China’s New AI Military Model with a Touch of Llama
Chinese research institutions linked to the People’s Liberation Army (PLA) have developed an AI military model, ChatBIT, by customizing Meta’s open-source Llama 2 13B.
In a recent paper made in June and reviewed by Reuters, researchers showcased the way an older version of Llama was used for the development of an AI military model dubbed ChatBIT, mainly designed for military intelligence gathering and decision-making tasks.
Adapting Llama for Military Applications
Sx researchers affiliated to China’s PLA, including the Academy of Military Science (AMS) made some alterations to Meta’s Llama in a way that it processes the military needs.
The team added to the Llama 2’s 13B some specific parameters that enables it to perform tasks within the military field.
ChatBIT was fine-tuned to hold conversations and answer questions that are related to military scenarios. During the testing phase of its effectiveness, the Chinese AI military model showed to compete with other large language models (LLMs). They said that the AI model achieved a performance level somehow similar to OpenAI’s ChatGPT, but did not provide any further details about its metrics or even if it was implemented operationally.
“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,” stated Sunny Cheung, associate fellow at the Jamestown Foundation, and specialist in China’s emerging and dual use technologies including AI.
Meta’s Response
It is also well-known that Facebook’s parent company had released many open-source AI models, including Llama, while restricting their use for military purposes, requiring users to ask for license from the company.
Meta’s control over publicly available AI models is limited, as highlighted by the unauthorized use of its Llama model by the PLA.
Molly Montgomery, Meta’s director of public policy, told Reuters that “any use of our models by the People’s Liberation Army is unauthorized and contrary to our acceptable use policy.”
A spokesperson from Meta emphasized that the US could maintain its competitive edge by supporting open innovation, highlighting that China’s ongoing expansion of AI investments, challenging the US tech sector.
Trying to Keep China Out of the Cookie Jar”
The use of AI in the military intensified debates about the risks of exporting US tech to other countries, specifically China. In response, US president Joe Biden signed an order in October 2023 that mandates the management of AI risks to mitigate security-related threats.
Another report showed that China is going beyond just AI models, as researchers from the Aviation Industry Corporation of China (AVIC) has also used Llama for electronic warfare training, in a bid to enhance its military capabilities.
Experts believe that limiting China’s access to open-source AI models is somehow difficult. William Hannas, an analyst at Georgetown University, commented that keeping China “out of the cookie jar” would be challenging, given ongoing scientific collaboration between U.S. and Chinese researchers.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.