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