Rohin Shah comments on [AN #114]: Theory-inspired safety solutions for powerful Bayesian RL agents