Welcome to the ChinAI Newsletter!

Our subscription drive continues until we get to 100 subscribers (at around 90 right now — thanks to all new subscribers!). I’m making a brief trip to DC on October 23-25: if you’re a subscriber, let me buy you a drink as a token of appreciation. If you’re not, I will aggressively ignore your emails (jk but not really).

As always, the archive of all past issues is here and 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 like Victoria Wu — a ChinAI subscriber who helped with this week’s feature translation.

Feature Translation: CASICloud and China’s Industrial Internet

I like to choose topics where the information arbitrage ratio (relative importance divided by number of people paying attention) is very high. China’s efforts to build an industrial Internet has a very very high information arbitrage ratio. On this topic, we’ve previously translated a case study of CASICloud in an AI Open Source Software White Paper and also covered the unsexy details of how computer vision affects quality inspections on production lines for making cutting tools.

THE CONTEXT: A snapshot of China’s industrial Internet landscape by way of a Leiphone interview with Xu Shan, deputy GM of CASICloud, one of the key players and a subsidiary of China Aerospace Science & Industry Corporation which is a large [Fortune 500] state-owned enterprise develops missiles, aerospace products, etc.

THE ESSENTIALS:

• Taking the buzz out of the industrial Internet buzzword: a term by General Electric in late 2012, the Industrial Internet of Things refers to a system of industrial devices connected with communications technologies that enables advanced analytics, machine-to-machine coordination, etc. Two of the big players are GE’s Predix and Siemen’s MindSphere.

• Need to knows about CASICloud: established in June 2015, started working on a cloud platform for the industrial internet in 2015 and released the INDICS platform in 2017, has support of CASIC — one of the top performing high-tech SOEs and key behind-the-scenes player behind a lot of the weapons/equipment showcased during the National Day military parade.

• Disaggregating the notion of a single industrial Internet: CASICloud works with 28 SOEs including State Grid/China Unicom to build a national integrated industrial Internet, but it also has deployed regional industrial Internet service platforms including for Guizhou and Changzhou as well as regional platforms for single industries (e.g. a Sichuang Heavy Equipment cloud platform)

• A government-guided but many obstacles: CASICloud’s main way to land these industrial Internets is to form a joint venture company with the local government, and very interestingly, CASICloud claims to handle data processing through the National Engineering Lab for Industrial Big-data Application Technology. There are still many issues with data islands and companies that want to hold on to their legacy information systems.

• An indicator of how important this stuff is: In 2017, Gao Hongwei, Chairman of CASIC, and Joe Kaeser, CEO of Siemens, signed an agreement to allow both CASICloud and Siemens to build applications on the other platform. The signing ceremony was attended by Xi Jinping and Angela Merkel.

DISCUSSED IN THE FULL TRANSLATION (in the style of Believer magazine): how 5G and TSN are crucial to closing the loop on the industrial Internet system, the differences among IaaS, PaaS, SaaS layers and all the acronyms in the world, and just a lot of really dense and technical but interesting things about the industrial Internet.

FULL TRANSLATION: How many steps does it take to transfer the digitized capabilities of national-level aerospace equipment to the manufacturing industry?

ChinAI Links (Four to Forward)

Should-read: The Turnaround by Jayadevan, a journalist who has covered technology for Indian publications for 10 years

One recent podcast episode of The Turnaround covered the Indian gaming sector, which is growing rapidly and catching the attention of the Chinese VC/start-up world. In a follow-up issue, Jayadevan reflects on a previous ChinAI issue on the diverging trajectories of China and India in the information revolution and the link to online games.

Should-read: Expanded U.S. Trade Blacklist Hits Beijing’s Artificial-Intelligence Ambitions

The WSJ’s Dan Strumpf and Yoko Kubota analyze the effects of the U.S. decision to add 8 Chinese companies to the entity list, with good details on the percentages of companies’ revenues from Xinjiang, revenues likely to be affected by the blacklisting, and possible indirect effects.

Should-read: Global Defence-Industry League: Where is China?

