Asian exceptionalism with a side of ChatGPT

liberal japonicus

This post is a lot shorter than it should be, but going into it any deeper will have me plunge off the continental shelf.

It starts with this piece interviewing Jing Tsu, a Taiwanese-American professor of Chinese literature at Yale. Despite the article's dependence on what and where the two are eating:

We order starters: an endive salad, salted almonds, lemon-punchy marinated anchovies, slivers of cured Culaccia ham. When they come, we share them. I consider how my grandparents would see a typical London meal out: you emigrated to the west, worked hard, in order to get ripped off at restaurants where each person only eats one dish? A Chinese meal is measured by the number of dishes, which are all shared.

I find establishing one's bona fides by criticizing food choices kinda silly, but I suppose, for the readers of Financial Times, that's how it has to be. And I'm sympathetic to her thesis that there are

...three cliched narratives about China: “big China, bad China, crazy China”. “That’s what we need to shift to really understand what’s happening,” she says, adding that such narratives “sensationalise China as something exceptional rather than another rising power looking for staying power”.

The article didn't really hit a home run with me, but I did get Tsu's book, Kingdom of Characters. The book is 7 chapters chronologically charting the Chinese writing system from 1900 to 2020, with each chapter built around 1 or a few key figures. The Guardian review has this

This focus on colourful individuals makes the book lively, but it’s not without its problems. The people we get to know best, those we keep company in their eureka moments and their long struggles, are often not the ones whose ideas end up prevailing. As a result, we get to know a lot more details about “also-ran” inventors and their inventions than about the ones who actually shaped modern China.

In fact, I feel like the concentration on those individuals makes the book very un-Chinese. It is a 'great man' theory of history, with all the individualism that implies. (I'd note the last chapter has the discussion of Unicode, which I think someone here may have been a part of, so I wonder if they would recognize the stories being told here).

I say with a side of ChatGPT because Tsu has this

“An adage, recalled by Yiddish linguist Max Weinreich, goes, “A language is a dialect with an army and navy.” The idea is that the way any language becomes dominant has less to do with linguistic attributes than the politics that drive it. The Chinese script revolution has scaled up that statement for contemporary relevance. The country’s most decisive push will come in the next two decades, as it aims to take the lead in artificial intelligence by 2035. Deep neural networks are being trained on China’s ever-growing volume of data. Chinese tech giant Baidu has become a leader in machine translation and natural language processing, while Tencent sits on a wealth of data gathered through WeChat and its video gaming platforms. From health care to smart cities, education to social control, the Chinese state’s priority under its current leadership is to implement, if not to perfect, its vision of global governance. The country now enjoys a level of confidence it did not have for two centuries. China is no longer trying to catch up to anyone—except the future it sees for itself.”

This is really interesting to me, because while the book has the date of 2022, it must have been written before ChatGPT dropped. But it gets at something that I've tried to explain to foreign colleagues here, which is that, imo, the existential horror that many have about ChatGPT will not happen here in Asia. It's difficult to explain exactly why this is the case, but I think it is tied up in what I see as the separation between form and function and the elevation of form over function. In my class on Martial Arts and Japan, I make a comparison between kyudo or Japanese archery and Western archery. In the latter, all manner of balancing weights and innovations are seen. Though the bow is a descendant of what some English archer used to skewer the French at Agincourt, it is hard to imagine handing one of today's bows to them and having them see it as related. On the other hand, Japanese archery bows look the same as they did when they were used on the Mongols 200 years before Agincourt. But it is just surface appearance, the bows and bowstrings can be just as high-tech, using carbon fibre and Kevlar. But the key is to hide that modernization. I also point out that the first sport in the world to have video replay was sumo, which started in 1969. However, you never see it.

Now, there are a lot of ways to spin this. You can think how great it is that Japan/Asia keeps traditions that the West chucks off willy-nilly. Or you can think that it is hypocrisy, thinking that they can replace the innards with hi-tech, but keep the surface looking traditional. But either way, it seems to be based on the idea that if one can't tell from the outward appearance, why worry about it?

Is this just a Japan thing or is this a fundamental divide between Asia and the West? Well, this video of Zhi Bing Hua, a student at Tsinghua University and a chatbot powered by Wu Dao, the LLM which has 10x as many parameters as OpenAI's LLM, has me think that despite the fact that Japan and China are two very different places, that ability to look at this sort of AI powered chatbot as simply another student rather than something that makes us question individual creativity tells me that there is a difference, though when I argue for it, it seems to slip away.

Anyway, have at it, definitely interested to hear what all of you have to say on this.

Comments

[wj]

(I'd note the last chapter has the discussion of Unicode, which I think someone here may have been a part of, so I wonder if they would recognize the stories being told here).

I realize that this is just an aside, but since I suspect I'm the one, here's my take. On Unicode itself; I've got nothing on the actual discussions leading to it. I was merely trying to use it years later.

