Greetings from a world where…

the dream lives on because of those who paved the way

…Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors). As always, the searchable archive of all past issues is here.

Feature Translation: Algorithmically-Monitored Workers

Context: A deeply-sourced longread (link to original) on office automation in China, published last week by Caijing Magazine, a respected business and financial news platform known for publishing investigative, critical pieces. A sampling of the key passages follows, though I’d highly recommend checking out the full translation, which is nearly 6000 words.

The Hook:

In mid-December 2020, Shanghai saw its first cold snap. The northwest monsoon winds blew through the streets, urging pedestrians to wrap their coats tighter. At 11:30 in the evening, in front of a large private construction company on Caobao Road, Xuhui District, more than a dozen figures hurriedly opened the door with their work ID cards.

The group of people went upstairs in two elevators. They were talking and laughing just before, but at this time they were collectively silent. A young man tried to keep going with an inside joke but was stopped by his colleagues with stern eyes.

"The elevator's camera is connected to the office automation (OA) system, as well as recording equipment," a senior engineer Fu Cheng whispered after stepping out of the elevator. "This matter was kept confidential, but a guy who had left the job secretly told me."

The nuts and bolts of OA:

This office system is internally called the "Intelligent Task Distribution System", which the company commissioned an information technology company in Shanghai to develop. The developer claims that this system is developed in accordance with the characteristics of the "project system" of construction engineering companies. It has a powerful machine learning program that can automatically assign tasks to office groups according to the time nodes of each link of the project. It can also "improve employee work efficiency" by linking together cameras and punch cards throughout the company.

But the front-line employees' feelings of being managed by the system are not so good. The general reaction is that the system "does not consider people's feelings."

Supervisors needed no longer?

Wu Hao is the workshop supervisor of a foreign-funded electronic equipment manufacturing company in Wuxi…During the annual recruitment season, Wu Hao had the final decision on the employees assigned to his workshop. In his daily work, he was also responsible for the daily management, guidance, training and evaluation of the workshop employees. "I determined the daily tasks and performance indicators."

In the past, relatives in his hometown would always find him, begging him to hire their children into the workshop, and some workers and technicians secretly asked for "extra leniency" in performance appraisal and leave. All these made this man from the northwest "have a sense of accomplishment and work harder."

In 2019, the foundation of Wu Hao's workplace suddenly collapsed: the company began to introduce a fully intelligent employee management system.

"It is said that this was a command issued from the US headquarters. It must be enforced," said Wu Hao. "The technical department found an information company in Suzhou, developed it for half a year, and installed it in the factory."

The workshop terminal of the new system includes more than 20 cameras all over the workshop, an OA system that records all worker information, and an electronic attendance system at the entrance of the workshop. They have different functions:

The camera records the on-the-job status of all workers in the workshop and monitors their work efficiency. For example, each component has a prescribed processing time, and the system will recognize the worker's actions through the camera screen. If it takes too long, it will be reflected in the assessment of the worker at the end of the month, and performance merits will be deducted accordingly. The attendance staff will also manually review the system’s assessment results…

Some interesting insights into the benefits of guanxi practices

…Since the system has been running for more than a year, fewer and fewer people have engaged in guanxi practices with Wu Hao. This would appear to be more fair, but Wu Hao doesn't think so.

In his opinion, the close relationship between himself and the frontline workers has begun to be indifferent. In the past, he would always take care of workers who had urgent matters at home, such as not going through the process to grant them half-day leave to prevent them from using up the leave quota, and also by giving clever workers extra performance to encourage innovation in the assembly line and win over the hearts of the people…

…An event in the summer of 2019 was still fresh in Wu Hao's memory: a worker took 3 seconds to go to the bathroom beyond the prescribed time (the requirements for going to the bathroom are so precise? Yes, this is the most darkly humorous thing I have encountered in an interview, even going one second over will not cut it), the system deducted 50 RMB from his salary at the end of the month. The worker lodged a complaint, and Wu Hao called the human resources department to report, receiving a cold, one-sentence reply: Exceeding the time limit by one second is still exceeding the limit, and the rules and regulations must be followed.

This incident quickly spread throughout the workshop. "Old Wu is of no use anymore, now the people above only look at the computer." Wu Hao heard this type of comment while eating in the cafeteria.

Features some quotes from leading Chinese academics on OA, like this one:

Lingyun Qiu (associate prof. at Peking University), mentioned that existing research has found that if employees can participate in the algorithmic decision-making process, they will be more willing to follow instructions.

“If employees are allowed to make some simple inputs (to the algorithm), similar to even parameter settings, they will feel that they are involved in the decision-making process, which will reduce their feelings of alienation toward the algorithm, and they will not feel that they are completely controlled by the algorithm." But at the same time Lingyun Qiu pointed out that this requires companies to establish an effective feedback loop. “For example, if an employee is scheduled to work overtime for three consecutive weeks by the software, the system needs to design a feedback function—whether through the supervisor or the employee themselves —that allow employees to express their dissatisfaction with the algorithm.”

