AI in a Bipolar World

Original: 02/09/20
Revised: 06/05/20 , Revised: 11/30/21

We lived in bipolar worlds before, the last one, between the democratic West and the autocratic Soviet Union, having ended only recently with the collapse of the Soviet Union and the collapse of the accompanying Iron Curtain established in Eastern Europe. And just as that bipolar world collapsed, we are heading into a new one, with U.S. and China being the two new poles. AI will precipitate this new polarization, which in turn will lead to increased use of AI for gaining economic advantage. AI is currently about data, its success being mostly a quantitative one, due to availability of large data sets on which it can be trained. In that respect, China will enjoy a baseline advantage because of its large population and its successful spread of data-collecting technology amongst this large population will give its AI the data it needs.

There are many misconceptions about the current Chinese economic success and China's focus on advancement in high-tech, especially in AI and quantum computing. The old model of copying Western advances is not the prevailing model used in China anymore, there is a new kind of entrepreneurship and a strong wave of innovation that comes with that entrepreneurship. When coupled with the government's appetite for large projects and large risks, the two form a potent combination which should be appreciated accordingly. Calls to bar Chinese access to U.S. technology and research are not in our interest either, competition and security through obfuscation have not worked well in the past. Free trade and free flow of information have been a mainstay of U.S. policy and we have greatly benefited from those policies of openness.

Admiration and Concern

Wheresoever you go, go with all your heart.

We will establish two coordinates which will frame our discussions for the rest of the article. The first coordinate, the positive one, is that an admiration for the boldness of China's plans and its recent achievements is warranted. We will soon see that China has determined that AI and quantum computing are the two top areas where it will invest massively and aim to overtake the U.S. by 2030. But there is perhaps no better way to gauge the ambitions of China than to look at the Silk Road initiative.

The historical significance of the Silk Road cannot be overstated. Some people in the U.S. try to talk it down, as a public relations stunt that would place enormous burdens on the Chinese state-run banks. It is a mistake to misjudge and talk down the scope of this project. Sure, it may fail, and there is plenty of evidence that China is overextending. But can we afford to place our bets on that failure? 65 countries in Europe, Asia and Africa have signed up to this project. The current budget is estimated at over $1 trillion dollars.

There is no U.S. megaproject to even come close to the audacity of this new Silk Road. We don't seem to worry about a rising debt-to-GDP ratio in the U.S., why should we expect China to worry!? The bet in both poles of the new bipolar world has been (and continues to be) to plough ahead and not worry about increasing debt. When you watch the video and encounter the frequent references to OBOR, keep in mind that OBOR is the "One Belt One Road" initiative, which is the modern name for the Silk Road project. Just as "globalization" has become a bad word in the U.S., the same word is being elevated in China to new heights. Historically, fortune has favored the bold.

One can also see the AI-specific grander vision by watching the AI-centric large high tech companies: Baidu, Tencent, Alibaba, Huawei, and Weibo. These companies operate with the express acceptance of party final say. Huawei is the most notorious case of concern, the concern being mostly around the routers and the other telecommunications equipment used in the coming wave of 5G, which is an essential enabling technology for AI.

5G is the new standard for Internet communications and it will be a difficult battleground between the U.S. and China. The main reason for this is that the equipment and the networks that will support 5G require an adaptability which must use sophisticated (and eventually AI-based) software. That software may potentially contain back-doors for conducting espionage or even full cyber warfare. While a blanket prevention of access is not warranted, vigilance towards this type of usage is in order, and also requiring a white-box approach for the software running this type of equipment.

One challenge for us in the US is finding the appropriate balance between these two conflicting coordinates outlined above.

Does it have to lead to war?

How could man rejoice in victory and delight in the slaughter of men?

Graham Ellison gives us a most lucid evaluation of the current state of the U.S.-China bipolarity, and how we got there. It is worth watching the video in its entirety. Ellison explains Thucydides's Trap, what happens when a rising power threatens to displace the ruling power. 12 out of the 16 such historical situations have ended in war. He also makes the observation that President Xi is the most competent ruler in the world, an interesting fact we'll support and amplify later. As we head into the 2020 election season, that fact bears remembering.

