A Pentagon advisory board has published a set of guidelines on the ethical use of artificial intelligence (AI) during warfare.
In “AI Principles: Recommendations on the Ethical Use of Artificial Intelligence by the Department of Defense,” the Defense Innovation Board (DIB) shied away from actionable proposals in favor of high-level ethical goals.
In its recommendations, the board wrote that the Department of Defense’s AI systems should be responsible, equitable, traceable, reliable, and governable.
Since AI systems are tools with no legal or moral agency, the board wrote that human beings must remain responsible for their development, deployment, use, and outcomes.
As far as being equitable, the board wrote that the Department of Defense (DoD) “should take deliberate steps to avoid unintended bias in the development and deployment of combat or non-combat AI systems that would inadvertently cause harm to persons.”
To ensure AI-enabled systems are traceable, the board recommended the use of transparent and auditable methodologies, data sources, and design procedure and documentation.
The board recommended that the DoD’s AI should be as reliable as possible, and because reliability can never be guaranteed, that it should always be governable. That way, systems “that demonstrate unintended escalatory or other behavior” can be switched off.
The board called for ethics to be an integral part of the development process for all new AI technology, rather than an afterthought.
“Ethics cannot be ‘bolted on’ after a widget is built or considered only once a deployed process unfolds, and policy cannot wait for scientists and engineers to figure out particular technology problems. Rather, there must be an integrated, iterative development of technology with ethics, law and policy considerations happening alongside technological development,” wrote the board.
Although the public sector, including the European Commission, the United Kingdom House of Lords, and ministries or groups from the governments of Germany, France, Australia, Canada, Singapore, and Dubai have all formulated AI ethics or governance documents, the US is unique in offering AI guidelines specific to the military.
“What is noteworthy when canvassing the plethora of available AI Ethics Principles documents is that there is no other military in the world that has offered its approach to ethical design, development, and deployment of AI systems. In this respect, DoD is leading in this space, showing its commitments to ethics and law” wrote the board.
Since DIB’s recommendations are not legally binding, it is now up to the Pentagon to decide if the board’s guidelines should be followed.