This acknowledges the pitfalls that Superior AIs could possibly be misused - by way of example to distribute misinformation - but claims they can even be a pressure once and for all.In reinforcement learning, the agent is rewarded for good responses and punished for bad kinds. The agent learns to pick responses that happen to be categorized as "gre