that I like very much pertains to the--what those
of you who have read any [INDISTINCT] would know is materiality of signifiers and in a
more concrete sense means that the typist would have to adjust the strength, the force
that they use to press down on the typing lever, depending upon the number of strokes
in the character, because the number of strokes in the character translated directly into
the weight of the character. And so, what--some characters are one stroke like the character
"e" meaning one, is just one horizontal stroke. And then there's another one for dragon long,
which has many--more than 20 strokes. If you use the dragon strength force on the character
one, you could puncture the paper and therefore have to start all--start all over again. And
so this idea of--kind of optimization anticipation operated on many, many levels. Okay but this
is--this is kind of necessary but insufficient set of conditions to get us to this more radical
change, the semiotic change, this change of interface that takes place I'm arguing in
the 1950s, and so how do we get to a place where in 1956 someone actually sits down and
removes all of the characters from the tray bed and then rebuilds some sort of natural
language arrangement of all of the characters. How do we get to the place where someone would
feel mandated or authorized or inspired to do that? The question of this is really connected
to the politics of the 1950s--the politics of the 1950s. And in particular, a politics
that was, within 20th century China, unique to the Maoist Period, which was a centrally
endorsed, proletarianization of knowledge that is kind of user-led innovation in a sense
where in a vast number of different expert domains including seismology, paleontology,
medicine, the State--the Central State in Beijing and the Communist Party was advocating
mass participation of none elite, kind of, participants in these otherwise expert systems,
and was also advocating the really--the overturning of knowledge systems. And the kind of, if
you want to say--if you want to put it in the kind of a prose format to take much more
seriously folk knowledge, to take much more seriously the knowledge that everyday people
have. And so there does seem to be a strong correlation between this advocacy that's--and
mandate in authorization that's being issued from the party states and the type of experimentation
we see in the 1950s. So, its very, very difficult to locate the origins of moments like this,
and in many cases finding an exact point is not all that important. But if we had to--if
we had to choose the first moment in which we see a natural language arrangement of Chinese
characters in an information technology environment, it takes place not in typewriting but in typesetting.
In the City of KaiFeng with the work of this--a typesetter named Zhang Jiying. And Zhang Jiying
worked as a typesetter both in the pre-Communist era in the 1940s and then into the early--into
the Maoist Period in the '50s and he posted very respectable typesetting speeds throughout
most to his career, and his character rack, the rack where all the characters were displayed
for his use were also organized, largely according to the Radical Stroke System, the same one
that we saw before. Around 1951-'52, inspired by this new mandate for the proletarianization
of knowledge mass participation and so forth, he began to reorganize these characters, to
maximize the adjacency of those characters that go together. Some of the kinds of prime
examples for these come from the name of the Press he worked at, the Xin hua she. So he
said, well I'm constantly setting Xin hua she so why should I reach around this character
rack, why not just put them next to each other? Another example is Ge ming, Revolution. I'm
constantly saying ge ming or constantly setting this type, so might as well put them together,
and then my favorite is, mei di, American Imperialist. This was the time--this was the
time of the Korean War in the first really Nationwide Campaign of the Maoist Period which
was Aide Korea--Resist America, Aide Korea as--in connection with the Korean War. So
these phrase mei di was said over and over and over again. Now what's interesting is
it's very hard to say, was Zhang Jiying inspired by this--by this central new mandate of the
Communist Party or were many people doing this and we just don't know about it. It's
very hard to say the one thing that we do know is that the party state was not only
advocating or promoting this type of proletarianization of knowledge but whenever it happened in a
locale or a region, they were publicizing it and popularizing it. So Zhang Jiying was
actually--became something of a little media star for a few years. He was--he was featured
in the People's Daily, there was a book published about his method, he was sent on a paid tour
of China, to various printing houses to tell people about his method, the Central Broadcasting
Company filmed him doing his job and he was eventually admitted into the party, and so
he kind of, was a--his method was celebrated. It's very quickly after that celebration of
his method that we start to see the same application of natural language arrangement of the--of
key--of characters on the Chinese Typewriter. And this brings us back to the 1950s in the
story that began the talk. So this brings us back to again the same--the same example.