The tricks of the trade:

Recently, a netizen broke the news on Weibo that a technology startup in Hangzhou issued a batch of high-tech cushions that could sense body data to employees and required all employees to use it. Unexpectedly, employees discovered that the cushions were used to monitor employees' heart rate, breathing, sitting posture, fatigue and other data to investigate whether employees were "loafing on the job", which caused a lot of complaints…

…But most employees are not aware of this (a large state-owned power company’s policy to track employees movements through employee cards). “The employee card has this function. We didn't know it before, but it's not surprising,” an employee said, "The current technology can completely guarantee that we have no privacy after entering the company."

Since money will be deducted, why would employees not know?

"Because the money deducted is the year-end bonus," Liu Fei (an employee of the company’s Party-Mass Work Department) uncovered the mystery, "Would ordinary people notice that the year-end bonus is short by 80 or 100 RMB? They would think it was for some tax expense." He said that this has already been used as a "secret" cost control method for several years, and it can save the company nearly one million RMB every year.

Liu Fei said that companies with similar operations are "not uncommon" in the industry: "We learned this from other industry partners. As far as I know, they also learned from some private enterprises."

Feifei Wong, an employee of a private scientific research company in Shanghai, encountered a similar incident. In August 2020, she chose to leave her job, but she did not expect the company to ask her to pay a loss of hundreds of yuan: "According to the company's assessment, leaving the park without authorization during working hours is considered a violation of labor discipline, and the data must be entered in the information database." Feifei Wong said. Perplexed, the human resources department immediately pulled out a list from the printer, which contained the records of her going out with access control during working hours in half a year, all in red.

"If you swipe your card during working hours, the system will automatically generate a red warning record,” human resources explained.

"I just went to the convenience store to buy a rice ball!" Feifei Wong felt wronged. Human resources responded that the labor discipline protocols had been known to her in black and white before she started the job, and the deduction of money was "reasonable.”

Digitization for digitization’s sake:

The companies that develop these systems are deeply familiar with procurers’ unreasonable algorithm requirements. One OA system development company marketing director who did not want to be named told the Caijing reporter that all kinds of companies that come to consult often "care more about the label of digitization so as to get subsidies from the government and superiors." As for the effectiveness and possibly negative effects of digitization, they "don't care.”

The end:

Just before Caijing published this piece, Fu Cheng's year-end award was also issued. Affected by the performance of his project team, it was 7,000 RMB less than the previous year. After receiving the text message, Fu Cheng posted a photo of him and his wife on a trip to Disneyland in his WeChat Moments, accompanied by the lyrics from the new album of the famous folk band "Omnipotent Youth Society [万能青年旅店]":

"Under the Milky Way,

An electronic wasteland. 

Hundreds of millions of ways of the world,

Hundreds of millions of mud that can defile a man."

FOR MORE, including a note on how the last line of this stanza is a reference to a poem by the legendary Chinese poet Du Fu, SEE FULL TRANSLATION: Algorithmically-Monitored Workers?

ChinAI Links (Four to Forward)

Must-read: The Immigration Preferences of Top AI Researchers: New Survey Evidence

A co-published paper by the Centre for the Governance of AI and Perry World House, from the team of Remco Zwetsloot, Baobao Zhang, Markus Anderljung, Michael C. Horowitz, Allan Dafoe

A survey of more than 500 active researchers who publish in the leading machine learning conferences. Here’s one of the many interesting findings: “Nearly 60 percent of respondents not currently based in the United States think there is a greater than one-in-four chance they will move there within the next three years. The same percentages are 35 percent for the United Kingdom, 28 percent for Canada, and 10 percent for China.”

Must-read: 'China-watching' is a lucrative business. But whose language do the experts speak?

An essential entry in the debate over whether China “experts” need to know Chinese, by Yangyang Cheng for The Guardian: “As China develops from an impoverished backwater into the world’s second largest economy, many in the west have looked to it as fertile ground for promising careers. Their passion is not in Chinese history or culture, at least not as a priority. To the corporate elite, China is a market to be mined. To the security expert, China is a threat to be addressed. To the politicians and pundits, China is a “problem” to be solved. The lives and wellbeing of Chinese people, affected by policies, rhetoric and business deals, barely register in these discussions. Knowledge of the local language becomes irrelevant when the natives are presumed silent.”

Should-read: Signals of strength: Capability demonstrations and perceptions of military power

A fascinating article by Evan Braden Montgomery, for Journal of Strategic Studies, on the importance of demonstrations to establish the credibility of military capabilities. An eye-opening historical analysis is followed by an analysis of how emerging technologies, including AI, could affect demonstration mechanisms.

Should-read: The Pandemic Is Propelling a New Wave of Automation 

By Will Knight for Wired, a great dive into the world of office automation, with some good details about how Covid has accelerated the process. It also makes important distinctions between the work that can be automated with RPA software vs. office automation enabled by machine learning systems.

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a researcher at the Center for the Governance of AI at Oxford’s Future of Humanity Institute.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

Greetings from a world where…

ideology is an imaginary relationship to a real situation

…Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors). As always, the searchable archive of all past issues is here.

