There are voices among experts that artificial intelligence will be humanity's last contraption. Not because the robots will join troops and raise against us and then carry out an act of annihilation; that's the selected variant of Hollywood's movie studios. We will face another crisis: humans will get lazy and mostly stupid. And that's the scariest aspect of AI.
We have previously outsourced the method of decision-making to software systems. Years ago, it was the easy activities that got automated. Now they target other, more specialized professions. In some fields, that's ok. Software that supports doctors in identifying diseases and proposing treatment is a positive use of this technology. But the moment we get complacent with relying on the outcomes, without questioning how the software worked it out, then it will get dangerous.
There already systems that are a total black box, mostly using deep learning algorithms (neural networks). Even those who design those systems have trouble understanding it. On a mass scale, it's terrifying.
The process of reasoning, arguing, or challenging someone's thinking process was, and still is, the hallmark of humanity's progress that moved us from Stone to the Digital Age. Once it's fully encoded in software logic - in algorithms - then it poses a threatening situation we may never recover from.