Global, Network / Cyber
The military is focusing on how to use AI. (Graphics from Breaking Defense, original images from Getty and DVDS)
WASHINGTON – Thirteen months after the United States released the Political Declaration on Ethical Military AI at an international conference in The Hague, representatives from the signatory countries will gather in a Washington suburb to discuss next steps.
“We have over 100 participants from at least 42 of the 53 countries,” a senior State Department official told Breaking Defense behind sharing details of the event for the first time. Representatives of military and civilian officials will meet in closed session. March 19th and 20th at the University of Maryland. college park campus.
“We really want a system that allows states to stay focused on responsible AI issues and keep a real focus on building practical capacity,” the official said.
Topic: All military applications of artificial intelligence, from unmanned weapons and combat networks, to generative AI like ChatGPT, and back-office systems like cybersecurity, logistics, maintenance, and human resources. The goal is to share best practices, discuss models like the Department of Defense's online Responsible AI Toolkit, and build personal expertise on AI policy to bring back to government.
This cross-pollination will help technology leaders like the US refine their policies, while also helping technology proponents in less wealthy countries “get ahead of the curve” before investing in military AI themselves. Dew.
This is not just a chatterbox for diplomats, state officials stressed. Next week's meeting will include a mix of military and civilian representatives, with civilians coming from foreign ministries as well as independent science and technology institutions in many countries. The official said the very process of holding the conference served a useful coercive function by simply letting signatories decide who to send and which agencies within their governments to represent them.
The U.S. government hopes this will be the first of an indefinite series of annual conferences hosted by member states around the world. In between these general meetings, smaller groups of like-minded countries should meet for exchanges, workshops, war games, etc., national officials explained.“Do whatever it takes to raise awareness and take concrete steps'' to implement the Declaration: 10 principles [PDF]. These smaller forums then report to the annual plenary session, which codifies lessons learned, discusses the way forward, and sets next year's agenda.
Related: How GIDE Grows: AI Combat Network Experiments Expand to Army, Allies, and Industry
“We value different perspectives and different experiences, and the list of countries supporting the declaration reflects that,” the official said. “We are very pleased with the broad and deep support we have received for our political declaration.
“53 countries are now participating,” the official said, up from 46 countries (including the United States) announced a few months ago in November. “Look at that list. This is not a US-NATO 'usual suspects' list.”
The signatory countries are certainly diverse. This includes core allies of the United States such as Japan and Germany. Even more troublesome NATO partners are Türkiye and Hungary. Wealthy neutral countries like Austria, Bahrain, and Singapore. Pacifist New Zealand. War-torn Ukraine (which is experimenting with AI-guided attack drones). Three African countries: Liberia, Libya, and Malawi. And even little San Marino. But not just the Four Horsemen (China, Russia, Iran, and North Korea) that have long driven U.S. threat assessments, but also the notoriously independence-minded India (despite years of U.S. cooperation in national defense). His absence is notable. A country with an Arab and Muslim majority.
That does not mean there was no dialogue with these countries. Last November, just a few weeks apart, China joined the US at the UK AI Security Summit to sign the far-reaching Bletchley Declaration on AI in general (not just military). Chinese President Xi Jinping agreed to the vague definition discussion on: we president Joe “The risks and safety issues associated with artificial intelligence,” Biden said after the summit in California. Both China and Russia are participating in the regular Geneva meeting of the United Nations Group of Governmental Experts (GGE) on Lethal Autonomous Weapons Systems (LAWS), but activists seeking to ban “killer robots” are They claim that talks have long since stalled.
Related: The Ethical Terminator, or How the Department of Defense Learned to Stop Worrying and Love AI: Looking Back at 2023
State officials were careful to say the U.S.-led process was not an attempt to circumvent or undermine U.N. negotiations. “Those are important discussions, and those are productive discussions. [but] “Not everyone agrees,” they said. “We know that disagreements will continue to exist in the context of LAWS, but in other forums and on other issues, we cannot allow such disagreements to prevent us from moving forward from where we collectively can. I don't think it's a good idea.”
Indeed, the 2023 Political Declaration, and the Department of Defense's approach to AI ethics that derives from it, is characterized by its reference not only to future “killer robots” and SkyNet-style supercomputers, but also to other military uses of AI. That's what I'm doing. It's not as dramatic, but it's already happening today. This includes day-to-day management and industrial applications of AI such as predictive maintenance. But it also includes military intelligence AI that helps designate targets for deadly attacks, such as America's Project Maven and Israel's Gospel (HabSora).
U.S. officials have long argued that all of these various applications of AI can be used to make military operations not only more efficient but also more humane. “This technology holds great promise,” state officials said. “We see big signs of improvement and we think this will allow countries to exempt her from IHL. [International Humanitarian Law] …So we want to maximize these benefits while minimizing potential downside risk. ”
To do so, the U.S. government believes it will need to establish norms and best practices “across the entire waterfront” of military AI. “It is important not to undervalue the need for consensus on how to use even back-office AI in a responsible manner,” the official said. [by] Having international legal reviews, appropriate training and auditable methodologies. …These are basic foundational principles of responsibility that apply to all applications of AI, whether in the back office or on the battlefield. ”
Updated at 11:48 a.m. ET on March 19 to more clearly reflect that the Political Declaration is part of a whole-of-government approach, not just a State Department initiative.