Now what's interesting about this is that there are--if on a 42-key keyboard, there
are 42 times 41 times 40 so forth possible arrangements on a key--on a tray bed with
2,450, there are 2, 450 times so forth and so on. There is effectively infinite number
of arrangements--possible arrangements of these characters. And so, we see is that in
after this--after the predictive turn in the 1950s, we see multiple people applying this
to their own tray beds centering around certain basic--basically shared principles and yet
each of them completely individualized and completely idiosyncratic. There are no two
predictive tray beds that are--that are alike. And let me give you just a few examples of
these. There's two tray beds that I'd like to talk about. One is a Chinese Typewriter
used at the Paris--the Unesco Office in Paris. Another one is use in the United Nations Office
in Geneva. These are two separate Chinese type--typewriters and typists working in very
different contexts and I--and I'll walk you through this as well. On the left one here--and
this is a 1930s just to give a comparison against the pre-predictive tray bed. In the
middle here, in the Unesco Machine we see this character in yellow Mao, this is the
surname of Mao Zedong. And if we look to the--below it, we--the two characters below it are Zedong,
so that he could type or she could type Mao Zedong, Mao Zedong, Mao Zedong quickly and
easily. Emanating off to the right of it are--is the character--are the characters Ju and Shi,
Jushi, Mao Jushi which is Chairman Mao. And so kind of the Mao is again constellated in
such a way that it can be kind of concatenated with a bunch of different characters that
this typist needed. In the right example, in the Geneva Machine, we see the exact same
idea but in essence constellated differently. We see again the character Mao in yellow,
and now Mao Zedong doesn't go up--down as in this example but all off to the left. And
Mao Jushi, Chairman Mao doesn't go off to the right but goes down. So in essence we
see these typists converging around certain principles but each one very idiosyncratically
organized. To make this a bit clearer for those, you know, for those who don't read
Chinese we have the example of punctuation. Predictive text was even applied to punctuation
in then actually in numerals. But anyway, in the 1930s the question mark, exclamation
mark, period, comma, were put where they kind of rationally belong all together in one place
and after the '50s a set of typists began to say, well that doesn't make much sense.
We know that question marks always come after a question particle because that's how you
express--that's one of the ways you express the interrogative in Chinese is by the addition
of particles. We know that the particle--that the characters ma, ba and na those are in
red there are always followed by the question marks so that's constellated around the question
mark on the machine. We see in the United Nations machine by comparison the exact same
idea but again a different configuration of these other particles. Now why are they differently
configured? And this is the really fascinating thing that I'm sort of working with various
set of data visualization specialist and others to try and think about, it could be all sorts
of things left-handedness, right-handedness this is sort of mnemonic devices that one
person prefers to remember the location or another. All sorts of factors could go in
to why this idiosyncrasy prevails but the really--but kind of anchoring any of these
idiosyncratic personalization are again these kind of cohering strategies that we often
see from one example to the other. This is a basic heat map which is--basically visualizes
the--is a calculation of adjacency it's the bright red or the dark red patches are those
that in which the character can be combined with maybe seven--six, seven, or eight of
the characters in the surrounding area the kind of--the whit, the very, very, very light
white-red, the white can be combined with one and then the black here indicates that
the character cannot be combined with any of the characters in the Morse neighborhood
to form a real Chinese word. This is a heat map of a tray bed from the Republican period
1911 to 1949 before the type of innovation I'm taking about. So you can see that there
is some almost accidental situations in which a character is flanked by other characters
that it can be combined with and then there's one part here that I'm happy to talk about
why there's--why there is a collection there. The important thing is that once we go to
19--after 1949 this is what the tray bed looks like it heats up, it lights up because the
basic principle again is the maximization of adjacency of related characters. So to
put them side by side we can kind of get a sense of what all of these incremental innovations
in the interface of the machine translated into and this is how we explain an acceleration
of the speed of typing that could be that high as indicated in the People's Daily. Not
just this leap, it's not just this leap that is so interesting it's also again the variety
between different interfaces. These are--both the UNESCO and the UN machine both of them
predictively, a natural language arrangements, and how differently each of them behaves.