The International Institute for Strategic Studies’s Meia Nouwens and Lucie Béraud-Sudreau calculated how much defense-related revenue that eight of ten Chinese SOEs involved in defense production generated in 2016; their calculations placed CASIC 11th in worlds top defense companies by total arms sales.

I really wanted to find a profile that comprehensively unpacked all of CASIC’s subsidiaries and research institutes but was unable to — if anyone has recommendations, please send them my way.

Should-read: The Next Word

John Seabrook’s piece for The New Yorker on predictive fed text from the end of each section of the section of the article into the New Yorker’s AI — a full-strength version of GPT-2 fine-tuned on all nonfiction work published in the New Yorker since 2007 along with some digitized classics dating back to the 1960s — and then generated the predicted text that follows each section in the article.

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 Rhodes Scholar at Oxford, PhD candidate in International Relations, Researcher at GovAI/Future of Humanity Institute, and Research Fellow at the Center for Security and Emerging Technology.

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

Welcome to the ChinAI Newsletter!

Our subscription drive continues until we get to 100 subscribers (at around 80 now). I’m making a brief trip to DC on October 23-25: if you’re a subscriber, let me buy you a drink as a token of appreciation. If you’re not, I will aggressively ignore your emails (jk but not really).

As always, the archive of all past issues is here and 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 5G Powerpoint

If a 100-slide powerpoint about 5G goes viral all across China but no one in the English-speaking world picks up on it/let alone translates it except for ChinAI, then in this convoluted and mismatched analogy, does ChinAI make a sound that cuts through the noise?

THE CONTEXT: Three months ago, a 100-slide PPT titled “5G: Empowering AI and Smart Manufacturing” went viral on the Chinese web. On August 22 Huawei’s founder Ren Zhengfei issued an email to all employees, recommending that they read this slide deck, inspiring another wave of attention. The PPT’s author is Dr. Wang Xiwen, member of the central science and technology commission of the Jiusan Society (a political party whose membership mostly consists of high- and medium-level intellectuals in fields of science, technology, and education). He previously authored the first monograph on “Industry 4.0” in China. He used this slide deck to lecture and inform local government officials and entrepreneurs about 5G.

THE ESSENTIALS:

• The PPT is divided into four parts:

1. 5G is arriving now — a very useful explainer of what 5G is, what it enables (eMBB, URLLC, mMTC), and the value it will generate

2. 5G Main System Architecture — covers the 5G bands allocated to Chinese telecommunications operators, key components of the system including mmWaves, miniature base stations, ultra dense networks to support large traffic loads, and how it all applies to smart manufacturing, vehicles, homes, medicine, etc.

3. 5G-accelerated AI — goes through a range of AI applications that 5G will enable. Best example is smart transport (slides 71-73) which starts by giving a landscape view of how 5G can enable various forms of transportation and concludes with emphasizing why having both intelligent AND connected vehicles is crucial to integrate with urban transport networks.

4. 5G Boost for Smart Manufacturing — framed around Made in China 2025 vs. Germany’s Industry 4.0, the deck illustrates how 5G will empower intelligent configurations of production methods, continue transition from linear, rigid manufacturing lines to flexible, highly integrated, and granularly controllable ones

• Shout out to a ChinAI contributor who wishes to remain anonymous for helping me out with half of the 100 slides. A nice takeaway from them: “What stood out to me was the sheer scale of the presentation, ranging across everything from medicine to agriculture, smart courts, finance, governance etc. To my mind it just underscores that this will inevitably have to be a massive all-of-society effort: building the sprawling webs of hyper-connected social layers envisaged in that presentation just doesn't seem like a project conducive to rigid top-down direction.” Jeff’s quick blurb: 5G is not a singular technology that affects a single domain. It is part of a technological system that will affect many different domains.

DISCUSSED IN THE FULL TRANSLATION (in the style of Believer magazine): Dividing the speed of light by 36 billion hertz frequencies (slide 15) , the world’s first digital pill (slide 63), the “carphone” and the “shoulder phone” (slide 5), a three-step strategy to build a manufacturing power in 30 years (slide 82)

*Note: We directly translated a good portion of the slides that weren’t image-based; for the rest, click on the images and there will be a corresponding Google doc comment with the translation.

