Nathan Helm-Burger comments on Evidence on recursive self-improvement from current ML