Think what you like. Opinions of those in the field are all over the map, and my own opinion -- and major concern -- is that without human-like biological foundations, artificial thinking is bereft of feeling (the largest, deepest, and most ancient part of our intelligence) which means, among other things: NO EMPATHY is possible. Humans share a common DNA language and much body and mental architecture with each other and with animals; we also share many of the same basic needs (food, air, water, etc) and desires. We feel love and fear and awe and pain, we sense our kinship with others from their body language, facial expressions, needs, and more. A supercomputer has none of that; at most it has programmed mimicry.

A machine AI will be a psychopath; it won't necessarily be motivated by malice (unless programmed, even unwittingly, to act in that fashion) but it also won't feel any deep connection with human beings.

That a cluster of supercomputers will only have third-hand knowledge of whatever daily human actions and experiences people typically have will also work against the AI's having correct and useful knowledge about many human situations, motivations, needs, and so on.