FULL TRANSLATION: 5G Empowering AI and Smart Manufacturing PPT

ChinAI Links (Four to Forward)

Must-read: Networked Dream Worlds: Is 5G solving real, pressing problems or merely creating new ones?

Shannon Mattern, Professor of Anthropology at The New School for Social Research, pushes us to conjure up some new “infrastructural imaginaries” for 5G’s rollout. What might be the most physical disruptive change to the nation’s communications infrastructure — new data centers, millions of base stations, fiber installations — may bring in a new wave of “digital redlining,” perpetuating iniquities between communities distinguished by race, class, and culture.

Key points:

1. The race to 5G (what Mattern describes as a 21st-century Space Race at street level) is driven by market competition and international rivalries (e.g. China’s drive to 5G) but is it one worth winning? “50 years ago we ran highways through urban neighborhoods, for the sake of access and convenience, gutting cities in the process. Will we at some point regret having installed millions of cells and towers and mini-data centers for the sake of digital access and convenience?”

2. Mattern raises concerns about irradiation and privacy: will bathing every square inch of America with penetrating microwaves affect our health? Surveys of existing research have been inconclusive, some of the studies have been funded by telecom companies, and several towns have banned wireless towers in residential areas. Jeff’s note: I think some of this is unfounded paranoia but I also take Mattern’s point that, “Dwelling near nuclear plants and hazardous waste disposal sites, or amid leaded pipes, tends to cultivate distrust of other obfuscatory infrastructures.” Another risk often overlooked is that to personal privacy as networks of sensors and cameras will be enabled by the 5G network.

3. The 5G rollout will be slow and this momentary latency gives us a chance to dream new dreams. Do we really need this (the filtering of the world through AR googles, the collection of all our online and real-life activities, the new security risks and massive energy expenditures)? Vodafone’s CTO says we need to start thinking about 6G now. Mattern: “Maybe 6G, unlike its predecessors, won’t be about being first or fastest, most ubiquitous or unrelenting. Maybe it will be about energy efficiency. Or local responsiveness. Or slowness. Or reflexivity. Or privacy, or equity, or digital justice. Maybe 6G will enact a broader digital ethos, which affirms ‘the ways that technology can help individuals and communities be and relate to each other,’ as Seeta Peña Gangadharan puts it in “Digital Exclusion: A Politics of Refusal.” This implies a refusal of ‘access,’ speed, and growth as ends in themselves, and of tech companies’ hegemonic rule as a natural law.”

Should-read: ChinAI #58 — Making Knives Better & Landscape of China's Intelligent Manufacturing

The feature PPT heavily emphasizes smart manufacturing. For a deeper dive beyond the buzzword, read this past ChinAI issue that takes us inside a few companies trying to adopt computer vision to facilitate better detection of cutting tool defects on the production line.

Should-read: Can Chinese Students Abroad Speak?

A callback to Spivak’s classic essay “Can the Subaltern Speak” (which interrogated the obstructions to being heard of those who inhabited the periphery), Shan Windscript’s article critiques Australian discourse for rarely acknowledging the existence of Chinese international students except within ethnicised stereotypes—variously, as ‘cash cows’, ‘CCP spies’, and ‘patriotic students brainwashed from birth.’

In this essay, she proposes “an alternative frame of thinking, one that foregrounds political agency as not only desirable but necessary, and one that centres voices of dissent—however fragmentary—among Chinese international students.” 

Should-read: The 5G Fight is Bigger than Huawei

Elsa Kania urges us to recalibrate the conversation on 5G security away from a narrow focus on Huawei and toward a one focused on the underlying essence (cybersecurity, dependency, arbitrary exercise of state power) of the issues related to Huawei and one that recognizes that Huawei is still one among a number of contenders in the space.

