With AI, vast bits of information are held and this increases every day. This means that by storing so many options, outcomes in some areas/tasks/results are better than man can arrive at. An example was 'Intellgence' beating the World's top Chess player. But I don't see this as intelligence. It was just selecting the known options programmed into the '*'to beat opposing moves. In a less simplistic way, Robots perform tasks in factories programmed by men in factories and industry with sensors stopping or changing directions in an ever more sufficiated way. Driverless cars are an example. Eventually, people driving will cause more accidents. But do you consider this as intelligence? As regards medicine, we know that AI can quickly compare and evaluate results but unless told so will not of itself evaluate something found at the bottom of the sea as a medical breakthrough unless instructed to do so. As far as I am aware, Robots cannot analyse feelings as people can, even if they can jump higher, run faster and for longer. They do as instructed. Some have been armed and actually used to fire at the enemy and it is likley that will be further developed.
So Robots will take over certain works and displace people as machines did during and after the Industrial Revolution but I don't think they will be sufficiently advanced of their own volition, to turn against mankind as a whole, as sometimes depicted in films. Man might use 'them' or 'it' to decimate populations just like any other advanced weapon that has been adopted aggressively. Sadly, some aggressive traits in man persist to this day. One could say 'AI' could be used as a weapon of war but that applies to other things too, Nuclear Weapns for example. I think 'AI' is more likely to be used to disrupt and cause problems just like we experience presently but to a greater extent It has the potential to be good and lead to advances in a number of fields. So let's look on the optimistic side.