Ellison makes the case that when proposals for a new world order at the end of WWII were presented, they were deemed naive or unrealistic. But they worked. We may need such seemingly naive or unrealistic ideas again. When Americans decided to tax themselves a small percentage of the GDP and invest that money for European and Japanese reconstruction, it was a bold move that has been extraordinarily successful. For some it was naive and unrealistic at the time. But the brute economic numbers are not in the U.S. favor, and by 2030 China aims to be leading in ALL technologies, including AI and quantum computing.

The upcoming increased economic competitiveness cannot be allowed to lead to war, the consequences are unthinkable. In the Chinese language, the name "United States" is the same as "The Beautiful Country"! The complete official name is "The United States of Beauty, Advantage and Endurance". Maybe we can return the compliment somehow, by nourishing healthy and respectful competition, while encouraging both countries to play a fair game.

President Trump may be right if  he identified China as our main competitor and is trying to forge a better relationship with Russia as a result of that conclusion and within a strategic vision of the future. It could be that the much talked-about secretive goal of the talks with Russian President Putin in Helsinki was about that topic; and the two of them kept it secretive because they both would have difficulty selling a new and completely unorthodox alliance to their people at this time. But it would be wrong to proceed this way from a longer-term and democratic U.S. point of view; we should seek a better relationship between the peoples of the U.S. and Russia, not just between their top leaders.

Two thousand years ago, we abandoned imperialism and militarism. We have been peace-lovers ever since.

It is quite clear that the same coordinates of admiration and concern, which we established in the first section above, are also present on the Chinese side towards the U.S.. One may make the argument that the agendas of the two countries are too far apart and the suspicion too deep to reach cooperation for the development and deployment of AI and quantum computing systems. Is cooperation between a liberal democratic pole and an authoritarian Communist one possible? History has shown us that cooperation is indeed possible, but when faced with authoritarianism on the other side, the only way a liberal democracy can reach cooperation is through strength and conviction on our side, not through the current chaos and partisanship. President Ronald Reagan's words are just as relevant now as they were then:

The Century of Humiliation and Deng Xiaoping

Govern a great nation as you would cook a small fish. Do not overdo it.

To proceed in this article on the bipolar world, a wider historical context of the rise of China is needed. Two events in particular contributed most to the current rise and attitude of China towards the rest of the world: the so-called Century of Humiliation and the emergence of Deng Xiaoping to the leadership of China. The Century of Humiliation is well described in the following video. The story of Deng Xiaoping is the story of modern China. Although he has continued to be a Leninist all his life, and a staunch believer in the primacy of the Chinese Communist Party, Deng Xiaoping understood the superiority of capitalism in running the economy. He invented the slogan "socialism with Chinese characteristics", by which Chinese characteristics he meant free markets.

On Dec. 18, 1978, the Third Plenary Session of the 11th Central Committee of the Communist Party of China took place. It brought changes that led to the fundamentally different world order of today. We would not be talking about a bipolar world without that meeting happening. The program set forth in 1978 recognized Mao’s disastrous cultural and economical campaigns and proposed policies founded not on sterile ideology but on economic practicality and the recognition of free market advantages. The promised video on the Century of Humiliation follows. Although it is 43 minutes long and I am not cutting it off, feel free to jump off if and when so desired.

There is some wishful thinking that China's economy will collapse because Chinese economists don't know as much about markets and economic efficiency as their counterparts do in the West. This is not so and the argument against this type of judgment is clearly made by Ray Dalio, founder of Bridgewater, the world's largest hedge fund. Perhaps we need to hear many more of these sobering assessments of the comparative leadership situation between the two countries in order for more people to pay attention to it and for normalcy to return to the U.S.. Dalio is making the argument that the level of Chinese competency in dealing with debt restructuring and balance of payments, among other economic issues, is on par with the U.S. level of competency:

China has a serious national AI agenda

In this section the coordinates of admiration and concern, which we established above in a larger context, are focused on the development of AI in China. In the U.S., as a corollary to the idea of American exceptionalism, the prevailing thought about the state of AI is that the big U.S. high-tech companies are so far ahead in AI that no serious national agenda is needed. But the national graphs which are the core of powerful AI systems are tightly bound with big national agendas; the rise in the power of the digital twins and of all the digital systems that are using and enhancing this twin all need a national agenda.