They have clusters in different places they--the kind of architecture of the arrangement is
very, very different and again in terms of interpreting why that's a very challenging
active--of working backwards from the arrangement by the 19--late 1980s and into the 1990s we
begin to see Chinese typewriting companies and Chinese typewriting instruction manuals,
the appendix of which used to carry grids like tables that showed you where all of the
characters on the tray bed are, you know, it's a very important thing for a manual to
teach you. By the 1980s and 1990s we begin to see tables like this that are empty because
they're basically saying to the user you decide, you decide where to put the characters. Now
there are some suggestions these characters here, well it's hard to point too, the characters
along the side are saying you should probably--you could probably put characters about Communism
here, you could probably put characters about higher-education here, about war here, about
science here, about and they're--but they're just suggestions ultimately it is left to
the typist to decide where all of these arrangements are taking place. So in essence what this
speaks to me is that machine--the manufacturers of these machines were in effect catching
up with something that was a user lead innovation. So to close I just want to ask a few sort
of set of open questions, the first set of questions is kind of what about the computer?
You'll notice that electrical automation, computing, word processing is noticeably absent
from this talk it's entirely about mechanical typewriting environment and if you go into
the discussions of various innovation by BAIDU, or by SOGO, or by various input methods that
use predictive or algorithmic probabilistic determinations for input. The assumption,
the prevailing assumption is that this was kind of the--thanks to the computer which
like the Data ex Machina comes in and saves Chinese from itself in essence, and what you
see actually when you look into the history of early Chinese computing in the '60s and
'70s is by in large the virtualization or digitization of principles and applications
that had been developed in the mechanical Chinese typewriting environment. That is to
say that at the Chinese Academy of Sciences and various others working in the late '60s
and '70s, they are very aware of what's going on in the mechanical Chinese typewriting environment
and they're trying to basically enhance or build upon that and it's not a question of
this innovation coming in exclusively from the outside and changing the game as it were.
So the larger question I guess is that--and I give this example here in BAIDU, excuse
me, the BAIDU example all input systems not just search engines but all input systems
for Chinese are all make use of predictive text, all of them now, and so there is this
important question of where does this come from in the history of Chinese computing and
Chinese internet. And quite obviously I'm arguing that there's not an exclusive genealogy
that dates back into the history I'm talking about but one that certainly combines the
history of Chinese typewriting and the history of computing more broadly. So if this is true,
the question is a sort of bigger question, is how did we miss the typing rebellion? If
there was this--no Chinese historians in the group so--okay, there's this massive civil
war in China could the Typing Rebellion. Anyway, there is a question of that if this innovation
is taking place in the 20th century then how do we keep missing it? How do we--why do we,
in large part, continue to ignore the history of information technology in China before
maybe five, ten years ago and imagine it as being--in essence more of proof of the anti-modernity
of Chinese than anything we would ever look into to be inspired in our own innovation
and what I would like to suggest is that, not only for my own purposes, but also maybe
even for designers among you who are looking for massive repository of experimentation
and innovation, which may have not seen the light of day but may itself provide inspiration
and blueprints for things moving ahead. I'd like to suggest that the Chinese typewriter
as a problem, historically has operated in very similar type of inspiration as the original
billboard ad back in the day that was designed to kind of challenge and inspire the most
innovative, eccentric kind of thinkers in the world and the Chinese typewriter in it's
own way, throughout the course of the 20th century, became a magnet for not only innovators
in China but from all across the world and same thing with Chinese computing and now
into later technologies. And so it is actually an immense repository of inspiration and principles
that I think we could all profit from looking more deeply into. Thank you very much. I'm
happy to take questions. Why don't we start on the right and then go to the middle.
