Top Chinese research institutions linked to the People’s Liberation Army (PLA) have utilized Meta’s publicly available Llama model to create an artificial intelligence (AI) tool designed for potential military applications, according to academic papers and analysts.
A June paper reviewed by Reuters reveals that six researchers from three Chinese institutions, including two affiliated with the PLA’s Academy of Military Science (AMS), detailed their development of “ChatBIT,” which is based on an earlier version of Meta’s Llama 2 model.
These researchers modified the Llama 2 13B large language model (LLM) with their parameters to build a military-oriented AI tool that can gather intelligence and provide accurate information for operational decision-making. The paper states that ChatBIT was specifically optimized for dialogue and question-answering tasks relevant to military contexts, outperforming some other AI models that were approximately 90% as capable as OpenAI’s ChatGPT-4. However, the researchers did not specify how they measured performance or if the AI model is currently in use.
This marks the first substantial evidence of PLA experts systematically researching and trying to leverage open-source LLMs like Meta’s for military purposes, according to Sunny Cheung, an associate fellow at the Jamestown Foundation specializing in China’s emerging technologies.
Meta has made many of its AI models, including Llama, publicly available but has imposed restrictions on their usage. These include a requirement for services with over 700 million users to obtain a license from Meta and a prohibition on employing the models for military purposes, espionage, or the development of weapons. Nevertheless, since the models are public, Meta has limited means to enforce these restrictions.
In response to inquiries from Reuters, Meta reiterated its unacceptable use policy and emphasized that any use of its models by the PLA is unauthorized. Molly Montgomery, Meta’s director of public policy, stated that the U.S. must embrace open innovation, highlighting China’s significant investment in AI technology.
The researchers involved in this project include Geng Guotong and Li Weiwei from AMS, along with colleagues from the Beijing Institute of Technology and Minzu University. They suggest that, with further technological refinement, ChatBIT could also be applied to intelligence analysis, strategic planning, simulation training, and command decision-making.
China’s Ministry of Defence did not respond to requests for comments, nor did the researchers or their institutions. While the capabilities and computational power of ChatBIT remain unverified, the paper indicates that it was trained on a relatively small dataset of only 100,000 military dialogue records compared to the vast datasets used for other LLMs.
Amid ongoing debates in U.S. national security and technology sectors about the implications of open-source AI models, President Joe Biden signed an executive order in October 2023 aimed at managing AI development due to associated security risks. This week, the U.S. also announced plans to limit investments in AI and other technologies in China that could pose national security threats.
Analysts have pointed out that China’s advancements in indigenous AI development, including the establishment of numerous research labs, are making it increasingly difficult to prevent the country from closing the technology gap with the U.S.
In a separate study, researchers from the Aviation Industry Corporation of China (AVIC), which has ties to the PLA, described how they used Llama 2 to develop airborne electronic warfare strategies. Moreover, the use of Western-developed AI in domestic security has also been reported, with a paper discussing how Llama is utilized for “intelligence policing” to process data and enhance police decision-making.
The state-run PLA Daily has commented on the potential of AI to expedite weapon development, improve military training, and enhance combat simulation efficiency. William Hannas, a lead analyst at Georgetown University’s Center for Security and Emerging Technology, expressed skepticism about the feasibility of keeping China away from AI advancements, citing significant collaboration between Chinese and U.S. scientists.