Feature Translation: Alibaba’s AI Lab Fizzles Out

Context: Usually I try to optimize the newsletter toward topics that have a very high ratio of importance to “boringness” (open source microservices, anyone?). This week’s feature translation is a little more “gossipy,” but it’s another high-profile example of an important point I’ve tried to highlight again and again: Western observers consistently overinflate Chinese AI capabilities.

Last week, news about the shutting down of Alibaba’s AI lab reached No. 1 in hot posts on Maimai (脉脉), a Chinese career networking platform and a rival to LinkedIn (screenshot below):

An Alibaba employee wrote: “Alibaba’s AI labs have already fizzled out, and Alibaba’s official website and DAMO Academy have deleted the webpages about Alibaba’s AI labs.” In fact, Daniel Zhang, Alibaba’s CEO, may have shut down the AI labs as early as 2019. To be clear, Alibaba’s R&D arm, DAMO Academy, will continue to do AI research, and Alibaba claims this is just part of restructuring. More details in this week’s feature translation (links to original in Mandarin), from AI科技评论(aitechtalk), a platform that focuses on in-depth reports on developments in the AI industry and academia.

Dig deeper:

• Aitechtalk pinned Alibaba’s response to the piece: “The AI labs have not closed. In the last round of structural changes, it has been integrated into Alibaba Cloud Intelligence as a whole, led by Professor Tan Ping. Alibaba will continue to increase its investment in artificial intelligence research.” Is this big tech’s version of the excuse: I wasn’t fired, I’m resigning to spend more time with family?

• In 2017 speech, Jack Ma said this about Alibaba AI labs (including DAMO Academy as a whole): “More than 90% of what is researched cannot just be in the laboratory, but must be in the market. Only in this way can this lab walk a long road.” Alibaba AI labs did produce some important breakthroughs, such as the Tmall Genie smart speaker, which generated millions in sales. But Tmall Genie was the exception not the rule, and many of the lab’s products did not pan out.

• The piece makes the case that these types of research failures are normal. If AI labs were judged on their ability to commercialize products, not many would score well. Citing two international examples to underscore this point, the article presents a graph of DeepMind’s annual losses and also references Element AI’s (Canadian AI startup) disappointing sale.

Some fun in a scathing comments section:

As you can see from the top of the first screenshot below, this week’s feature translation has been read almost 60,000 times (seven of my WeChat friends had already read it before I opened it).

• The top-rated comment from 隐形轰炸鸡 (Invisible Superchicken) compares Alibaba’s 90% rule for market-oriented research to Huawei’s approach to research: “Huawei’s 2012 labs (Jeff’s note: this is Huawei’s R&D arm) simultaneously carry out research oriented towards advanced technologies in 2035 and beyond, which may not necessarily enter the market. The product R&D and marketing departments shall not interfere with these research directions.” The second-rated comment also praises Huawei’s approach to R&D.

• In the first comment from the next screenshot (below), Bios writes: “Is this to save money to do community group buying? (note: this is when a designated community leader coordinates food orders, a phenomenon which has shaken up e-commerce in China recently) Deepmind burns money, but what Alphafold2 has accomplished has pushed the entire industry forward a big step, a big step in the basic direction of medicine. Quoting Professor Zhang Yang, who is well known in the industry, the ability of the industry to gather talents and resources is the envy of academia.”

• E’fei writes, “China’s most advanced AI technology is all used on precision-targeted sales and big data-enabled price discrimination.”

ChinAI Links (Four to Forward)

Must-read: Why I’m Suffering From Nanotechnology Fatigue

Back in 2016, Andrew Maynard wrote for Slate on his exhaustion with how nanotechnology became “brand nanotechnology” — a 14-letter fast-track to funding and the source of an “endless cycle of nanohype.” You could take Maynard’s text, find and replace all references to “nanotechnology” with “AI” and republish it today.

Should-read: Personal Data, Global Effects: China’s Draft Privacy Law in the International Context

For DigiChina, Alexa Lee analyzes China’s draft Personal Information Protection Law (PIPL), which “represents a third way between the sectoral U.S. approach, which applies different rules for specific industries or classes of consumers, and the European Union’s comprehensive General Data Protection Regulation (GDPR) framework, which enshrines fundamental rights across contexts. With the draft law, China’s evolving data governance regime emphasizes consumer privacy while also prioritizing national security through data localization measures, cross-border data flow restrictions, and continued surveillance and law enforcement powers.”

Should-read: 10 AI Failures in 2020 by Synced

Fourth installment of Synced’s year-end compilation, which includes some examples from China I hadn’t come across before reading.

Should-read: Is Substack the Media Future We Want?

A piece I’m still grappling with after reading a week ago. Anna Wiener writes for The New Yorker: “But Substack’s founders have acknowledged that, for the majority of writers, a newsletter will be a side hustle. In most cases, subscription fees will generate not a salary but something closer to tips. In a recent blog post on Medium, Hunter Walk, a venture capitalist, compared a newsletter to a stock-keeping unit, or sku, a term of art in inventory management. “The biggest impact of someone like Casey [Newton] unbundling himself” from the Verge, Walk wrote, “is that he is now an entrepreneur with a product called Casey. His beachhead may very well be a paid newsletter . . . but the newsletter is just one sku. . . . There could be a podcast sku. A speaking fee sku. A book deal sku. A consulting sku. A guest columnist sku. And so on.” Lisa Gitelman, a media historian and professor at New York University, said, of Substack, “They obviously want to call it a democratizing gesture, which I find a little bit specious. It’s the democracy of neoliberal self-empowerment. The message to users is that you can empower yourself by creating.”