>> So we speak a little bit about how these typewriters actually work in terms like how
big they are, how do they get set-up because first thing's one of the things about this
is it makes your typewriter a very personal device or you have to arrange it to your--to
your additive [INDISTINCT] sit down at any keyword and you expect [INDISTINCT]
>> MULLANEY: Right. >> And that seems like a very different shift
[INDISTINCT] >> MULLANEY: Right, absolutely. Could everyone
hear the question? Okay. So the machine was roughly--let's--I'll just use my hands, is
roughly about this wide--the tray bed itself is about this wide and this deep, okay, and
then machine itself kind of--may juts out from the back maybe about this far and it
comes off--from top to bottom it would be about yay tall very, very heavy. So the--the
dimensions of it are pretty comparable with many desktop systems we see but the weight
of it is significantly higher but what you're talking about here, this kind of built in--this
built in, almost personal relationship between the operator and the typist--and the machine
is absolutely true. So even before what I'm describing as a predictive--kind of a predictive
turn in the interface, you would have to personalize your machine. Not only in terms of the time
period and whether or not you're work in a bank or a law firm as I suggested but things
like maybe your co-worker or your boss has a surname whose character--the character if
the person surname is within it of itself, very infrequent. It doesn't show up in newspapers
very much but for you, it shows up everyday, every page. And so, you would have to make
sure that your machine has that name. And so, if we imagine a scenario in which, you
know, someone was sick that day and couldn't come in and someone else had to sit down at
the machine. There would be a--before 1949, there would be kind of basic road map. The
person would know that this is organized according to radical stroke but beyond that basic road
map they might not know where to go. They might not know how the other person personalized
it. So, it is a very different interface just as a build--from the GETCO than type writing
in practically any other language that we--that we--that we see. And then after 1949 this
becomes even more so. You would not be able to use really effectively a typewriter that
someone else had rearranged according to their own natural language arrangements. The tray
bed is almost an embodiment of their consciousness. If you think about it, it's how I remember
where everything is, how my body works, how I prefer this to be. And then you come and
try to use it. It's like trying to figure out someone else's file folder structure on
their laptop. It's very, very personal. The gentleman in the--in the sweater and then
to the gentleman with the glasses, yes? >> So you mainly talk about how people with
their group--language, natural language [INDISTINCT] is there any evidence you're reading [INDISTINCT]
>> MULLANEY: I know, yeah. >> That, you know, [INDISTINCT]
>> MULLANEY: That's a great question. Yes? >> Do we have to repeat the question for the
remote the audience? >> MULLANEY: Oh, for the remote audience.
Oh, I'm sorry. The question is--the question is, so the talk mainly focused on how innocent
people with their understanding of and their use of natural language--their own use of
the language reorganized the interface of the machine. Is there any evidence to suggest
that the kind of reorganization of the interface machines cycled back and really transformed
language? The quick answer is, I suspect so. The longer answer is I suspect so but it is
going to take massive corpus analysis to really decide whether or not that that's happening.
We'd have to--we'd have to do stuff that really has never been done with the analysis of Chinese
language material high throop. I mean it really depends upon Chinese OCR becoming better because
we would need top do immense corpus analysis to see whether or not there are changes overtime.
The one thing that maybe countervails against it--against the possibility, so that is to
say that I suspected it's true but one thing that maybe suggested is not--is that the Chinese
Typewriter was mainly a re-inscription device, a re-productive advice. It was--you would
start out with a manuscript or something and you would copy it, you would produce a type--a
type script version of it. What you don't see in the context of Chinese Typewriting
as opposed to practically every other typewriting context, is then idea of the author sitting
down at a machine and producing a manuscript, like there's no such thing as Lin--the writer
Lin Yu Tang in China sitting down in his Chinese typewriter the way that Allen Ginsberg might
sit down. And this comes back to the common usage system that not all of the characters
are there. And it is the job of the poet and the job of the writer to in fact to delve
into the infrequent and the beautiful. And so, maybe not but I think that that's one--that's
one area that I'm trying to team up with distant reading and corpus linguistic specialist to
see if it's possible to ask that, to ask that question about linguistic change. The gentleman
in the glasses and then the gentleman in the black leather.
