Defending Against Adversarial Attacks in Go

Can Go AIs be adversarially robust? Tom Tseng explores defending Go bots against adversarial attacks, using adversarial training and architecture change.