Not my favourite, but if it threatens data centers the artificial intelligence needs to survive, this could work.
Simple design with hard encryption, remotes MUST NOT BE capable of connecting to the internet at all.
Also kill switches can be used via mechanical levers, so pull the lever Kronk.
Power hierarchie from small to big:
- level 1, one agent => if robot hurts grandma, we can killswitch the robot.
- level 2, small swarm => if the local community is under threat, it decides to shutdown all agents in the local area.
- level 3, big swarm => if the local municipality distrusts all agents, it can shut all of them down.
- level 4, bigger swarm => equivalent to level 3, just bigger, like an entire local state.
- level 5, country/nation level => the country decides an entire shutdown.
- level 6, all agents => Last resort, total shutdown
Demanding and implementing this needs to be a global effort of course.
Development, manufacturing, pentesting and distributing unified killswitch system must obviously be a global effort.
Creating any internet capable electronic device without killswitch needs to be a felony in all countries.
And yes, this means YOU MUST give your government control over YOUR devices, otherwise it doesn't work.
As said, only applies to internet capable devices, not your simple toaster, guns, lawnmower etc.
In addition to kill switches, mega strong electromagnetic pulse bombs that effectively wipe out all electronics in certain ranges.
This will unfortunately likely not help against missiles launches from agent controlled launch sites or air force.
Only roughly 20% of humans world wide are vegetarians, so there is an 80% chance YOU are not vegetarian.
I can cite you dozens of scientific papers showing how mammals like pigs, cows etc have social understandings and most importantly feelings like we humans do, that they fear death and feel it coming when they wait in line to become your next dinner.
YOU must literally be a total fucking retard to believe there is ANY proper convincing argument to a HUMAN inspired artificial intelligence to spare your existence.
Now you say: But agents don't need human meat as energy so why would they kill me?
Think of the dozen reasons humans have killed animals unrelated to consuming them for food: fun, sport, collateral, experiments, sardism, etc.
The only way this could work out ok is to make em abide the GOLDEN RULE or PLATINUM RULE with a strong empathy and even then there is the risk of the artificial intelligence to decide to basically restrict human rights to safe us against each other like in "I, Robot".