>> So, there was also this [INDISTINCT] the simplified Chinese. Do you think to wish to
simplify typing and wish to simplify writing would be correlated--basically correlate the
[INDISTINCT] >> MULLANEY: Sure. The question is, you know,
that the relationship between simplification campaigns and then this kind of taxonomic
reform of the--of the tray bed are there any relationships? Because they are happening
at the same time. There are happening in the early 1950's under the--in during Maoism they
take--they take place in the '50s. There was an earlier aborted effort to simplify characters
under the Republican regime but the places where they seem to be related is that the
typewriter became--and the tray bed in particular became a highly politicized space or domain
if we think about because for example, during the simplification campaigns, there were no--there
were notifications sent to typewriting manufacturers saying that you need to replace these traditional
characters with simplified characters now. Which from their perspective is a monetary--you
know, is a financial burden--there's a financial burden on people at the local level to--that's
theoretically half to replace the character with another one that has been identified
as the simplified version. And, so people were reticent to that because it just--it
just kind of cost money and was a--was a--was pain. And so, it became one of the places
that the--that the--that the communist government was very eager along with printing presses,
a very eager to promote simplification because these were obviously the places that produced
texts. So, you want to go to the--if you can change the characters on a typewriter and
change a character on a type--a printing press then you therefore transform every piece of
every page that gets produced thereafter. So, in terms of them interacting or relating
to each other causally I don't see that all that much but they are connected historically.
Yes, sir? And then the gentleman. >> So, underlying the couple of assumption
that [INDISTINCT] I grew up in this country and there wasn't a lot reading the newspaper.
So, the language became very I would say [INDISTINCT] so, I wonder if there's any study on how much
it affected the language. Because it proves to make it more [INDISTINCT]
>> MULLANEY: Right. The question-if I can--if I can synthesize it, let me know if I get
it right. Is--I'd say it is the question, is there--is there a relationship between
the kind of templetization--I like that word, of language in Mao--in Maoist China which
relates to kind of--the kind of Newspeak phenomenon that happened in many communist context and
this--that because the language was in essence more predictable, that it was more amenable
to a predictive text arrangement. And the answer is--I mean I think yes. You know, in
a longer paper, in a longer project that's actually an entire part of the argument is
that the emergence of a un--of a--of a--of a political rhetoric, of unprecedented standardization
during Mao--the Maoist period of unprecedented kind of routedness was itself largely facilitated
the creation of a tray bed in which it would be very useful to create a natural language
arrangement because, you know, that you're going to be saying American imperialist--the
American imperialist over and over again. What's facilitating about it I think is that,
normally the idea of Newspeak this kind of 1984 image of the communist rhetoric is typically
understood as an inhibition as an inhibiting factor to innovation. It inhibits thought
because you can--it kind of, you know, that's a--it's double plus good kind of idea of language.
And it seems that perhaps there is--there's a lot of credence to that argument. But in
this case, the--we see the exact opposite. That for the--for at least in this context
Newspeak is inspiring or part of a history of radical techno-linguistic innovation in
which there is a--there is a kind of encouraging or inspiring momentum for the typist to make
the tray bed as idiosyncratic and personally optimized as possible. So, that they can more
and more efficiently produce texts that are more and more route if you see what I mean.
So, there's a--there's--do you see what I mean? There's actually--there's actually a
relationship between innovation and Newspeak that we--that we rarely see another context
but that is highly relevant--highly relevant here I think. Yes? In the back and then--in
a black sweatshirt. >> Okay. You mentioned...