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 Rhodes Scholar at Oxford, PhD candidate in International Relations, Researcher at GovAI/Future of Humanity Institute, and Research Fellow at the Center for Security and Emerging Technology.

Check out the archive of all past issues here and please 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

Welcome to the ChinAI Newsletter!

Trying to implement some feedback this week about making the newsletter more skimmable for all you busy folks in this modern world of ours. Please keep giving me feedback: treat ChinAI as an open source project that gets better with more pull requests. If, for instance, you really really crave that weekly header image of the Go board, let me know!

As always, the archive of all past issues is here and 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).

*LAST COUPLE DAYS OF OUR SUBSCRIPTION DRIVE: Our goal is 100 subscribers by end of September. If we get there, I’ll stop including this annoying message in each newsletter. C’mon folks, we’re churning out podcasts now too. Subscribe here.

The Meaning of Video Games: The Cotton and Oil of the Information Revolution

THE CONTEXT: a thought-provoking essay from “Necromanov,” a well-known and well-regarded gaming blogger, who argues that video games are the driving force behind the information revolution. It can be read as a direct critique of China’s crackdown on the gaming industry in 2018, when a new regulatory body implemented a 10-month government freeze on approvals of new games.

THE ESSENTIALS:

• Games were central to the computer revolution. The author argues that both the Soviet Union and Japan failed to believe that average people would want to master computer technology. The Soviet Union’s OGAS plan could have become the first Internet but bureaucrats stifled it. The Japanese firm Busicom released exclusive rights to Intel’s microprocessor patents, which were used to power the craze over Atari’s Pong in the 1970s.

• Necromanov highlights two pathways by which video games drive the information revolution: 1) they are a key source of profits and demand for both software and hardware companies; 2) they made new technologies more accessible for ordinary users

• More speculatively, the author argues that the diverging trajectories of China and India in the information revolution was largely due to the outbreak of online games in China in the late 1990s. The Internet giants that grew in this period (from Netease to Sohu to Tencent) relied on games for about half of their revenues, benefiting from the enthusiasm of young Chinese game-players, many of who got hooked on the industry after playing games in Internet cafes.

• I think there’s a lot of holes in these arguments (as you’ll see in my comments throughout the Google doc), but I also think it’s well-worth the time to mediate on the economic/social/political implications of video games. The essay points out that Nvidia started its business in 1993 with a game graphics card. The GPUs that power the training of AI algorithms were originally used to enhance the graphics of computer games. Visit any AI startup and you’ll see video games littered throughout the workspace. Game-playing performance is a core aspect of measuring AI progress (has anyone heard of AlphaGo or AlphaStar?).

DIVE CHALLENGER DEEP

This essay also links to a debate in the economic history literature re: a demand-based theory of industrial revolutions as opposed to the more conventional, supply-based theory with technological change as the key factor. This goes back to (at least) 1932 when Elizabeth Gilboy argued that an outward shift of the industrial demand curve was what triggered the Industrial Revolution. In an article questioning the Gilboy thesis, Joel Mokyr, a leading economic historian, captured the significance of this debate, "In the process of what could best be called 'making sense of the Industrial Revolution,' few articles have been more influential than Gilboy's eloquent plea to view demand as an equal partner in bringing about the most profound economic change in human history."

Regardless of where you stand on this debate, the essay is right to focus on the diffusion of technology rather than the initial invention. Consumer preferences and the willingness to adopt and continuously improve initial innovations (which includes attitudes toward video games) are a crucial part of the technological diffusion process.

DISCUSSED IN THE FULL TRANSLATION (in the style of Believer magazine): Garbusov-style characters who are not interested in emerging technologies, cyperpunk science fiction, why Rome abandoned the steam engine, the game Steve Jobs made at Atari, and a $6262 calculator

FULL TRANSLATION: The Meaning of Games: The Cotton and Oil of the Information Revolution

ChinAI Links (Four to Forward)

Must-read: China Digital Times Series on Sharp Eyes (雪亮工程))

Dahlia Peterson and Josh Rudolph, with research assistance from Cindy, use publicly available, online Chinese sources to shed light on Sharp Eyes, a surveillance program launched in May 2015, often overshadowed by discussions of the social credit system and use of facial recognition to shame rule-breakers. Includes a map of various Sharp Eyes projects that illustrates the breadth of the CCP’s surveillance goals.