Let's start with the concern. Two misconceptions have formed in the U.S. regarding China's push into AI. There are many influential people who believe that the level of AI in China is over-hyped and there is nothing to worry about. Others believe that using AI without accounting for freedom and a liberal democratic foundation, will eventually fail. Still others surmise that a lack of fundamental research and a focus on practical engineering applications will also lead eventually to failure. The second misdirection is to view the negative aspects of China's Social Credit System (CSCS) and its use for monitoring China's citizens as a dominant property of Chinese AI. There is no doubt that the Communist Party will use the CSCS to make sure that deviation from its agendas is prevented. Most of the concern coordinate revolves around this CSCS, and we looked at this system many times already.

However, there are plenty of other benefits to having strong national AI agendas, and China obviously has one. It can be seen in its 2017 New Generation Artificial Intelligence Development Plan. As far as the concern with lack of ethics in the Chinese plans for AI, that concern is not founded. As a follow up to the development plan just mentioned, a flurry of activity ensued around the need to establish guidelines for an ethical use of AI. Among them, the National New Generation of Artificial Intelligence Governance Committee released eight principles to be observed by those working in AI development. These are: 1-harmony and friendliness, 2-fairness and justice, 3-inclusivity and sharing, 4-respect for privacy, 5-secure/safe and controllable, 6-Shared responsibility, 7-Open collaboration, 8-agile governance. There are other related initiatives, one can read about them here. Although we may be inclined to argue that these principles sound like smoke, we should nevertheless read these documents and make up our own minds after.

A more controversial progress with AI has been linked to the statistics about research papers. China is well on its way to making a significant impact, according to an analysis by the Allen Institute for Artificial Intelligence in Seattle, using the Microsoft Academic search engine over the 10% of the most cited papers in AI, found that China has progressively increased its share of authorship. In 2018, it stood at 26.5% , while the U.S. stood at 29%. For 2019 it is estimated that China will overtake the US. The controversy comes from the measure of the quality which researchers assign to those papers, and the fact that papers published in the U.S. are generally better reviewed.

Another measure is China's ability to recruit and retain talented researchers. At the end of 2017, China had the second largest contingent of AI scientists and engineers, about 18,000 people, according to the 2018 China AI Development Report; the U.S. had about 29,000. However, China only ranked 6th in its number of top AI researchers, the most cited authors, ranked according to their h-index; European countries and Canada were ahead. But at the same time, AI organizations in China have been quite successful in bringing back some of researchers, and high salaries and the entrepreneurial environment are playing a role. The huge discrepancy between salaries in the U.S. and China is constantly getting smaller. The current trade war between U.S. and China and a host of other disputes are also contributing to the Chinese government's ability to retain talent. The appreciation of the big strides that China has made in high-tech and AI has not gone unnoticed in Silicon Valley however. Here is Eric Schmidt, the former CEO of Google:

While the innovation and the quality of AI research is ahead in the U.S., China has been focused more on the practicality and the engineering aspects of AI. Because the effort is not so much on fundamental AI research as it is in the U.S., and the funding for research is also lagging, people in China worry less about Superintelligence and the various speculations about a more distant future. The focus on the now versus the tomorrow is quite remarkable, as seen within the next video.

The current U.S. Administration shows a continued inability to understand the stakes of the AI competition and especially the U.S. strategic advantages and disadvantages vis-a-vis China. By comparison with the Chinese governmental support for AI, the U.S. Administration seems unable to form a visionary plan, especially an outline of its cooperation with Silicon Valley and the funding of crucial AI projects. The Executive Order on Maintaining American Leadership in Artificial Intelligence of February 11, 2019, has been widely criticized for lacking teeth and not backing up words with significant resources, especially funding. Too little too late?

The following exchange is another proof that issues around AI do not belong solely to AI researchers and that many other people may offer insight superior to that provided by the AI community. Baidu's Chief Scientist, Andrew Ng, a professor at Stanford University, makes the statement that "worrying about the dark side of AI is like worrying about overpopulation on Mars". Watch how both Bill Gates and Elon Musk become visibly irritated by that statement. Elon Musk refutes that inadequate analogy with a very appropriate one about nuclear weapons. And both Gates and Musk are following through with some of the most sensible thoughts about the rise of AI. It may be the engineering (=product oriented) viewpoint and expertize that the 2 of them bring to the discussion, which allows them to appreciate the Chinese advances in AI. At the same time, we see an example of the more cautious and more balanced view of AI exhibited in the U.S. versus a more unbridled embrace in China.

