Intelligence without moral sense is the worst case

From the conversation

After that, chances are it won't come on its own, these capabilities, but people get lost in definition of what is human-level general-purpose AI, AGI, when who cares what definition we're going to take. The only thing is: "Has the AI reached a certain level of competence that can become dangerous?" And it doesn't have to be about everything. It can even be superhuman in one area and stupid in others. In fact, that's the worst case. An AI that lacks the moral sense of what we consider virtuous in human beings, but is otherwise highly intelligent, is the horror.