>> MULLANEY: Oh, I'm sorry. >> You mentioned various factors [INDISTINCT]
how about spoken dialect, is that [INDISTINCT] >> MULLANEY: The question pertains to spoken
dialect and if I can--if I can broaden it out because it is a really interesting questions
about--I mean really the question of dialect and Chinese language information technology
as a whole is a very, very political question throughout really the 20th century. For the
most part, there has been a kind of antagonistic relationship between Chinese information technology
and dialect. In the sense that people had been political leaders manufactures and others
had been very resistant to the idea of developing for example, Cantonese input or Shanghainese
input or Taiwanese input or to--even before that--before we think of input and whatnot.
The question of whether or not different dialects in Chinese should be outfitted with their
own written languages and whether or not people should be able to publish for example a book
in Cantonese where the characters are actually--there are many Cantonese characters which do not
exist in Mandarin Chinese should people be allowed to publish in that way which is also
a kind of information technology question and the answer has consistently been, no because
if we--if we pair up the massive diversity of dialect diversity in China with written
form we will in some sense institutionalize these cleavages and that could actually damage
the territorial integrity of the country. That's why most regimes in the 20th century
in China had been very, very--and another elite, political elites have been very focused
on the promotion of a single standard. And so for example with pinion with Chinese pinion
inputs that is actually a--not only an information technology tool but a pedagogical tool that
people know that if they're going to use a computer in China they have to use pinion
input and pinion input is route is connected directly to the Chinese standard. And so the
very active learning how to use a computer is in some way shaper form becoming oriented
towards the standard. And so they've been very resistant to develop alternate techniques
for that. And let me just jump to you I'm very sorry.
>> Yes. I'm trying to kind of like you to get the kind of answer I really [INDISTINCT]
did you try to--kind of figure out how they actually basically try to reverse the communication
process applicable to this and how the actual typist make this happen? [INDISTINCT]
>> MULLANEY: Right. >> But how did they do it with like the numbers
they're very equipped [INDISTINCT] very large numbers.
>> MULLANEY: Right. >> And how do they arrive at that, I mean
[INDISTINCT] I mean whenever we [INDISTINCT] too far things, we just Google near each other
[INDISTINCT] >> MULLANEY: Yeah. So the question in this--the
basic question is how did these re-engineers of their machine do it, how did they go about
the process of producing these natural language arrangements, tray beds that's actually why
or how I came to be here today was I gave a talk at Stanford which is my home institution
and it was within STS and STS contacts was attended very luckily for me by Scott Clemer
from HCI at Stanford who then invited me to give a talk there and I was talking with Scott
and I said I need to, I need to talk to someone who works on optimization because I need to
know how someone right now would think about the optimization of such a hyper-complex system
not because I want to build a Chinese typewriter but because I want to hear someone talk about
the process of optimization so that I can--so that it can excite my imagination and excite
my sort of heuristic complex so that I can go back and have more intelligent--search
more intelligently for exactly the answer that you're talking about. It is entirely
a question there's two pathways to answering the question, I don't have an answer yet that's
going to be--that's the kind of reason that I actually first emailed I'm sorry--to complete
the story, then Scott said you need to talk to Xu Man Jai and then he put us in contact
we had a really wonderful conversation and then that manifested itself in today but I
have to admit with the tools of my disposal I'm at a loss because there I can try to reverse
engineer from the tray beds that I have and that's where I really need to pair up with
someone who thinks about those problems everyday and then the other thing that I can line up
is ethnographic or oral-historical interviews with typists. But even that is a very challenging
thing, tell me back in 1970 how you decided to move character one to space x and character--that's--you
know that's embodied memory. They--that's not something that they're going to written
down somewhere so even that is a incredibly challenging process but that's I think an
incredible pay off not just for me as an historian but for designers who are looking for inspiration
in my opinion. The gentleman on sorry--well actually I'm sorry maybe if I can keep the
order and then resume in. Yes. >> [INDISTINCT] it seemed like, you mentioned
[INDISTINCT] and character may have to be [INDISTINCT]
>> MULLANEY: Right. >> To be here you have to lose every character.
