Chaina and US :According to academic papers and analysts, prominent Chinese research institutions associated with the People’s Liberation Army have developed an artificial intelligence (AI) tool with potential military applications by utilizing Meta’s publicly available Llama model.
Six Chinese researchers from three institutions, two of which are part of the Academy of Military Science (AMS), the leading research body of the People’s Liberation Army (PLA), described how they had used an early version of Meta’s Llama as a base for what it calls “ChatBIT” in a June paper reviewed by Reuters.
A military-focused AI tool was built by the researchers using an earlier Llama 2 13B large language model (LLM) that Meta META.O, incorporating their own parameters, to gather and process intelligence and provide precise and trustworthy information for operational decision-making.
According to the paper, ChatBIT was improved and “optimized for dialogue and question-answering tasks in the military field.” It was discovered that it performed better than some other AI models, which were about 90% as capable as OpenAI’s powerful ChatGPT-4. The researchers did not specify whether the AI model had been put into use or how they defined performance.
Sunny Cheung, an associate fellow at the Jamestown Foundation who specializes in China’s emerging and dual-use technologies, including AI, stated, “It’s the first time there has been substantial evidence that PLA military experts in China have been systematically researching and trying to leverage the power of open-source LLMs, especially those of Meta, for military purposes.” He is also an expert on China’s AI.
Meta has embraced the open arrival of a significant number of its computer based intelligence models, including Llama. It requires services with more than 700 million users to apply for a license from the company, among other restrictions on their use.
The terms also say that the models can’t be used for “military, warfare, nuclear industries or applications, espionage,” as well as for developing weapons and content that “incite and promote violence.” These are all activities that are subject to US defense export controls.
However, Meta has limited means of enforcing those provisions due to the public nature of its models.
Meta stated that it took measures to prevent misuse and cited its acceptable use policy in response to questions from Reuters.
In a phone interview, Meta’s director of public policy Molly Montgomery told Reuters, “Any use of our models by the People’s Liberation Army is unauthorised and contrary to our acceptable use policy.” Montgomery was speaking to Reuters.
Meta added that open innovation must be encouraged in the United States.
A Meta spokesperson stated in a statement, “The alleged role of a single, and outdated, version of an American open-source model is irrelevant in the global competition on AI, when we know China is already investing more than a trillion dollars to surpass the US on AI.”
Geng Guotong and Li Weiwei from the AMS’s Military Science Information Research Center and the National Innovation Institute of Defense Technology are among the Chinese researchers, as are researchers from Minzu University and the Beijing Institute of Technology.
According to the paper, ChatBIT “will not only be applied to intelligence analysis in the future, through technological refinement, but also… strategic planning, simulation training, and command decision-making will be explored.”
The institutions and researchers did not respond to a request for comment, nor did the Defense Ministry of China.
Although the researchers noted that ChatBIT’s model included only 100,000 military dialogue records, which is a relatively insignificant number in comparison to other LLMs, Reuters was unable to confirm ChatBIT’s capabilities or computing power.
According to Joelle Pineau, vice president of AI Research at Meta and a professor of computer science at McGill University in Canada, “that’s a drop in the ocean compared to the majority of these models (that) are trained with trillions of tokens so… it really makes me question what do they actually achieve here in terms of different capabilities.”
The research comes amid a contentious debate about whether companies like Meta should make their models available to the public in US technology and national security circles.
In October 2023, US President Joe Biden signed an executive order to control AI development. He said that while innovation can have significant benefits, there were also “substantial security risks, such as the removal of safeguards within the model.”
Washington stated this week that it was finalizing regulations to restrict US investments in China’s artificial intelligence and other technology sectors that could jeopardize national security.
John Supple, a spokesperson for the Pentagon, stated that the Department of Defense was aware of the advantages and disadvantages of open-source models and that “we will continue to closely monitor and assess competitors’ capabilities.”
“Cookie Jar” Some observers contend that China’s efforts to develop its own AI, including the establishment of numerous research labs, have already made it difficult to prevent the nation from closing the technology gap with the United States.
Two researchers from the Aviation Industry Corporation of China (AVIC), which the United States has designated as a company with ties to the PLA, described using Llama 2 for “the training of airborne electronic warfare interference strategies” in a separate academic paper reviewed by Reuters.
Domestic security has also been a part of China’s use of AI developed in Western countries. Llama had been used for “intelligence policing” to process large amounts of data and improve police decision-making, according to a June paper.
Commentary on AI’s potential to “accelerate the research and development of weapons and equipment,” “help develop combat simulation, and improve military training efficiency” was published in April by the state-run PLA Daily.
Can you prevent China from entering the cookie jar? William Hannas, lead analyst at Georgetown University’s Center for Security and Emerging Technology (CSET), told Reuters, “No, I don’t see how you can.” According to a paper published in 2023 by CSET, 370 Chinese institutions had researchers who had published papers on General Artificial Intelligence, which contributed to China’s national strategy to be the world leader in AI by 2030.
