I can't see how pure AGI can be "safe". Huge part of human intellect revolves around the need to survive,be it danger,lack of food,or less rational choices base

I can't see how pure AGI can be "safe". Huge part of human intellect revolves around the need to survive,be it danger,lack of food,or less rational choices based on emotions.If computers can rationalise positive behaviour of humans,they may not be able to do so well with greed, jealousy, hunger for power,that aren't very logical processes but create a lot of positive and negative nonetheless.