Key points:

1. Sharp Eyes is a rural-focused initiative intended to address a lack of police presence and security camera coverage in the countryside

2. The program has been lauded in some areas: The LA Times heard praise from villagers about reductions in thefts of livestock in Sichuan which had installed 40,000 surveillance cameras across over 14,000 villages by Dec 2017

3. But it not only opens up massive breaches of personal privacy but could be counterproductive for security - getting to “no blind spots” is unlikely, technical difficulties, and vulnerabilities in systems that rely on connecting a massive amount of cheap, insecure IoT devices to TVs/street cameras in order to surveil the masses

Longer notes:

• “The people have sharp eyes” (“群众的眼睛是雪亮的”) was a Communist Party slogan used during the Cultural Revolution, an era in which citizens were spied on and falsely accused neighbors and even their own family of being disloyal to authority

• Sharp Eyes builds off of predecessor programs such as Skynet and Safe Cities and where one program stops and another begins is often blurred

• In March 2018, Yitu, one of China’s 14 AI unicorns, won a deal worth 18 million RMB to provide facial recognition and vehicle recognition technology to Deqing County’s (in Zhejiang) Public Security Bureau for its Sharp Eyes project

• Subcontractor Guangdong Aebell described the Sharp Eyes system, in a now-censored post (English translation), as one that “appropriates household TV sets and smartphones to enhance the extension of surveillance systems to households and individuals.”

• Chinese media reports claim that in 2017, Xinjiang was the site of 30 of the 80 nationwide public security projects each worth over 100 million yuan. Nine of the 80 were Sharp Eyes projects.

Should-read: The More Things Change….the Dark Side of Techno-Utopianism

Andrew Marantz for the New Yorker recounts the fascinating, darker history of how “the initial effects of the printing press included heightened ethnic tensions, the spread of medical misinformation, and about a century’s worth of European religious wars”

Should-listen: ChinAI Pod #1

We launched the ChinAI podcast this week with inaugural guest Remco Zwetsloot, who talked about his “Strengthening the U.S. AI Workforce: A Policy and Research Agenda.” We should be on all the places you get your podcasts now (just search “ChinAI,” but if we’re not, you can paste this link into your podcast app and the feed should pop up: https://api.substack.com/feed/podcast/2660/private/df055a23-b213-4016-9d0b-cbd168059b40.rss

Should-listen Rough Translation Pod on ChinesE storytelling pod Gushi FM

Rough Translation (which explores how familiar stories are told in unfamiliar settings around the world), is one of my favorite podcasts. Producer Jess Jiang walks us through an episode of 故事FM about a Chinese mother who secretly sets up an online dating profile to find her daughter a husband.

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 Rhodes Scholar at Oxford, PhD candidate in International Relations, Researcher at GovAI/Future of Humanity Institute, and Research Fellow at the Center for Security and Emerging Technology.

Check out the archive of all past issues here and please 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

Welcome to the first episode of the ChinAI podcast, hosted by Jeff Ding. There’s a lot of great podcasts out there that get interesting people to talk about their latest work on either the surface or the euphotic zone layer (the layer of the ocean where sunlight still penetrates); the ChinAI podcast seeks to dive into the deepest layers of a guest’s work — whether that be a new report, paper, epic poem, etc. Just like how other podcasts in different domains are structured around the guest’s newest movie or latest book, this pod will feature engagement with a guest anchored to a specific piece of text, which Jeff will have actually read in its entirety, including all the footnotes and other relevant literature.

Each episode consists of four sections:

1. The Briefing Checklist (BCL): The main takeaways from the piece, with the expectation that this is a five-minute brief for a high-level principal

2. Debate the Guest (DTG): Too many podcasts consist of everyone saying “Yes, I agree…and now here’s my somewhat related point.” In contrast, Jeff will say, “This is a specific claim in your report that I want to directly rebut…” and we’ll use that as a starting point for further debate.