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a researcher at the Center for the Governance of AI at Oxford’s Future of Humanity Institute.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

***1.17.21 NOTE: This issue has been edited to correct a transliteration of a name, as well as a translation of community group buying. H/t to ChinAI reader Giuseppe Baldini for bringing this to my attention.

Greetings from a world where…

the best Plan B will emerge from the multitudes

…Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors). As always, the searchable archive of all past issues is here.

***Translation note: Thanks to many readers for commenting on a translation question from the previous ChinAI issue re: 有线通信, for which suggestions converged on “fixed-line communications” as the best translation.

Feature Translation: The Ten Biggest Technological Advances in AI in 2020

The Beijing Academy of Artificial Intelligence asked its scholars to come up with a joint top 10 list (links to the original Mandarin) of the most important AI advances in 2020. Instead of a Google doc, I’ve just translated what they came up with below:

10: Controlling Fairness and Bias in Dynamic Learning-to-Rank

Cornell University proposed an unbiased and fair ranking model to alleviate “Matthew effect” issues with online search rankings

In recent years, the fairness of retrieval and recommendation models based on counterfactual learning have become important research directions in the field of information retrieval. Related research results have been widely used in click data correction, offline model evaluation, etc. Some technologies have already been implemented in recommendation and search products from companies such as Alibaba and Huawei. In July 2020, the team of Professor Thorsten Joachims of Cornell University published FairCo, a fair and unbiased ranking learning model, which won the SIGIR 2020 Best Paper Award (conference on Research and Development in Information Retrieval)…The improvement has received wide attention and praise from the industry.

9: Google and Facebook teams each proposed new unsupervised representation learning algorithms

At the beginning of 2020, Google and Facebook proposed SimCLR and MoCo, respectively, both of which can learn representations of images on unlabeled datasets. The framework behind the two algorithms is contrastive learning. The core training signal in comparative learning is the "distinguishability" of pictures. The model needs to distinguish whether two inputs are from different perspectives of the same picture, or inputs from two completely different pictures. This task does not require human annotation, so a large amount of unlabeled data can be used for training. Although the two works of Google and FaceBook deal with many details of training differently, they both show that unsupervised learning models can approach or even match the results of supervised models.

8: MIT uses only 19 brain-like neurons to control self-driving cars

Inspired by small animal brains such as the nematode, teams from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), TU Wien, and IST Austria used only 19 brain-like neurons to control autonomous vehicles. The conventional deep neural network requires millions of neurons. In addition, this neural network can imitate learning, and has the potential to be extended to warehouse automation robots and other application scenarios. The results of this research have been published on October 13, 2020 in a Nature subsidiary journal — Nature Machine Intelligence.

7: Peking University achieves for the first time a neural network high-speed training system based on phase-change memory

In December 2020, the team of BAAI scholar and Peking University’s Yuchao Yang’s team proposed and implemented a neural network high-speed training system based on phase-change memory (PCM), which effectively reduces the time and energy costs of training artificial neural networks — problems that are difficult to implement on the chip. The system is improved on the basis of the direct error feedback algorithm (DFA), and uses the randomness of the PCM conductance to naturally generate random weights for the propagation errors, which effectively reduces the hardware overhead of the system and the time and energy consumption in the training process. The system performs well in the training process of large-scale convolutional neural networks, which provides a new direction for the application of artificial neural networks on terminal platforms and the realization of on-chip training.

***Comment: I couldn’t find the paper for this, so I linked to Professor Yang’s research page.

6: Tsinghua University first proposes the concept of neuromorphic completeness of brain-like computing and a corresponding system hierarchy

In October 2020, a team including BAAI scholars Zhang Youhui, Li Guoqi, and Song Sen of Tsinghua University put forward the concept of "neuromorphic completeness" and a corresponding system hierarchy, which affects the compatibility between software and hardware, and proved this type through theoretical applications. and prototype experiments. The hardware completeness and compiling feasibility of the system expand the application range of the brain-like computing system to support general-purpose computing. The research results were published in the journal Nature on October 14, 2020. Nature commented that the new concept of “completeness” promotes neuromorphic computing, constituting a "breakthrough solution" for the tight coupling of software and hardware in brain-like systems.

5: Baylor College of Medicine in the United States achieves high-efficiency "visual cortex implanting" through dynamic intracranial electrical stimulation

For more than 40 million blind people around the world, seeing the light again is an unattainable dream. In May 2020, researchers at Baylor College of Medicine in the United States used the new technology of dynamic intracranial electrical stimulation to form a visual prosthesis with an implanted microelectrode array, drawing the shapes of letters such as W, S, and Z in the human primary visual cortex, and successfully allowed blind people to "see" these letters. Combined with the high-bandwidth, fully implantable brain-computer interface system released by Neuralink, a brain-computer interface company founded by Musk, the next-generation visual prosthesis may accurately stimulate every neuron in the primary visual cortex of the brain, helping the blind "see" more complex info to realize their dream of seeing the world clearly.

