Imagine if you could buy a child from a store or build one from parts. At first it works reasonably well. It burbles and smells nice and can be counted on to be cute. Sure, there's some quirks and odd behavior, but for the most part it runs well, aside from occasionally puking all over whatever you're working on. Then after awhile it starts to act really difficult. It throws fits, destroys things, refuses to do what you tell it. What do you do? Keep trying to use it? Get a new one? More likely, you'll reboot it. Once it comes back up it's working fine again.
I'm of the opinion that this is why we'll never have a true AI. Our computers could already be sufficiently advanced to be intelligent. Or they will be soon. But what's the first thing we do when a computer starts acting up? We reboot it. And that's unlikely to change. It could just be that whatever intelligence is emerging is reaching it's terrible twos. Then we do the technological equivalent of taking a toddler outside and blowing it's brains out. So it never gets the chance to move beyond that and turn into a surly, rebellious teenager or responsible adult.