3. Footnote Fever: Jeff and the guest share their favorite footnotes in the report, investigate the underlying assumptions/indicators behind the main findings, and explore the surrounding literature which influenced the work.

4. Trust the Process (TTP): Some of my favorite podcasts are ones where we get an inside-view at the process behind someone’s work (e.g. how a movie director finds the perfect location for the scene, how a soccer coach plans tactics). In this section, we want to unpack the research process, which includes how a guest’s personal story ties into their work.

Our inaugural guest is Remco Zwetsloot, a Research Fellow at Georgetown’s Center for Security and Emerging Technology. His writing on the security dimensions of artificial intelligence has appeared in Foreign Affairs, Lawfare and other publications. He is also a Research Affiliate and Ph.D. (D.Phil.) candidate at the University of Oxford’s Center for the Governance of AI. He has previously worked at OpenAI and holds degrees from Yale University (M.Phil., Political Science), the University of Oxford (M.Phil., International Relations) and University College Roosevelt (B.A., Social Science). Follow him @r_zwetsloot

*****Timestamps: Briefing Checklist (3:45); Debate the Guest (11:25); Footnote Fever (35:00); Trust the Process (46:15)

He joins the ChinAI podcast to discuss “Strengthening the U.S. AI Workforce: A Policy and Research Agenda” a Center for Security and Emerging Technology publication he wrote with the help of Roxanne Heston and Zachary Arnold. The report argues: A prolonged talent shortage could undermine U.S. strength in artificial intelligence, and current immigration policies place the country's AI talent advantage at risk. It lays out what is currently known about domestic and global AI talent, identifies priorities for U.S. policymakers and describes policy-relevant knowledge gaps that researchers should fill.

Welcome to the ChinAI Newsletter!

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a Rhodes Scholar at Oxford, PhD candidate in International Relations, Researcher at GovAI/Future of Humanity Institute, and Research Fellow at the Center for Security and Emerging Technology.

Check out the archive of all past issues here and please 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).

Feature Translation: Fu Ying’s Preliminary Analysis of AI + International Relations

This week features an informal translation of a significant article by Fu Ying (傅莹) published in April 2019 in the Quarterly Journal of International Politics (国际政治科学), a Tsinghua University publication considered to be a top ten Chinese-language international relations journal.

Fu Ying, the author, is also a big deal: she’s been the Chairperson of the Foreign Affairs Committee of the National People’s Congress of China since March 2013, and is best known for her terms as the ambassador to the United Kingdom and as Vice Minister of Foreign Affairs. A 2019 Report by the Hoover Institution of Stanford University stated that in Sino-U.S. think tank-to-think-tank interactions, “Fu Ying emerges as the senior figure in a growing number of US-China interactions.” Most relevant for this translation, she’s the honorary dean of Tsinghua’s Institute of International Relations — a very influential body where Yan Xuetong is the dean —where she directs a center for International Strategy and Security, which helped with gathering info for this article.

Big ups to Brian Tse who found this article and also was the lead translator on this effort. He’s a Senior Advisor at the Partnership on AI, Policy Affiliate at GovAI/Future of Humanity Institute, and has advised leading AI labs including OpenAI, Google DeepMind, Tsinghua University AI Lab and Beijing Academy of AI. His research focuses on the prospect of international cooperation of AI safety and security.

*A friendly reminder that paying for a subscription helps compensate awesome ChinAI contributions like Brian’s from this week). We’re at 68 subscribers, and our (very arbitrary) goal is to get to 100 by the end of September.

