Analysis of early development trends of generative artificial intelligence in the U.S. Department of Defense

Introduction: The application of generative artificial intelligence has exploded in the past year. They use large amounts of historical data to learn and can quickly generate text, images, videos and other content based on simple guidance prompts. Since its emergence, the U.S. Department of Defense has rapidly paid attention to the development of generative artificial intelligence and taken some initiatives.

Generative artificial intelligence (AI) capabilities such as large language models (LLM) are growing in popularity, capability and influence globally. These capabilities are trained on large-scale data sets to produce detailed and apparently coherent content that previously required human authoring. These capabilities bring new opportunities, but also significant new risks. There is an urgent need for the defense sector to explore the application of this technology and the potential of the scale, speed and interactivity of these models to improve mission effectiveness while identifying appropriate protective measures and mitigating various associated risks.

Establishment of "Lima" working group

The U.S. Department of Defense announced the establishment of an artificial intelligence technology task force, "Task Force Lima", on August 10, 2023, to focus on researching and deploying generative artificial intelligence tools such as large language models. The Lima Working Group, led by the Pentagon’s Chief Digital and Artificial Intelligence Office (CDAO), will develop, evaluate, recommend and monitor the implementation of generative AI technologies across the Department of Defense to ensure that the Department can responsibly and securely design, Deploying and using generative artificial intelligence technology ensures that CDAO remains at the forefront of cutting-edge technology while safeguarding national security. The "Lima" working group will provide guidance and make recommendations for consideration by relevant decision-making bodies.

Comprehensive goal

To enable the responsible and unified adoption of this cutting-edge technology, the Lima Working Group will provide the Department of Defense with the framework and initiatives to achieve the following comprehensive goals:

• Accelerate promising generative AI initiatives and joint solutions;
• Unite disparate development and research efforts into DoD communities of practice to accelerate innovation and implementation;
• Evaluate solutions in principles, organization, training, materials, leadership, personnel, facilities, and policies;
• Promote technology education and build a culture of responsible implementation and use; ensure DoD collaborates with cross-agency, international, education, civil society and industry partners on the responsible development and use of generative AI.

preliminary work

On September 18, 2023, the "Lima" working group stated that it will fully understand the needs of the military and the Department of Defense for generative artificial intelligence in the next 18 months to accelerate the understanding, evaluation, and deployment of this technology and monitoring, large-scale integration of relevant technologies with mission adaptability. During this period, the work of the "Lima" working group included:

• Comprehensively understand the working principle of artificial intelligence and the application process of external data, and focus on the connection between military use cases and combat processes, and cooperate with other departments to train multiple types of large language models;
• Develop interim guidance, frameworks and workflows based on generative artificial intelligence for the U.S. Department of Defense, and participate in the Department of Defense’s actual research and deployment of large language models;
• Establish a periodic meeting mechanism to allow internal personnel at all levels to keep abreast of the plans and methods being formulated by the "Lima" working group;
• Issue a request for information to the industry to understand and explore the risks and benefits of the development and integration of innovative artificial intelligence.

Summary

The mission of the "Lima" working group is actually to develop, evaluate, recommend and monitor the generation of artificial intelligence capabilities for the entire U.S. Department of Defense, and to responsibly ensure that the Department of Defense integrates these technologies at scale and in a controlled manner. The Lima Working Group has divided the overall mission into several phases. The current first phase is to learn as much as possible about everything out there, inviting the public and industry to help the Department of Defense "further understand and explore the development, acquisition and integration of artificial intelligence." The Risks and Benefits of Intelligence” to take a quick look at the uncertain and transformative power generative AI presents in defense, another short-term task is to provide the Pentagon with interim guidance on generative AI.

The fundamental benefit of this preliminary work is that the Lima Working Group can now establish many industry-recognized standards and security measures when looking to acquire low-risk commercial software. According to the White House order issued on October 30, the Department of Defense is a perfect test bed for the government to responsibly advance and realistically train generative artificial intelligence technology in a closed environment. Whatever guidelines, best practices or standards the Lima Task Force codifies must be shared across the federal government to inform the work of other departments and agencies.

Regarding whether generative artificial intelligence technology can be widely used, the key issues that the US military hopes to solve include: how to avoid the problem of machine illusion, how to measure the performance of the model, how to formulate trustworthy artificial intelligence metrics, and how to manage the data used for training the model. data, etc., and the US military is also working with industry to jointly solve the above problems. There are still many unknowns about whether the U.S. Department of Defense will eventually deploy generative AI at scale, how and for what purposes. (Source: Beijing Lande Information Technology Co., Ltd.)

The above is the detailed content of Analysis of early development trends of generative artificial intelligence in the U.S. Department of Defense. For more information, please follow other related articles on the PHP Chinese website!