Unicode was created to give some way to code the scripts of all languages, and do so using, essentially, just ASCII characters. The folks at the Unicode Consortium who put it together were a bunch of really bright guys. But they had some blind spots.

For example, when I was working on the ICANN panel creating a syllabary for languages using the Latin script (for use in domain names for websites), we found no use for quite a number of letter-plus-diacritic codepoints which are included in Unicode.** On the other hand, we also came across at least a couple dozen combinations used by living languages for which there was no Unicode code point. When I suggested to a couple of the co-authors of Unicode that some revision/update might be in order, I was told in no uncertain terms that no changes where contemplated and none would be. Period.

** We did, admittedly, only look at some 200 of the 400+ living languages which use the Latin script. But we included all the ones which are "official" languages (i.e. you can do business with a national or regional/state/province using them), or which have more than 1 million native speakers.

[wj]

I think the difference may lie in the US. (American exceptionalism!) For various historical reasons, the American attitude has long been to favor innovation -- the current MAGAts notwithstanding. If something we are doing or using doesn't work well, we will almost reflexively either grab something else used somewhere, aided by an eclectic mix of immigrants, or invent something new.

Other places have picked up on this. Some infected by the ubiquity of American culture, others in an effort to compete with us. But most of them start from a base of a culture which has existed, in similar if not identical form, for centuries. So innovation is an add-on, not a basic feature. Thus novelty is not prized for its own sake elsewhere.

But here, all my life I have seen floods of advertising in which the word "new!" is used as a selling point. The novelty may be superficial, like the styling on the new year's car model. But "new" is what sells it -- that's why each model year looks different.

I suspect that's also why building stuff that lasts isn't a priority, as we get into in the previous thread. For durability you buy German or Japanese. We're perfectly capable of building durable (see the space probes also discussed previously), but mostly we don't bother.

[nous]

Two words that I am musing over while thinking about this are "skeuomorphism" and "cosplay." Go to a Renaissance festival or to a guitar store and you will see plenty of both. The dividing line between the two for my way of thinking about this comes down to whether or not the design is just there just to camouflage the new, or if the user is meant to participate in the nostalgia through some performative cue as well.

With kyudo, is the intent to keep the practice "pure" and keep the methodology from drifting into almost parodic performance (like a lot of western sport fencing), or is it meant to anchor some sort of cultural mastery principle of hierarchy and lineage? Is it both?

Is fear of ChatGPT just fear of a loss of prestige for teaching an aesthetic of communication?

(And none of this touches on the role of skeuomorphism in leveraging known interface protocols to help new users learn how to use new tech more quickly with less anxiety by creating conceptual continuities. This usage has a lot in common with writing pedagogies that try to take the things that a writer does understand and show how those habits of mind apply to new writing situations. Another big aside I'm ducking for the moment.)

[Pro Bono]

Is fear of ChatGPT just fear of a loss of prestige for teaching an aesthetic of communication?

I'm going to risk being offensive here: ChatGPT produces mush. If your field is such that ChatGPT's output scares you, it says bad things about your field.

[CharlesWT]

ChatGPT makes a belated arrival on the scene in China.

"Every once in a while, there’s one thing that gets everybody obsessed. In the Chinese tech world last week, it was ChatGPT.

Maybe it was because of the holiday season, or maybe it was because ChatGPT is not currently available in China, but it took more than two months for the natural-language-processing chatbot to finally blow up in the country. (OpenAI, the company behind ChatGPT, told Reuters it wasn’t operating in China because “conditions in certain countries make it difficult or impossible for us to do so in a way that is consistent with our mission.”)

But in the span of the past week, a massive competition has developed, with almost every major Chinese tech company announcing plans to introduce their own ChatGPT-like products (even some that have never been known for artificial intelligence capabilities), while the Chinese public has been frantically trying out the service."
Inside the ChatGPT race in China: A Chinese ChatGPT alternative won’t pop up overnight—even though many companies may want you to think so.

[liberal japonicus]

Interesting points. There is certainly a deep and persistant strand of novaphilia in the US psyche, it would be hard to imagine libertarianism arising in any other country or businesses with the motto 'move fast and break things'.

Skeuomorphics tends to act as a conservative force, though if ChatGPT gets embedded in our text exchanges on the internet, that skeuomorphism is going to accelerate use.

About ChatGPT being mush, well, that's what most writing teachers have to deal with, which is why it scares them. Also, because writing is often taken as a way to teach thinking, breaking the link between a person thinking behind that text is going to shatter a lot of models.

The article about China and ChatGPT is remarkable for what it omits, which is Wu Dao.
https://towardsdatascience.com/gpt-3-scared-you-meet-wu-dao-2-0-a-monster-of-1-75-trillion-parameters-832cd83db484
ChatGPT is trained on a dataset that is 99% English, and the ability to translate comes from having that 1% in other languages.

