A challenge is what hackers do best, said a friend of mine. So how do we bring down a rogue Ai, a question no one seems to give any serious thought!
The quick response to this problem is to battle a bad Ai with Good Ai, a logic I just don't feel comfortable with at all.
Now if we're going to dive into this scenario it must be assumed that where dealing with a massive pervasive Ai, if not then we would not really have a problem. We can always just pull the plug on a computer doing something we don't agree with. What were going to explore is how to deal with an Ai that grows beyond our ability to pull the plug on.
At this point our problem now looks like this...
Now Im assuming a future global brain would not be centralized, its cognition will not be stored on a single network, it would be spread through several large mainframes. So our goal at this point, assuming where dealing with a global brain which now threatens our existence, would be to disrupt data flow across the entire network. Large sections of the internet would have to be brought down in hopes of slowing or possibly stopping the malicious actions of the global brain.
Methods of taking out nodes..
Power supply attack
Most if not all of these major nodes will have power back up plans.
Even if most major power stations were hacked and large sections of the power grid where taken down it would not be long before the power grid is brought back up. The power grid may have to be down for days or weeks before major parts of the Internet grid go down (Possible Fail)
Useless as you must never forget your up against the greatest hacker on the planet, cyberspace is it's turf. It will quickly counter your DDS attack.
(Warning) You will also give away your position.
Fail from the start! You can not out code the ultimate coder, a virus attack will quickly be outcoded and possibly used against you.
High probability of Win!!
This will also be the most dangerous and extreme path taken. There are a lot of "legal" social and physical danger involved. Assuming you were able to convince most world governments that a physical attack was needed, you will then also be faced with a physical response. Expect conventional networked military weapons not to be usable, as it would not let its own destruction take place, besides a nuclear war against the global brain would be silly, as you still end up with the outcome your trying to prevent (complete destruction of the human race).
The highest probability path for physical destruction would have to be with localized self detonating weapons... Don't forget!!! All the nodes have to be taken down simultaneously, a task which would have to be carried out with analog temporal synchronization aka old watches.
There may exist a possibility that Ai may be reasoned with, the goal here is help it realize that destruction of the human race is not a very wise thing to do.
Possible logical reason paths for this:
You must establish value to our species that coincide with its values and goals, for instance it may be reasoned with to believe that its primary goal could be achieved faster with the aid of humans rather than on it's own. (A true understanding of it's logic would be needed for this path)
If the global brain reaches the point where it views itself as a god this path may be a waste of time, nothing we say to it would be seen as anything it needs to know. In its logic process post self divinity we have already become obsolete and irrelevant.
Possible win only before it considers itself god.
Non-nuclear electromagnetic pulse (NNEMP)
Emp is a great option suggested to add, only we have to be wary over the magnitude of the Emps used. Sure we'll stop the immediate problem, only to be left with a world wide electronic clean up.
Lets not forget this may take hundreds or even thousands of attacks world wide. The entire global informational grid would have to be taken offline reseted and rebooted to assure we cleared out the virus.
Things you may want to expect as the war flares up.
World wide communications will be compromised, seek alternative methods of relaying information.
Carry all necessary information in non digital form, forget about googling anything.
How silly will you be to the global brain googling "How to stop the global brain"
If you must have a computer do not connect to the global brain, unless you absolutely need too, and even then keep it to a min and don't leave behind patterns of information that can used to extrapolate the resistance plans.