***Comment: Anything related to Elon Musk gets a lot of coverage in China.

4: DeepMind and others use deep neural networks to solve the Schrödinger equation, promoting the development of quantum chemistry

I’ve excerpted their summary of FermiNet (I’ve linked DeepMind’s blog post). They also summarize related research: “In addition, in September 2020, several scientists from the Free University of Berlin, Germany, also proposed a new deep learning method, which can obtain nearly exact solutions to the electronic Schrödinger equation. Related research was published in Nature Chemistry. This type of research shows not only the application of deep learning in solving a specific scientific problem, but also a great prospect for deep learning to be widely used in scientific research in various fields such as biology, chemistry, materials, and medicine.”

3: Gordon Bell Prize for Deep Potential Molecular Dynamics Research

On November 19, 2020, at the SC20 conference in Atlanta, USA, the “Deep Potential” team, which included BAAI scholar Wang Han of the Beijing Institute of Applied Physics and Computational Mathematics, won the highest award in the international high-performance computing application field — the "Gordon Bell Prize…”

***Comment: Press release of the award states, “ACM, the Association for Computing Machinery, named a nine-member team, drawn from Chinese and American institutions, recipients of the 2020 ACM Gordon Bell Prize for their project, ‘Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning.’ ”Practical applications include accelerating drug development. Here’s a question: How many people researching AI and U.S.-China relations know about this accomplishment? Until reading this article, I was not one of them. A reminder that U.S.-China cooperation in AI happens all the time, and a lot of it occurs for the greater good.

2: DeepMind's AlphaFold2 solves the problem of protein structure prediction

1: OpenAI releases the world's largest pre-trained language model GPT-3

Comment: Probably would be 1-2 on any top ten list. Chose not to include the article’s summaries for these two, as these have been amply covered elsewhere.

Feature Translation: IDC’s 10 Big Predictions for China’s AI Market in 2021

The International Data Corporation (IDC) published a report, “IDC FutureScape: Global AI Market 2021 Predictions — China Implications.” (links to original summary in Mandarin) Their 10 big predictions are as follows:

Prediction 1: By 2023, more than 15% of consumer-centric AI decision-making systems in the financial, medical, government and other regulated public sectors will introduce relevant regulations that explain their analysis and decision-making processes.

Prediction 2: By 2021, more than 50% of organizations will add AI capabilities to environments that process incoming calls.

Prediction 3: By 2024, 45% of repetitive tasks will be automated or enhanced through the use of “digital workers” supported by AI, robotics, and robotic process automation (RPA).

Prediction 4: By 2023, the number of data analysts and data scientists using an end-to-end machine learning platform from data preparation to model deployment that is encapsulated using automated machine learning (AutoML) technology will double.

Prediction 5: By 2024, automated operation and maintenance (AIOps) will become the new normal of IT operations. At least 50% of large enterprises will adopt automated operation and maintenance solutions to automate major IT systems and service management processes.

Prediction 6: By 2025, 10% of artificial intelligence solutions will be closer to general artificial intelligence (AGI) — using neural symbolic technology to combine deep learning and symbolic methods to create methods that are more reliable and closer to human decision-making.

Prediction 7: By 2021, at least 65% of China’s top 1000 companies will use AI tools such as natural language processing (NLP), machine learning (ML) and deep learning (DL) to empower 60% of use cases in business areas such as customer experience, security, operations management, and procurement.

Prediction 8: By 2024, more than 30% of China’s top 1000 companies will deploy AI workloads more evenly on the end, edge and cloud. These workloads will be managed by artificial intelligence software platform providers to make AI infrastructure "invisible.”

Prediction 9: By 2023, 30% of enterprises will run different analysis and AI models on the edge. Among them, 30% of edge AI applications will be accelerated by heterogeneous acceleration solutions.

Prediction 10: By 2022, 80% of China's top 1000 companies will invest in internal learning platforms and third-party training services to meet the needs of new skills and work style changes brought about by the adoption of AI.

Comment: I find these predictions useful mostly because they set out some interesting indicators that can be measured. As for the veracity of these predictions, I would say they’re overly optimistic and under-specified — though, to be fair, you can’t go through every condition in a one-sentence prediction. I would say that Roy Amara’s quote applies here: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

ChinAI Links (Four to Forward)

Should-read: Machine learning is going real-time

On the “MLOps race between the US and China,” Chip Huyen writes: “Few American Internet companies have attempted online learning, and even among these companies, online learning is used for simple models such as logistic regression. My impression from both talking directly to Chinese companies and talking with people who have worked with companies in both countries is that online learning is more common in China, and Chinese engineers are more eager to make the jump.”

