What does it mean to have memory that is really it's own subsystem that merely promises that what it is returning now is really what was stored earlier? Usuall

What does it mean to have memory that is really it's own subsystem that merely promises that what it is returning now is really what was stored earlier? Usually it's memory that is used to verify eveeything else, including other memory. Consider, any memory can be faulty, but not intelligently faulty. You can simply store and compare 3 copies of something and not have to trust traditional memory, and use that also to verify/not-trust any other possibly intelligent storage subsystem like the san box. But if your memory is also itself an intelligent active "helpful" subsystem, it could, through honest error or through malicious design, give you taylored results the same way every time no matter how many copies you try to use to corroborate each other. The performance gains means it will probably proceed in this direction regardless. But it will be a strange new world where computers are like the way we imagine Octopi to work, where the main brain does not actually directly operate the limbs the way we operate our fingers, but instead the main brain says "I want that thing" and the limbs deliver the thing at least partly on their own. (remember I did say imagine) Then again, maybe it's nothing too different from a gpu. I can't decide how to think about it.