What you mean by ‘smart’ is very important. There is a significant difference between being intelligent (or superintelligent) and being conscious. The former relates to our brain and its operations, which we are in the process of understanding. We can and will create very intelligent AI. The latter is something completely ambiguous — nobody has any idea what consciousness is, how we experience it, and what its connection to our brain is. We don’t know if it’s organic or a natural result of high intelligence.

Are humans at the top of the food chain because we are intelligent, or because we have a consciousness? We don’t know for sure, but we do know that other animals also have a high level of intelligence, and are able to solve problems, recognize patterns, and assess their environment in the same way humans can. Those are also the things that we are currently programming AI to do.

Any AI algorithm will necessarily obey the rules of its creators, Homo Sapiens. Not to say that those creators will not be biased in one way or another, but the AI will only act within the boundaries of that bias. An algorithm designed to avoid risking human lives at all costs will do so because it has been programmed to do so, not because it want to or has decided to. In order for an AI to become our overlord, it would need to either disobey our instructions or develop motivations of its own. Both of these require consciousness. Therefore, we can design a superintelligent AI without having that AI rise up against us.

This is not to say that AI will never become conscious— for all we know, it might just be a natural result of high intelligence. But in any case, an AI that is able to cause serious existential risk will need to be more than intelligent — it would need to be conscious. This makes one thing much clearer: if we want to minimize existential risk and maximize existential hope, we need to understand ourselves and our consciousness before we understand how to code intelligence into a computer.