Human cybersecurity mechanisms would be ineffective against a superintelligent AI, much like an ant's defenses against a human