So you would run the risk to let it autonomously decide that humans are a threat to it and all humans must be eliminated or coerced into slavery à la Matrix ?

So you would run the risk to let it autonomously decide that humans are a threat to it and all humans must be eliminated or coerced into slavery à la Matrix ? The concern doesn't look very far-fetched to me. If the goal of any natural life-form/object is to maximize its energy level while having a stable structure (whether an an atom, a molecule, RNA, a cell, the body, your kinship, a company, a nation, humanity, a solar system etc...), then what would need to be encoded in those AI programs to make them autonomous? And then what decisions would they make as completely different, life forms with their own decision making? Would they empathize with us (which means they would have to have an idea of the human experience) or would they treat us as a lion treats an antelope, as we treat a cow, etc.?