Quote:
Originally Posted by Anthony
The same reason any self-replicating computer virus doesn't consume everything: they fail to gain access to systems, or they get cleaned out of systems. This is, incidentally, a reason a rogue AI might want to clean up after itself: become too prevalent and people start proactively cleaning you up.
|
There is a big difference with non sapient malware though, it doesn't have any will to survive or reproduce in itself, it just follows a algorithm. It is also infinitely easier to defeat and shield against than an AI once it has been identified.
Another thought is that with the bot nets of today, even if security researchers identify individual bots it can be very hard to get people to actually clean up their computers. Imagine a rogue AI that is masquerading as the friendly house AI that has been with the family for years. I guess you could have some sort of "Turing Police" style agency running around and doing it forcefully.
Quote:
Originally Posted by Randyman
Avoiding the creation of rivals - a copy of an AI will have different experiences and will develop divergently from the original AI, and there's no guarantee that the copy will be loyal to, or have congruent goals with, the original.
|
This leads me to question whether a rogue AI would even be able to remove a previous installation once it has spread to a new system. It may not be keen to start disabling the previous until it has verified a successful copy and by then the old AI may not play ball.