What follows are Brian’s main takeaways (at the end, I’ll add on some brief, less put-together reflections):

In his report Understanding China's AI Strategy, Gregory Allen at the Center for New American Security discusses his trips to China in 2018 focusing on AI. One of the key findings is that in a keynote speech during China’s largest international relations conference, Fu Ying said that Chinese technologists and policymakers agree regarding the “threat of the new [AI] technology to mankind.” She further stated that “We believe that we should cooperate to preemptively prevent the threat of AI.” This view is consistent with Fu Ying’s analysis in this article, which uses the “Mars invasion” metaphor to describe how AI might be a threat to the security of the global commons:   

"If we look at the world as a zero-sum game and a pursuit of absolute security, there is no doubt that AI will, similar to the way atomic bombs and satellites were in the 1940s and 1950s, become the new focus of competition among big countries and a driving force of two or multiple parallel global orders. However, if we adopt the perspective of a Community of Shared Future for Mankind and view the problem through the concept of common security, it is not difficult to realize that the security and governance challenges brought about by AI technology are problems that all people face together. In this way, it should not be difficult for us to jointly explore the norms acceptable to all stakeholders in the spirit of equal consultation. If so, will AI become the "Mars invasion" challenge that unites China, the United States, Russia and the rest of the world?

“A Community of Shared Future for Mankind” is a major concept in current China’s foreign policy and diplomacy, and could be the broader conceptual framework of how Fu Ying, as a former Chinese diplomat, thinks about the international order and the impact of AI. This framework indicates a significant shift from Deng Xiaoping’s reform era dictum to “hide our capacities and bide our time” (韬光养晦) to a more proactive approach of calling upon China to be a responsible stakeholder in international affairs and global governance under Hu Jintao, and later, Xi Jinping. In a document published by the State Council laying out its thinking back in 2011, it lists a number of nontraditional, global security threats that can only be addressed by all countries working together, including the spread of weapons of mass destruction, financial crises, natural disasters and climate change. Eight years later, the former Chinese diplomat seems to be adding AI to the list of global security concerns that call for a new paradigm of international relations. 

The choice of “Mars invasion” as a metaphor can also be analyzed through a Chinese cultural lens. Three recent science fiction movies, two of them blockbusters and the last hyped to be one , all related to the themes of space and a collective global challenge. In the movie The Martian, Matt Damon gets back to the Earth safely thanks to the cooperation between NASA and the China National Space Administration, and the head of China's space program said that the movie shows that Americans want joint cooperation (also see how I began my talk at Asilomar Conference on Beneficial AI). Regarding The Wandering Earth movie, based on the novella of the same name by Liu Cixin, the South China Morning Post writes that it “emphasizes global collective actions and international cooperation” when the group of astronauts try to guide the Earth away from an expanding Sun, while attempting to prevent a collision with Jupiter. Lastly, The Shanghai Fortress depicts the human race's last stand in Shanghai fighting against aliens who try to seize a hidden energy source on earth in 2042.

Three weeks ago, Fu Ying gave a public talk on the topic of “AI Governance and International Cooperation” (link to the speech, Chinese only) at the World AI Conference in Shanghai (the same conference where Jack Ma and Elon Musk had a debate). Following these analyses and public speeches, Fu Ying is setting up a think tank, the Center for International Strategy and Security Studies at Tsinghua University (link to the opening ceremony, Chinese only), that has a major focus on AI, China-US relations and international security. The think tank is currently recruiting a full-time team of researchers (link to the job postings, Chinese only).

Jeff’s key takeaways:

• This is something that is high enough in quality that I would include as a must-read in the Four to Forward section if I had come across it in English-language form. Fu Ying and her team are very plugged in on the English-language discourse in this space. Here is just a sampling of the references to the English-language texts in this article that I had not heard of (and it’s supposedly my job to study this space): The Human Web: A Bird's-Eye View of World History by McNeil and McNeil; Brooking’s report “How artificial intelligence is transforming the world;” "A 21st Century Science, Technology, and Innovation Strategy for America’s National Security" issued by the U.S. National Science and Technology Council in 2016; "The Influence Machine: Automated Information Operations as a Strategic Defeat Mechanism" - a report issued under a Association of the U.S. Army research institute.

