Human Decisions Still Needed in Artificial Intelligence for War
The NSCAI maintains that a global treaty that prohibits the development, deployment and use of artificial intelligence (AI) enabled weapons systems is not in the interests of the United States and would harm international security. It argues that Russia and China are unlikely to follow such a treaty. A global ban, it argues, would increase pressure on law-abiding nations and would enable others to utilise AI military systems in an unsafe and unethical manner.
This is an unsophisticated way of thinking through a complex problem. Negotiations and conversations at the United Nations have been occurring on this matter since 2014. The voices of AI scientists, Nobel Peace Laureates and civil society were not represented as part of the NSCAI’s advice. If science argues against AI weapons, it is difficult to maintain that their development and use would benefit US interests and international security.
Instead of following the NSCAI’s advice, President Biden could take the lead and create an innovative international treaty requiring human control over AI military systems. This means that AI could continue to be used in some aspects of military operations including mobility, surveillance and intelligence, homing, navigation, interoperability and target image discrimination. But when it comes to target acquisition and the decision to kill, states would be required by the treaty to retain human decision-making. This positive obligation should be legally binding.