AI Learns Naval Defense Tactics Through Trial and Error

Leidos and Domino Data Lab conducted a wargame that used Reinforcement Learning to defeat attacks on a naval fleet without harming nearby commercial vessels.

Reinforcement Learning enables machines to learn through trial and error, similar to how young children learn. It maximizes rewards over time to sharpen performance.

Reinforcement Learning is well-suited for preparing autonomous vehicles for novel situations that can't be fully pre-programmed.

Reinforcement Learning could automate complex military operations like managing drone swarms within tight timelines.

Reinforcement Learning raises governance concerns about when self-learning systems should and shouldn't be used in combat situations.