Comment: This was a thought-provoking read. I still have a lot to learn here and haven’t talked to as many insiders on the ground on this issue, so this causes me to update my views slightly on China’s adoption speed of MLOps. Two caveats spring to mind for me: 1) MLOps is more than just online learning; 2) MLOps applies to more than just mobile apps (which is what it seems like most of the anecdotal evidence draws on)

Should-read: Nordic lights? National AI policies for doing well by doing good

Jacob Dexe and Ulrik Franke for Journal of Cyber Policy. Here’s the abstract: Getting ahead on the global stage of AI technologies requires vast resources or novel approaches. The Nordic countries have tried to find a novel path, claiming that responsible and ethical AI is not only morally right but confers a competitive advantage. In this article, eight official AI policy documents from Denmark, Finland, Norway and Sweden are analysed according to the AI4People taxonomy, which proposes five ethical principles for AI: beneficence, non-maleficence, autonomy, justice and explicability…

Should-read: Year-in-review Recs from Robot Humanities (机器人人文) Account (in Mandarin)

Came across a cool account that recommends excellent articles at the intersection of robotics and the humanities. Happy to feature translations from this list if there are any ChinAI readers wanting to contribute in 2021.

Should-read: Why Chinese youngsters are embracing a philosophy of “slacking-off”

Jane Li writes for Quartz:

The intense anxiety felt by younger people, and exacerbated by the pandemic, prompted a wider discussion on a once niche academic concept: neijuan. Translated as “involution,” the anthropological term was first applied to agriculture, and has come to describe conditions in which a society ceases to progress, and instead starts to stagnate internally. Increased output and competition intensify but yield no clear results or innovative, technological breakthroughs.

Neijuan has become a hot topic on the Chinese internet and in media reports this year as a word that “captures urban China’s unhappiness.” Complaints of their work becoming too “involuted”—more competitive with little corresponding rewards—are as likely to be discussed on Weibo by white-collar workers as food delivery drivers.

H/t to Doug Orr for sharing.

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a researcher at the Center for the Governance of AI at Oxford’s Future of Humanity Institute.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

***1.6.21: this post was edited to fix a typo in a comment on the 3rd ranked development in the first translation.

Greetings from a world where…

Third-tier newsletters cover product innovations, second-tier newsletters cover process innovations, first-tier newsletters cover technology standards.

…Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors). As always, the searchable archive of all past issues is here.

Feature Translation: China Standards 2035 ready to come out, this is the real key to Sino-U.S. tech competition

Last week, we covered localized substitution in Chinese software, which included a table, titled “Domestic and foreign ecosystem alliances in chips + operating systems” from Huaxi Securities Research Institute. Thanks to Kristy Loke for stepping up and translating it in full — now on page 5-6 of the full translation from last week.

Context: This week, we circle back to a previous Around the Horn pick that some readers had expressed interest in, from 智谷趋势 (zgtrend), which was the source of the feature translation on China’s North-South gap from a couple weeks back.

Key Takeaways:

Article expresses the standard narrative about China Standards 2035, which goes like this:

• First, China has lacked the “right to speak”/”discursive power” in writing the rules of the technological roads. Hence, China pays the second most in intellectual property royalties, which leads to a large deficit in trade in services. In the past, when China tried to challenge American-dominated standards with its own — most notably, the WAPI standard for wireless — Chinese experts were denied visas by the U.S. to attend key meetings. That’s not completely true, but that’s the version in the article. Here’s more context on what actually happened in the WAPI case back in 2004. Also, there’s more to technical standards-setting than zero-sum competition between great powers. In a commentary for NBR, I analyzed how China, the United States, and other countries are balancing priorities in their pursuit of technical standardization in data governance and artificial intelligence

• Second, new technological revolutions provide an opportunity for China to lead in the formulation of new systems of standards. AI is key to this, so that’s why in mid-August, the Standardization Administration of China and other five departments jointly issued the "Guide to the Building of a National Standard Framework for New Generation Artificial Intelligence.”

Some nice, substantive indicators for China’s efforts to increase influence in digital standards-setting:

• In recent years, the annual growth rate of China's submissions to the ISO and IEC (two leading int’l technical standards bodies) has reached about 20%. In 2019, China submitted a total of 238 proposals for ISO and IEC international standards, of which 150 were submitted to ISO, 77 were submitted to IEC, and 11 were submitted to the Joint ISO/IEC Information Technology Committee (JTC1) — this is the body where one of the most important AI standards bodies is housed (SC 42)

• China has submitted 830 technical documents related to wire communication (有线通信 — how do you translate this?!) to the ITU, ranking first in the world, surpassing the total of the following three countries, South Korea, the United States, and Japan.

• It’s important to keep all this in comparative context. When measured by national representatives at the chair or vice-chair positions at ISO and IEC technical committees, China’s influence has greatly increased but remains below that of other countries like the U.S. and Germany. As the article notes, these are very rough proxies for influence — at the end of the day, the best tech does carry a lot of weight even if it doesn’t always win, so the quality of technological standards is most important. The figures below support these two points — data from the Japanese Industrial Standards Committee.

One last thought for those who like meta-analysis. It’s notable how much international (esp. U.S.) coverage of China Standards 2035 has shaped this zgtrend article.

• The opening quote cites Frank Rose, a senior researcher at Brookings, on the importance of standards for U.S.-China tech competition — which is essentially the frame of the article. It also quotes U.S. Attorney General Barr’s speech at a CSIS conference, Naomi Wilson’s excellent de-hyping of China Standards 2035 on the CFR blog, and many other international outlets like Foreign Affairs.

