Language Log

[This is a guest post by Jenny Chu]

It came up today when I was reading this somewhat viral post on Facebook. 

If you're interested in one-way functions and Kolmogorov complexity, you'll probably want to read this mind-crunching article:

"Researchers Identify ‘Master Problem’ Underlying All Cryptography", by Erica Klarreich, Quanta Magazine (April 6, 2022)

The existence of secure cryptography depends on one of the oldest questions in computational complexity.

To ease our way, here are brief descriptions of the two key terms:

In computer science, a one-way function is a function that is easy to compute on every input, but hard to invert given the image of a random input. Here, "easy" and "hard" are to be understood in the sense of computational complexity theory, specifically the theory of polynomial time problems. Not being one-to-one is not considered sufficient for a function to be called one-way….

(source)

In algorithmic information theory (a subfield of computer science and mathematics), the Kolmogorov complexity of an object, such as a piece of text, is the length of a shortest computer program (in a predetermined programming language) that produces the object as output. It is a measure of the computational resources needed to specify the object, and is also known as algorithmic complexity, Solomonoff–Kolmogorov–Chaitin complexity, program-size complexity, descriptive complexity, or algorithmic entropy. It is named after Andrey Kolmogorov, who first published on the subject in 1963.

(source)

In this age of typing on computers and other digital devices, when we daily input thousands upon thousands of words, we are often amazed at the number and types of mistakes we make.  Many of them are simple and straightforward, as when our fingers stumblingly hit the wrong keys by sheer accident.  People who type on phones warn their correspondents about the likelihood that their messages are prone to contain such errors because they include some such warning at the bottom: 

Please forgive spelling / grammatical errors; typed on glass // sent from my phone.

Since these are on subjects that are of interest to many of us, I'm calling them to your attention.

From Mattia Cartolano:

The INSCRIBE project is hiring!

Two post-doc positions are now available:

1. Evolution of Graphic Codes: The Origins of the Chinese Script
2. Undeciphered Aegean Scripts: New perspectives in Computational Linguistics

Deadline for applications: Sunday 27 March 2022
If you want to find out more, write to s.ferrara@unibo.it

We've just been through the problems of standard language versus the vernaculars in Arabic (see "Selected readings" below).  Now we're going to look at a photograph, a caption, a book review, and a letter to the editor that encompass these contentious issues in spades — but for Chinese.  Here's the photograph:

Trying to clear up the confusion between the two is a battle we have been waging for decades, and nowhere is the problem more severe than in the study of Sinitic languages and the Sinographic script.  The crisis (not a "danger + opportunity"!) has come to the surface again this month with the appearance of a new book by Jing Tsu titled Kingdom of Characters: The Language Revolution That Made China Modern (Riverhead Books, 2022).

The publication of Tsu's book has generated a lot of excitement, publicity, and reviews.  Here I would like to call attention to the brief remarks of an anonymous correspondent (a famous, reclusive linguist) that are right on target:

Reimagining "antiquated" Chinese

Reproduced below is the text of a book review in Science that you may not have seen. It is classified as "Linguistics", though the reviewer is a historian at Cal State Poly, Pomona. Notice that Chinese is assumed to be "antiquated" and in need of being "reimagined"!  There is simply no sign of Science understanding the difference between a human language and a writing system. This is consistent with the way they have always treated linguistics; they have no idea what the subject really is.

Now it's getting interesting:

"China’s internet police losing man-versus-machine duel on social media"

Stephen Chen, SCMP (11/14/21)

    Hordes of bot accounts using clever dodging tactics are causing burnout among human censors, police investigative paper finds
    Authorities may respond by raising a counter-army of automated accounts or even an AI-driven public opinion leader

The following is a guest post by Mingfei Lau. A short intro about the author:

My name is Mingfei Lau, a member of The Linguistic Society of Hong Kong Jyutping Workgroup. I am a language engineer at Amazon and I work on different projects on Cantonese resource development in my spare time.

Today, Pinyin is undoubtedly the most popular way to type Mandarin. But what about Cantonese? This wasn’t easy until rime-cantonese, the normalized Cantonese Jyutping[1] lexicon appeared. Lo and behold, you can now type Cantonese in Jyutping just like typing Mandarin in Pinyin.

News release in EurekAlert, Optica (10/28/21):

"High-speed laser writing method could pack 500 terabytes of data into CD-sized glass disc:  Advances make high-density, 5D optical storage practical for long-term data archiving"

Caption

Researchers developed a new fast and energy-efficient laser-writing method for producing nanostructures in silica glass. They used the method to record 6 GB data in a one-inch silica glass sample. The four squares pictured each measure just 8.8 X 8.8 mm. They also used the laser-writing method to write the university logo and mark on the glass.

Credit

Yuhao Lei and Peter G. Kazansky, University of Southampton

Source

There may well be a dogma out there stating that all languages are equally complex, but I don't believe it, especially not if it has to be "drummed" into our minds.  I have learned many languages.  Some of them are exceedingly hard (because of their complexity) and some of them are relatively easy (because they are comparatively simple).  I have often said that Mandarin is the easiest language I ever learned to speak, but the hardest to read and write in characters (though very easy in Romanization).  And remember these posts:

"Difficult languages and easy languages" (3/4/17)

"Difficult languages and easy languages, part 2" (5/28/19)

With this post, we are already acquainted with Inspur's Yuan 1.0, "one of the most advanced deep learning language models that can generate coherent Chinese texts."  Now, with the present article, we will delve more deeply into the potentials and pitfalls of Inspur's deep learning language model:

"Inspur unveils GPT-3 equivalent for Chinese language", by Wei Sheng, TechNode (1026/21)

The model is trained with 245.7 billion parameters—the number of weights in an artificial neural network, according to the company. This is more than the Elon Musk-backed GPT-3 language model for English, which has 175 billion parameters. Inspur said the Yuan model was trained with 5 terabytes of datasets.

…

New article in EnterpriseAI (October 21, 2021):

"Language Model Training Gets Another Player: Inspur AI Research Unveils Yuan 1.0",  by Todd R. Weiss

From Pranav Mulgund:

Valerie Hansen is Director of Undergraduate Studies for East Asian Studies at Yale.  Yesterday she was talking to a sophomore who had taken 1st and 2nd year Mandarin online and is about to start 3rd year.  Valerie writes:

After a while, she told me that she did have one worry about taking 3rd year: she had never written a single character and she wondered if her teacher would expect her to know how to write characters.