Posted 2/20/2024, 11:50:24 PM
Pentagon Eyes ChatGPT for Military Uses But Worries About Reliability and Security Risks
- Pentagon exploring military uses of large language models like ChatGPT; sees potential but also risks like vulnerabilities and unreliability
- Possible applications include intelligence analysis, decision-making support, war-gaming
- However models still frequently hallucinate inaccurate information, a major obstacle
- Adversaries could potentially hack models and make them spill sensitive data
- US sees urgency to develop LLMs given China's AI ambitions, plans $2 billion investment in AI to stay competitive