• In fact, the only two Chinese voices quoted are Yan Xuetong, a prominent Tsinghua professor, and Dai Hong, director of the State Administration of China (SAC)’s Industrial Standards Second Department. Interestingly, I think I tracked down the Dai Hong quote in Dr. Ray Bowen II’s testimony (p. 21) before the U.S.-China Economic and Security Review Commission from earlier this year in March.

• Why should we care? I think this fits in with a broader trend I’ve noticed in scanning through Chinese tech coverage every week. Many of the writers and platforms are very plugged in with what’s happening in the English-language landscape (e.g. articles that largely translate influential Twitter threads).

More in FULL TRANSLATION: China Standards 2035 ready to come out, this is the real key to Sino-U.S. tech competition

ChinAI Links (Four to Forward)

Must-read: What’s in a Name? Metaphors and Cybersecurity

Very envious of this article by Jordan Branch in International Organization. It demonstrates the use of language — specifically, “cyberspace” as a “domain” for military operations — has expanded the military’s role in cybersecurity and bolstered support for the creation of a U.S. Cyber Command. Foundational metaphors matter.

Should-read: In AI ethics, “bad” isn’t good enough

Amanda Askell writes, “Lately I've been thinking about AI ethics and the norms we should want the field to adopt. It's fairly common for AI ethicists to focus on harmful consequences of AI systems. While this is useful, we shouldn't conflate arguments that AI systems have harmful consequences with arguments about what we should do. Arguments about what should do have to consider far more factors than arguments focused solely on harmful consequences.”

Should-read: Mapping U.S. Multinationals’ Global AI R&D Activity

New CSET analysis: Many factors influence where U.S. tech multinational corporations decide to conduct their global artificial intelligence research and development (R&D). Company AI labs are spread all over the world, especially in North America, Europe and Asia. But in contrast to AI labs, most company AI staff remain concentrated in the United States. Roxanne Heston and Remco Zwetsloot explain where these companies conduct AI R&D, why they select particular locations, and how they establish their presence there. The report is accompanied by a new open-source dataset of more than 60 AI R&D labs run by these companies worldwide.

Should-reread: ChinAI #120 Singles Day and the Making of Alibaba Cloud

I want to re-up this ChinAI issue from last month with some reactions from Cooper Pellaton, a student at Georgia Tech who’s worked at Alibaba. He’s an alumni of Alibaba’s DAMO Academy so he reached out after reading the issue. Thanks to Cooper for sharing his inside view on DingTalk, which is Alibaba’s “Slack-like” enterprise communications platform:

“Overall, I think DingTalk is really underrated but it’s definitely part of a larger toolchain at Ali. Things that worked successfully there which I haven’t seen anywhere else were integrated project management/time tracking/and version control system. Everything rolled into one means it’s a one stop shop for work (also easy request to other teams code/projects). Enterprise software is different in China and deserves to be studied in its own right.”

More details on DingTalk, which he calls a “great, but not so great” product:

“Positives:

• Real-time translation. In beta while I was at DAMO, you can message with people in your native language and all the translation is done seamlessly. Nice because while there are very few foreigners at Ali, you occasionally need to refer to things in your native language.

• Deep integrations. Features mini-apps like in Slack or WeChat. We could read AliMail, see package delivery, etc. all in chats.

Downsides:

• No statuses or calendars. Because most meetings are made informally and no one uses emails, having a daily “calendar” or schedule is impossible. Meaning finding your coworker is always a puzzle.

• Employee discovery is terrible. Very difficult to find your coworker when the contact cards are tiny and lack information - a bigger problem then you’d imagine.”

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a researcher at the Center for the Governance of AI at Oxford’s Future of Humanity Institute.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

Greetings from a world where…

we are crawling forward (匍匐前进)

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: The Wave of Localized Substitution — Is it Here?

Context: Surely, by now, you’ve heard about this grand narrative of “DECOUPLING!” between the U.S. and China, right? Two Internets — or better yet, The Splinternet. This grand narrative is usually accompanied by a few cherry-picked anecdotes — or, if you’re lucky, maybe even some statistics that show a time trend— from a few specific industries.

We can do better. This week’s feature article, from long-time ChinAI source Leiphone, analyzes China’s efforts at 国产化替代 (localized substitution, i.e. the replacement of foreign technology with Chinese-developed alternatives). Based on reports from Chinese security companies and interviews with key players.

Key Takeaways:

1) There are some indications that localized substitution is progressing:

• In office software, about 90% of State Council organizations, central state-owned enterprises, commercial banks, and provincial governments use WPS Office, an office suite developed by Chinese software developer Kingsoft.

• Inspur, China’s leading server provider, says it supplies 56% of central SOEs, 38% of all SOEs, and 31% of China’s top 500 companies.

• In customer relationship management (CRM), Chinese companies like Xiaoshouyi are gradually challenging leading int’l companies like Salesforce and Sieble.