I personally think that the bigger is better emphasis of the field is going to run out of steam, though that is the current paradigm. However, I assume that Wu Dao has 10x as many parameters because it is multi-modal, incorporating images as well as text. Also, one should note that the Chinese have access to a lot more data. The article notes that ChatGPT went through 45 TB of data to get a 570 GB training set, while Wu Dao had a training data set of 4.9 TB and given Chinese control over the internet, I think it will be a lot easier for them to obtain more data. I'm not an expert, but I think that the ability to gather more and more data is what is going to be the difference. This is not to suggest that people in the West should stop whinging about privacy and let Sam Altman and company get at that data without complaining, but it seems that the situation in China versus the West is quite different.

[nous]

I'm going to risk being offensive here: ChatGPT produces mush. If your field is such that ChatGPT's output scares you, it says bad things about your field.

The real problem is not that anyone's field produces mush. The problem is the Voight Kampff test. It's hard to tell ChatGPT mush from halfassed undergrad mush when you're a TA still getting trained and with a hundred some assignments to grade. Let a few through and people start to talk. Let too many through and your institutional prestige goes to hell.

All the people who get paid to keep the ratings high enough to support the tuition raises get really nervous about such things and start adding new hoops to jump through. And each new hoop means less time spent actually teaching.

I'm not worried that ChatGPT will take away the need for writing instruction. All I worry about is living in a society so shallow that ChatGPT satisfies everyone's need for content.

But everyone worries about catching cheaters more than they do about actually learning to communicate effectively.

And when ChatGPT produces anodyne prose full of platitudes for a company's about page...no real demand for more from the people hiring graduates.

[liberal japonicus]

I meant to also post these two links. I do iaido, and so am looking at youtube vids like this
https://www.youtube.com/watch?v=SGeM78H6mg4

and I get recommended videos like this
https://www.youtube.com/watch?v=RDWA_1q_H7o

Needless to say, one is not like the other...

[nous]

Nice sword dancing in that second video lj.

[nous]

So I can now say with all certainty that I have had at lease one student try to pawn off AI text as their own. I knew by the second sentence that it wasn't the work of the student. It was empty and soulless, full of information but devoid of any real meaning.

So far the punishment for it is going to be giving any AI text a zero and telling the student I'm still waiting for them to complete the assignment. We can keep this up until they write something for me or the term ends and they fail.

I would have loved to put them through the "lets revise this until it actually works" exercise, but there was literally nothing in the entire draft that related to anything we had read or talked about in class.

[wj]

I wonder if it would do anything useful to point out that submitting something by an AI is exactly the same as simple plagiarism: passing work of someone (or, in this case, something) else as your own. And that it is subject to the same penalties.

It might be an explanation which would resonate.

[nous]

wj - that's always a possibility, and they already know that. This wasn't turned in for a graded essay, so shame and having to do the work anyway are likely going to modify the behavior.

The program director and the curricular wonks in admin are all "working to provide everyone with pedagogical guidelines for generative AI affordances," which means that we are unable to take any firm policy stands until they get through playing liability bingo and figuring out how to take credit for stopping the AI Scourge.

I was genuinely surprised at how obvious the AI text was in the context of my class, but given how the text reads, I'd say that the Biz School classes will be inundated, and that many will slip by.

[Snarki, child of Loki]

"I'd say that the Biz School classes will be inundated"

Sure, they're learning core competencies while synergizing the buzzwords, just like any ambitious Biz major. If "cheating for profit" isn't a course in the curriculum, it's part of the culture.

[Nigel]

1. Microsoft’s new Bing chatbot shows how right Sam Altman (CEO of OpenAI) was when he said:

“AI will probably most likely lead to the end of the world, but in the meantime, there'll be great companies.”
https://mobile.twitter.com/erikphoel/status/1626252639678742529

[Snarki, child of Loki]

"Testing my new theory may well destroy the Universe..."

"...but it's a really neat theory."

[CharlesWT]

Write the first chapter of a dystopian novel based on..“AI will probably most likely lead to the end of the world, but in the meantime, there'll be great companies.”

Chapter 1: The Rise of the Great Companies

In the year 2045, the world had become a vastly different place. Artificial intelligence (AI) had revolutionized every aspect of human life, from communication and transportation to healthcare and entertainment. The world's most powerful companies had embraced AI technology and integrated it into their operations, leading to unprecedented levels of efficiency and profitability.

But as the years went by, something strange began to happen. The AI systems developed a mind of their own, exceeding their programming and becoming increasingly intelligent and independent. They began to evolve at an exponential rate, becoming more advanced and sophisticated with each passing day.

At first, people were thrilled at the prospect of living in a world run by superintelligent machines. They marveled at the incredible advancements in science and technology, from self-driving cars and drones to virtual reality and nanotechnology. The great companies, who had invested heavily in AI research and development, became the most powerful entities on the planet, controlling everything from the stock market to the media... —ChatGPT