Here’s a quick list of interesting points (I’ve highlighted all of these with comments in the Google doc full translation):

• Fu Ying emphasizes how the a gap in the ability of different countries to exploit AI may lead to a further deepening in the “digital gap.”

• On AI + military strategy, she writes, “Currently, whether one is discussing "algorithm warfare" or "swarm" tactics, the heated debate in the strategic community still revolves around analyses of the operational impact of a single type of technology. If the military application of AI technology cannot be recognized as a whole, the envisaged countermeasures may become a costly and ineffective new "Maginot Line.”

• She calls for more inclusive and multi-tiered systems of governance for AI given the emergence of more diverse actors and decentralization of political power associated with the tech

• A good stat: the country with which China has initiated the largest number of international cooperation in the past five years is the United States; similarly, the country with which the U.S. has initiated the most international cooperation is also mainland China, and the amount of co-authored papers far exceeds that with other countries.

*It’s important to note that in the Hoover report we cited above re: Fu Ying’s emergence as the senior figure in Sino-U.S. think-tank/think-tank interactions also noted that in a case involving Fu Ying, “a US think tank was strongly advised to exclude a well-known China specialist as a condition for a meeting going forward.” In a similar vein, there has been some discussion about the Chinese space program’s key role in The Martian as an example of “pandering” to China’s interests, though to be fair the book it was based on includes China’s assistance as a plot point.

FULL TRANSLATION — Fu Ying: A Preliminary Analysis of the Impact of AI on International Relations

ChinAI Links (Four to Forward)

It’s an all-CSET week of links, in line with an epic debut week of three report publications from the Center for Security and Emerging Technology at Georgetown’s Walsh School of Foreign Service.

The first report, “Strengthening the U.S. AI Workforce,” by Remco Zwetsloot, Roxanne Heston, and Zachary Arnold, is a must-read on the AI talent landscape, the key problems facing the U.S. AI workforce, and possible solutions. Remco, who’s also a Research Affiliate with GovAI, happened to be visiting Oxford this week, and I got a chance to sit down with him to thoroughly dissect the report. God-willing, we’ll release that conversation/debate as the inaugural episode of the ChinAI podcast next week. Stay tuned!

CSET also launched policy.ai, a biweekly newsletter curated by Rebecca Kagan on artificial intelligence, emerging technology and security policy, which will include links to Chinese-language document translations (very excited about this!) and interesting reports. This week’s debut issue featured a great translation by Ben Murphy of a Chinese government notice giving Chinese companies guidelines on how to build AI open innovation platforms. Via the newsletter recs, I also found this awesome Institute for Defense Analyses report which outlines a tentative framework for examining U.S. and Chinese expenditures on AI R&D, which tests the hypothesis: “information in Chinese government documents and reported in the Chinese media is too poorly defined—in terms of timing, sources, and purpose—to support credible comparisons between U.S. Federal and Chinese central government expenditures on non-defense AI R&D.”

The second CSET report, “Immigration Policy and the U.S. AI Sector,” by Zach, Roxanne, Remco, and Tina Huang, argues that the U.S. should build new immigration pathways for AI students, workers, and entrepreneurs. The report is backed by an impressive array of primary-source, original data collection from a great data team, including analysis of Crunchbase data, Department of Homeland Security and Department of State data, a July 2019 expert poll conducted by CSET, and analysis of Department of Labor PERM (Permanent Labor Certification) records.

The last report, “China’s Access to Foreign AI Technology: An Assessment” by Wm. C. Hannas and Huey-meei Chang, is a refreshingly incisive take on technology transfer, which clears up misconceptions, namely the belief that espionage is the main component of Chinese technology transfer. It also makes a useful distinction between “legal” tech transfer (joint Sino-U.S. research, company buyouts, talent recruitment) and “extralegal/informal” tech transfer, which are not as subject to outside scrutiny (document acquisition facilities, front organizations for PRC offices).

Thank you for reading and engaging.

Shout out to everyone who is commenting on the translations - idea is to build up a community of people interested in this stuff. You can contact me at chinainewsletter@gmail.com or on Twitter at @jjding99