I believe the biggest consideration about AI is this: if a machine managers to reach the human cognitive capabilities then by definition and construction it will be able to surpass these capabilities multiple-fold in a really short time (having access to a vast knowledge pool through internet). That in turn will lead to a superior cognitive entity, which we will not be able to follow, let alone manage and control. Nobody knows were this can lead, and this is exactly why Elon Musk and Stephen Hawking commonly published a call for prudence towards the scientific community.