Optimization Regularization through Time Penalty — AI Alignment Forum