ChinAI Newsletter #5: China's Views on Military AI + a Strong Chinese Advocate against Strong AI Emerges
|Jeffrey Ding||Apr 9, 2018|
We're at 400 subscribers in four weeks - thanks for supporting the newsletter by sharing!
These are Jeff Ding's weekly translations of writings on AI policy and strategy from Chinese thinkers. I'll also include general links to all things at the intersection of China and AI. Please share the subscription link if you think this stuff is cool.
I'm a grad student at the University of Oxford where I lead the Future of Humanity Institute - Governance of AI Program's research on China AI happenings.
White Paper on Civil-Military Fusion Presents Realistic Barriers to Integrating AI in the China's Military
Theme of this issue is China's views toward AI militarization, so all three of this issue's translations are related to military AI.
The first is a translation of a White Paper on Civil-Military Fusion, which contains a section on how AI and other emerging technologies fit into China’s (still nascent) military industrial complex. H/t to Lorand Laksai of the Council on Foreign Relations for his great piece on China’s civil-military fusion, which linked me to this white paper. My main takeaways, which were included in a recent Shephard Media article (paywalled/link in the ChinAI links section):
Major barriers for the application of AI in China's military modernization efforts can be characterized in two categories
1. Issues unique to AI technologies:
-Military equipment requires a high degree of reliability, so emerging technology products (e.g. AI-enabled military tech) will not be quickly adopted in military weapons and equipment. The more likely scenario is that AI-enabled military tech will be used as technical reserves, which will allow these technologies to be repeatedly tested and verified.
-Advancements in AI are moving too quick for the Chinese military to catch-up. China's product cycle/modernization cycle for military technology is notoriously slow, and as AI technologies continue to improve quickly, any AI-enabled military technology that is developed and scaled will likely be outdated by the time it is deemed reliable enough to be deployed in the field. This leads me to conclude that AI technology will not be widely used in China's military in the short term, however, there will be opportunities for rapid integration of AI technologies in the future.
2. Broader issues with China's civil-military integration efforts which will present barriers to China's militarization of AI:
- China's military industry, oriented toward monopolies controlled by state-owned enterprises, suffers from unsustainable valuations and corruption. Talented researchers are not incentivized to work for these enterprises because there are strict restrictions on the shares available for employees. Moreover, investment institutions, who could potentially spur development in the military industry, are deterred because companies do not share information (due to confidentiality/national secrets concerns) for these institutions to properly value these companies.
- Compare this to countries like the U.S., which has a robust military industrial complex, including in the realm of defense informatization (e.g. In-Q-Tel, the CIA's venture capital fund, was the initial sole investor in Palantir technologies, which is now involved in many important national security contracts across the U.S. military). China is trying to introduce its own strategic investors into these state-owned enterprises to help incorporate more private enterprises into the military industry, but this will take some time as it will require changes in culture and leadership. If China hopes to incorporate AI into its military modernization efforts, it will need to significantly improve its civil-military integration efforts.
Zhou Zhihua: Why We Should Boycott Korean Universities Developing Autonomous Weapons
Zhou Zhihua is an emerging strong voice in the Chinese AI community against Strong AI. Based on the articles I've read, strong AI is defined as AI that exceeds human capabilities across many domains.
From my translation: Recently, the Korea Advanced Institute of Science and Technology (KAIST) and the military enterprise "Hanwha Systems" cooperated in the research and development of autonomous weapons, encountering a joint boycott by more than 50 AI scholars worldwide, among which Professor Zhou Zhihua of Nanjing University was the only mainland scholar.
Zhou Zhihua states that strong artificial intelligence “cannot be done and should not be done!” From the perspective of technology, research direction, as well as the threat posed to humans, he elaborates on the point that researchers should not touch strong artificial intelligence.
Zhou Zhihua is the Director of the Department of Computer Science and Dean of the Institute of Artificial Intelligence at Nanjing University.
Tencent Book Chapter 24: Moral Machines
In a chapter on Machine Ethics, the book discusses AI in warfare: At present, many countries are actively researching and developing military robots, and an important development trend of military robots is the continuous improvement of autonomy. For example, the X-47B UAV developed by the U.S. Navy can achieve autonomous flight and landing. Countries such as South Korea and Israel have developed sentry robots. They have automatic modes and can decide whether to fire or not. Obviously, if military robots are not controlled in a certain way, they are likely to have no sympathy for human beings and will be merciless to achieve their goals. Once they are started, they may become truly cold-blooded "killing machines." In order to reduce the harm that military autonomous robots may cause, they need to be allowed to abide by human ethical standards, such as not harming non-combatants and distinguishing between military and civilian facilities.
I learned a lot about machine ethics in the course of reading this chapter. It engages with Wallach and Allen's book on Moral Machines and talks about value alignment as well. Includes strong rhetoric on the need for AI ethics: However, in essence, self-driving cars, intelligent robots, and various virtual agent software are no longer passive tools in human hands, but they are agents of humanity, with autonomy and initiative. This presents a major challenge to ethics and morality. Previous ethical norms for humans and human society now need to be extended to intelligent machines, and this may require a new ethical paradigm.
This Week's ChinAI Links
My recent work:
- Paywalled article on my China Dream report and thoughts on civil-military fusion
- A short piece for the China Policy Institute at the University of Nottingham that summarizes the main takeaways from my China Dream report
Battlefield Singularity - great work by Elsa Kania at CNAS on how AI could empower China's military revolution
Thank you for reading and engaging.
Please feel free to comment on any of my translations in the Google docs, and you can contact me at firstname.lastname@example.org or on Twitter at @jjding99