Artificial Intelligence and the US-China Balance of Power
Name
chang-bachang-PhD-political science- 2021-thesis.pdf
Description
Thesis PDF
Size
5.28 MB
Format
Adobe PDF
Checksum (MD5)
a00ecd5db35d856063f47f187e040c0e
Author(s)
Chang, Benjamin Angel
Advisor(s)
Narang, Vipin
Date Issued
June 2021
Publisher
Massachusetts Institute of Technology
Abstract
How will artificial intelligence affect the US-China balance of power? While a nascent literature debates whether AI may upend strategic stability or revolutionize the nature of warfare, existing discussions suffer from both imprecise conceptualization and scarce data. In three essays, this dissertation evaluates the impact of AI on the nuclear balance, the conventional balance, and long-term US-China competition more generally by focusing on deep learning, generating data through simulation and supply chain analysis.
The first essay defends the focus on deep learning, then presents an end-to-end conceptualization of how its technical qualities translate into usefulness across different categories of modern military tasks, which in turn affect, when contextualized to the particular dyad under study, the strategic balances across different domains of US-China competition. At each analytic layer, the paper condenses deep learning’s effects into several generalizations, tying AI to existing debates in security studies and setting an agenda for future research.
The second essay simulates US-China nuclear war in Python to assess AI’s impact on the strategic balance, focusing on the tracking of mobile platforms on land. It finds that AI reduces the total “effective counterforce area” – the area the United States would have to destroy with nuclear weapons, to carry out a splendid first-strike – by one to two orders of magnitude. Under low to medium alert, the simulation finds this would enable successful US nuclear counterforce. While countermeasures are available to China, the essay predicts heightened nuclear tensions as a result.
Finally, the third essay exploits supply chain datasets to assess each side’s ability to bring AI-enabled autonomous weapons to bear in future conventional conflicts. I find that control over the production of advanced AI chips by the United States and allies almost certainly means the United States would better exploit such weapons, if they emerged as decisive in modern warfare, within at least the next ten years. Potential Chinese policy responses, such as cannibalizing its civilian sector or substituting with older chips, would likely fail for technical reasons.
MIT Department
Massachusetts Institute of Technology. Department of Political Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