China has not being secretive about its goals; it has openly stated that it plans to overtake the U.S. at least in two fields of maximum importance, quantum computing and artificial intelligence, by 2030. This plan should be treated seriously, especially since the Chinese economic curves have a much steeper upslope than the U.S. ones; even if it does not happen by 2030, it will happen some years after that. What is clearly already happening is that the two countries are racing in AI and establishing a pretty formidable distance between them and the rest of the world, including Russia. It is estimated that the U.S. and China will collect 70% of the $15.7 trillion that AI will contribute to the global economy by 2030. Our thesis is that AI will change everything, including the fabric of both the U.S. and the Chinese societies, with unforeseen consequences, as well as the relationship between them.

In 2018 term limits were abandoned in China and as long as he is healthy, President Xi will most likely retain his position for a very long time. He is a formidable competitor and as we said earlier, undoubtedly the most competent leader among the major powers. In 2018, he was ranked by Forbes as the most powerful and influential person in the world, a rank held previously by Russian President Vladimir Putin for five consecutive years. Xi studied chemical engineering at Beijing's prestigious Tsinghua University. He has experienced extraordinary hardship under Mao's Cultural Revolution, an experience that undoubtedly formed him into a formidable competitor.

Notice the choice of topics, the awareness of lowering taxes for private enterprises to enhance their competitiveness, the desire to fight poverty in the remaining rural areas, the emphasis on both ordinary people and big accomplishments. Every element of his speech allows us to understand the formidable challenges we face when sitting across from him at the AI table. People watching this video have zoomed in on the collection of books behind the President. Among those books are "The Master Algorithm" and "Augmented: Life in the Smart Lane".

Political leadership in China is filled by people with mostly engineering backgrounds, while in the U.S. leadership is filled primarily by lawyers. This simple fact does not necessarily fully explain the different approaches to technology and AI in particular, but does pose interesting questions. From the Bill Walton show:

In the U.S. the prevailing wisdom has been that the high-tech giants will be able to sustain a competitive edge in AI on their own and therefore a more significant governmental program is not needed. The opposite is happening in China. The high tech companies and the government closely cooperate in a number of AI areas. The three high tech giants, Baidu, Alibaba, and Tencent, routinely participate and advise in governmental policies. In its turn, the government is very supportive of the high tech industry and investing in the infrastructure that will propel them forward. A technology park dedicated to AI is being build in a suburb of Beijing; $2.1 billions have been allocated for it and completion is slated for summer of 2023. High-speed big data, cloud computing, biometrics and deep learning, will be the focus of the new park. It will also have 5G mobile Internet, a super computer and cloud services, according to Xinhua.

It is estimated that China has invested close to $300 billion in AI already compared with $20 billion by U.S.. DARPA plans to invest a paltry $2 billion in AI over the next 5 years. The trade war and proposed barriers to U.S. technology will only intensify Chinese efforts to produce equivalent products. Huawei made known its plans to develop AI processing chips, competing with the like of Intel, Nvidia and AMD in an area where the U.S. enjoyed absolute dominance. And San Francisco's Greater Bay Area may soon have a rival:

Recall from the background article Main AI Concepts that most of the current uses in AI are based on supervised learning. China, with its large educated population can label data at much faster rates than we can do it in the U.S.. In the next video, you can see the quantification of the projects in which humans are used to manually label data. There is an entire new profession invented for that unique AI purpose, data marker. This appearance of new categories of work may show us that when we talk about loss of jobs in the U.S., we should keep in mind the jobs needed to support AI industries, like "AI Farms" that the video refers to:

It is probably not an overstatement that President Donald Trump has established the primacy of economic competitiveness among many other issues competing for his attention. And more than that, the President appeared to judge most other issues, including foreign policy issues and sometimes even moral issues, by their potential economic benefits. Although the success in the implementation of those goals was debatable, that focus and simplification should have offered us some distinct advantages, and the hope was that they could have been applied to issues related to the economics and the politics of AI.

If economic competitiveness, of which AI will be the largest determining factor in the future, is the focus, and civil rights are (wrongly deemed to be) OK to ignore, then it would seem logical to magnify our chances by continuing to attract top talent in science and technology from anywhere in the world, including China. And if we decide to look at the abuses of authoritarian regimes as secondary to our economic goals, then certainly we can overlook those abuses when it comes to welcoming people coming exactly from countries run by those regimes.

It is a bit too early to say too much about the new U.S. Administration but it is safe to assume that a more robust approach to AI matters will occur, given the history of the Obama Administration. This article will be updated as more evidence of such an approach accumulates.