The biggest concerns (and reassurances) of China’s military AI research

Concerns about China’s military applications of Meta’s Llama AI models have sparked debate about U.S. AI export controls and competitive strategy in the global AI race.

Recent developments: Chinese military researchers have published studies showing their use of Meta’s Llama AI models for military applications, prompting calls for stricter U.S. export controls.

• The People’s Liberation Army (PLA) reportedly used Llama’s 13b version for intelligence analysis and military information queries
• Researchers also applied Llama 2 to support electronic warfare and jamming strategies
• These revelations triggered immediate congressional concern about U.S. companies sharing advanced technology

Technical context: The AI models in question are far from cutting-edge technology in today’s rapidly evolving landscape.

• According to Lymsys’ Chatbot Arena rankings, Llama 2 ranks 98th in capability
• Llama 13b ranks even lower at 155th place
• The technology’s relative obsolescence makes it comparable to using Windows XP for military purposes

Credibility concerns: The source and timing of these research publications warrant skepticism.

• PLA research has a history of releasing potentially inflated claims for political effect
• A similar incident occurred in 2022 with questionable claims about quantum computing breakthroughs
• The timing and nature of these studies suggest possible political motivation

Strategic implications: The situation reveals more about competitive dynamics than technological threats.

• China’s use of Meta’s products demonstrates the success of U.S. open technology strategy
• American technological openness has historically led to global adoption and investment
• This default status provides crucial revenue and data streams that fuel continued innovation

Critical challenges: The real concern lies in China’s commitment to practical AI implementation.

• Chinese businesses lead U.S. companies by 18% in applying generative AI
• Success in potential conflicts may depend more on effective use of available technology than having superior systems
• China’s ability to innovate within constraints demonstrates strategic adaptability

Policy considerations: Rather than restrictive measures, the U.S. needs to focus on enabling better AI development and implementation.

• Hasty export controls could impede American global competitiveness
• Regulatory mapping exercises could help identify and remove barriers to responsible AI use
• Maintaining technological openness while accelerating implementation should be prioritized

Looking ahead: The U.S. position in the global AI race depends more on successful implementation than restricting access to existing technology. While maintaining vigilance against genuine threats, policymakers should focus on removing barriers to innovation and encouraging practical applications of AI technology across industries.