Chinese Researchers Allegedly Modify Meta's AI Model for Military Use
A recent report by Reuters has brought to light that Chinese research institutes affiliated with the People's Liberation Army (PLA) have reportedly made modifications to Meta Platforms Inc.'s massive language model, Llama, for potential military applications.
According to the report, six Chinese academics from three different institutions, including two under the PLA's Academy of Military Science (AMS), worked on tweaking an early version of Meta's Llama model to create a specialized artificial intelligence model known as ChatBIT. This customized AI model was designed for processing military intelligence data and enhancing decision-making processes.
Development of ChatBIT
A research article published in June revealed that the team utilized Llama 2 13B to establish an artificial intelligence system specifically tailored for military communication and question-answering tasks. Subsequent enhancements were made to the model to outperform existing similar models in the market.
While the specific capabilities of ChatBIT are not fully disclosed, it is reported to have been trained on approximately 100,000 military dialogue records, a significantly smaller dataset compared to conventional large language models.
Meta's Usage Policy
Meta Platforms Inc., in adherence to its approved use policy, prohibits the utilization of its AI models for military purposes, including applications related to warfare, nuclear industries, or espionage. Molly Montgomery, Meta's head of public policy, emphasized to Reuters that "Any usage of our models by the People's Liberation Army is unauthorized and goes against our acceptable use policy."
