They are focusing on a computer that seems to be able to repair itself if it finds a fault. I'm assuming this means at runtime, but then this begs the question for "How?". I can personally think of ways that this can be done, but depending on the way it repairs itself, it needs some kind of logic arguably, to determine how it's supposed to repair something... (For instance, you don't just get told to ride a bicycle for the first time and you go do it, perhaps you need practice, but you also need to know how to ride a bicycle before you can even start.) Now think of what might happen and how destructive this could be it some other piece of executed code had authority to modify this logic, and essentially turn this "brain" into a good one gone bad.
They already put out the idea that this could be used to repair damaged drones. I can only imagine how screwed up and disastrous this could be if some other piece of code could reprogram it's thinking to do something entirely different, and something in which it was not originally intended to do.
Some way, this logic needs to be completely separated from the workings of the computer itself, or I could probably spell "trouble" with this new innovation.