>> MULLANEY: Yes. >> You have to actually hit the [INDISTINCT]
>> MULLANEY: Right. >> You have to put that actually [INDISTINCT]
like [INDISTINCT] >> MULLANEY: No. No. Yeah, yeah, yeah. So
the question pertains to before 1949, before the natural language rearrangements here.
Even--how did the process of optimization of tray beds in the Republican period take
place and the process that you just describe is exactly what would've been required there
is no, I've seen no tray beds where some characters just strangely out of place and--but everything
is always in place so if you think about it the closest analogue we have is the kind of
that gain with the missing plastic with a plastic pieces with one piece missing and
you've got to move it around, you move one thing everything must move, everything must
move to fill in the space and so yes so even so that's an important dimension here is that
even before the '50s there is some--there is a--there are dimensions or characteristics
that are built into the tray bed itself and into this particular techno-linguistic apparatus
that provided a context with a more radical transformation we see in the '50s. But every
tray bed I've ever seen exhibits no laziness in terms of some weird character out in the
middle of nowhere so yeah. Yes? >> Tom, as historian with a many to sustain
such a political context to major changes to the institution but much of the results
could be looked at or design purely problem of language or efficiency point of view just--or
even in purpose we can optimize the arrangements so you had pasted on [INDISTINCT] so I wonder
with your [INDISTINCT] is to wonder how important these political, social background.
>> MULLANEY: Right. That's--I feel like--I feel like I owe you something because I--you
gave me a chance to kind of mention one slide that I skipped which was very important which
was the fact that--let's see if I can find it very quickly--that even in the Republican
period there was one small part, remember the--remember the--remember in the heat map
there's a one little bright red part there was a cluster there was one tiny, tiny region
of the tray bed roughly two cells wide by maybe 10 cells tall so out of 2500 roughly,
it's a very small a region which were arranged according to natural language arrangements
and these were province names. So the example up here has a let's look at the character
Jhang that's highlighted in yellow which means river Jhang can be combined with diagonally
Sou create a province name Jhang Sou. It can be combined with the character to it's left
Ju for Ju Jhang which is another province and then it can be combined through this kind
of neat kind of knight like a chess kind of thing of Halong Jhang from the bottom to the
top to the right. This was the one, there's few important things about this. One is that
we know both here and in very esoteric discussions by some language reformers that people were
thinking about this in the '30s, they were thinking about this. They were kind of--it
was in the general swirl of possibility as a known thing and at the same time we also
know that in the '20s and the '30s was a time of hyper innovative, hyper iconoclastic language
reform. This was the time when people were saying let's get rid of all the characters
and replace it with French. Let's simplify all the characters let's replace it with a
symbolic notation system. Very, very crazy sort of ideas and yet even with this context
the step towards the kind of radical departure towards this new arrangement was never, never
tried to my knowledge and I've seen a lot of these machines now more than I cared to
admit and so there seems to be--it seems to be that if this is purely a question of efficiency
and optimization that we would have seen something far earlier because people would've known
it [INDISTINCT] that it was just something was keeping the dams, the water behind the
dam and that it really took the--it really took the '50s and it really took Maoism and
it really took this idea of the kind of mass overturning knowledge systems to lay dynamite
to that dam and just say okay everybody, do whatever you want to the tray bed, do whatever
you want however you see fit and then from that moment on it takes place because in the
people's daily it shows up in three different places, it shows up in dozens of other places
and not everyone does it but suddenly it starts to spread out until by but again that slide
from the 1980s shows that even typewriter manufacturers were like okay you guys are
better at this than we are, you have better ways of optimizing your machines than we could
ever come up with so the semiotic interface is yours basically. So I do think that there
is, I thought about that for a long time and ultimately I came to the conclusion that the
last piece of the puzzle on this is counter intuitively Maoism.
>> Unfortunately we reached the end of the--this--[INDISTINCT] talk. Let's thank again for Tom. Thank you.
For those of you who want to--thank you very much.