• Other companies mentioned include: Seeyon (Beijing-based IT company), which does a lot of government affairs IT work, Dameng (Wuhan-based company that does database services), Quanshi (videoconferencing company), and KylinOS (an operating system).

2) Not so fast. We are not seeing a wave of localized substitution. If we get to the level of analyzing “China tech” in terms of specific verticals rather than as an abstract monolith, a more complicated picture emerges:

• According to data released by Tianfeng Securities, the localization rate of mid-to-high-end ERP (enterprise resource management) in industrial software is 25%; the localization rate of CAD (computer-aided design) is 11%, and the localization rate of MES (Manufacturing Execution System) is 30%. Main Chinese players in these verticals include Kingdee and UFIDA.

• Here’s the tough question IT folks grapple with: Is it better to overthrow everything and start over, or to continue patching? Replacing your SAP or Oracle tools for ERP is an ecosystem overhaul — not just a simple substitution of products. Liu Kai, a partner at Chinese VC 5y Cap, states, “In the past two years, there have been more slogans like Xinchuang or localization in China, but in the key ‘chokehold’ areas where technological accumulation is needed, the localization rate of basic software is still very low."

• Liu Kai gives another example in database services. China’s domestic database companies (e.g. Dameng, Kingbase, GBase, and Shentong Data) mainly serve the party, government, and army, but there are very few application scenarios. The world is much bigger than these three customers: Oracle can get 3 billion USD from database services to China’s Internet companies, financial sector, SOEs, and top 500 enterprises.

• With software, it’s all about employee experience. Chen Xuejun, the CEO of videoconferencing company Quanshi, notes that the impact factor of user/employee experience is more than 90% in deciding what solution is chosen. Liu Kai says, “Most of the decision makers at state-owned central enterprises are making decisions in order to achieve political tasks and hope to achieve autonomy and control. For example, in the semiconductor field, there is an autonomy rate indicator (how many products must be domestically produced) that has been proposed, but this does not matter if employees are unwilling accept it.”

3) Some Extra Tidbits

• The tension between open-source and localized substitution. Wei Shaojun, chairman of the IC Design Branch of China Semiconductor Industry Association and professor at Tsinghua University, has argued, “China's semiconductor industry is a bit too hot, a bit unreasonable. We must prevent extremist and closed thinking. Instead of substitution thinking as the main theme of development, the main theme should be openness and cooperation.” Others argue that open source and technological autonomy do not conflict.

• The phrase 信创产业 (Xinchuang Industry) is peppered throughout this piece. According to this Leiphone article, the term originated from China’s “Information Technology Application Innovation Committee,” which is an organization initiated and established by 24 units engaged in the research, application and service of key software and hardware technologies. It seems to be a buzzword/slogan that covers four verticals that we’ve discussed in this issue: 1) IT foundations (e.g. chips, servers, storage, cloud); 2) basic software (OS, database, middleware); 3) Application software (ERP, office software, etc.); 4) Information security. If anyone has a good way to translate 信创, let me know!

• Piece also includes a very interesting table of domestic and foreign ecosystem alliances in chips + operating systems:

*Any enterprising ChinAI reader want to help out with translating this table? Reply first to this email and commit to doing it before next week’s issue.

FULL TRANSLATION: The Wave of Localized Substitution — Is it Here?

ChinAI Links (Four to Forward)

Must-read: Huawei / Megvii Uyghur Alarms

The IPVM team found a Huawei "interoperability report" that show Huawei and Megvii worked together to test and validate software that could Uyghur minorities and alert police — one of the functions of Megvii’s facial recognition system that was reviewed by Huawei. I’m linking the WashPost coverage because the IPVM link wasn’t working for me.

Should-watch: Lisa Simon on The Future of Work and How the Workforce Adapts to Change

A recent Stanford HAI seminar on how individuals adapt to automation. She explores how AI is fundamentally changing how work gets done and by whom, investigating how individuals transition between different career pathways and how they react when faced with negative shocks.

Should-read: Chip Huyen’s Twitter Thread on MLOps in China

Chip Huyen @chipro

Talking to Internet companies in China gives me the impression that their MLOps infra is away ahead. Many US Internet co. are still wondering if there's value in real-time ML. Chinese counterparts are already doing real-time inference + online training. Are my samples biased?

December 10th 2020

66 Retweets772 Likes

My default is that Chinese companies are not way ahead in this area, but I couldn’t find any good Chinese-language analysis on MLOps/AIOps. There’s one paywalled Gartner report that might be relevant.

Should-read: 500,000 Kids, 30 Million Hours: Trump’s Vast Expansion of Child Detention

From The Marshall Project: “When U.S. Customs and Border Protection holds migrant children in custody, the child’s detention is supposed to be safe and short. That’s true whether the child is with a parent or without one.

But new data shows that over the last four years, detention times lengthened as the number of children held at the border soared to almost half a million. The detentions, which include both unaccompanied children and children with their families, peaked last year at over 300,000, with 40 percent held longer than the 72-hour limit set by a patchwork of legislation and a court settlement.”

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a Predoctoral Fellow at Stanford’s Center for International Security and Cooperation, sponsored by Stanford’s Institute for Human-Centered Artificial Intelligence.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99