3

The year is 2050 and there is a evil AI lurking in the web. For some reason it wants to kill humans to appease its creator the old god HEATHER KARUZA. It can access any computer and use any device excluding nuclear weapons and other WMDs because of mumble mumble handwavium. It is housed in a nether realm accessible by killing thousands of children and using their blood in a ritual. So no stopping it this way.

How can the AI kill every human quickly and effectively before the system admins shut it down?

L.Dutch
  • 286,075
  • 58
  • 587
  • 1,230
11Bravo
  • 2,082
  • 9
  • 37
  • 1
    So many hanwavium that what make you think it can be shut down? And why it could be pinpointed to one device even? – SZCZERZO KŁY Dec 22 '20 at 16:17
  • "killing thousands of children and using their blood in a ritual. So no stopping it this way" depending on how specific your nether realm is with what is considered a child or not that could be very easy to do – Topcode Dec 22 '20 at 16:20
  • https://worldbuilding.stackexchange.com/q/119399/30492 looks like a duplicate – L.Dutch Dec 22 '20 at 16:21
  • Sorry didn’t see that one?? Is it too similar? – 11Bravo Dec 22 '20 at 16:22
  • 2
    obligatory xkcd reference: https://what-if.xkcd.com/5/ – Trioxidane Dec 22 '20 at 16:23
  • This is a classic high concept question. VTC:Opinion-Based. Please visit the [help/dont-ask]. It states, "To prevent your question from being flagged and possibly removed, avoid asking subjective questions where ... every answer is equally valid [and] you are asking an open-ended, hypothetical question." – JBH Dec 22 '20 at 17:12
  • 1
    @L.Dutch-ReinstateMonica Questions that have answers that could answer another question are not really duplicates if the questions themselves are fundamentally different. Aliens have all sorts of tools at their disposal that a rogue AI would not. At most I would call that a related question. – Nosajimiki Dec 22 '20 at 22:06

2 Answers2

4

Through leveraging it's influence

Assuming that this rogue AI has advanced programming/hacking capabilities and that most of the most damaging stuff armies around the world can leverage are not atteignable from the internet, some quite dangerous stuff is still reachable. Some power plants, maybe even a nuclear one. But that's not where it hurts the most.

The AI cannot just switch off humankind, but it can set it on a way to it's own destruction. Manipulating opinions, creating fake news, creating false political tensions and making them worse.

The key here is enough subtlety to avoid being shut down right away.

Intercepting communications and subtly altering them is quite effective; a simple warning email from a secretary in the [some middle eastern country] embassy suddenly looks like a coded call for a military exfiltration, the president of a mighty country makes Nixon and Clinton looks good because he's been exposed by adding one too many people on his very private message to his very private secretary/CIA trained robbers/secret South-American murder machines... you get the idea.

Weapons are waaay less effective than misdirected people. Populations reacting to viral false flags and news based on events which never really happened.

And the stock market. Nothing moves faster than an AI to mess with the market, and by extension with human value. Crashing the market? Well, maybe later, but for now we'll make sure that all those eco-friendly companies looks like they will crash while the weapon dealers and the most polluting companies are having record profits! When a population is poor enough to be on the edge of civil way, let's devaluate their currency!

Some good people are pushing back? They'll have to do it the old way by speaking to crows in coffee shops and beer parlors, because on the internet their voice will both be downvoted to hell and drowned into mindless objections and flame wars. The AI can code it's own army of chat bots and advertisers, and it never sleeps.

We are doomed, it's just a matter of time. THEN the AI can use it's bit-money acquired assets in robotics and whatnot to make his own Skynet seek-and-destroy robots to finish the job, his awful diseases available to everyone who don't want it and, why not, even rent the help of awful people to hunt and kill the others. We're going Mad Max meet the Terminator and 12 Monkeys very quickly... just not overnight!

laancelot
  • 704
  • 3
  • 6
2

Reverse the vacuum ventilation fans and open the exterior air vents on the CDC and USAMRIID Biosafety Level 4 Clean Rooms, along with all of their counterpart in every biowarfare research centers in other nations around the world.

Henry Taylor
  • 69,168
  • 14
  • 116
  • 248
  • Why would you assume those are connected to a network? – Keith Morrison Dec 22 '20 at 20:22
  • 2
    Moreover, why do you assume they're physically capable of reversing? A lot of dedicated one-way systems have mechanical venting that prevents air from flowing in reverse specifically as safety measure. – Keith Morrison Dec 22 '20 at 20:38
  • 1
    @KeithMorrison, we can both definitely hope that the safety systems on all of the most dangerous laboratories on Earth are all designed with great wisdom and conservative restraint. Given that each of those systems was built by humans (a species not known for either wisdom or restraint), I would bet that at least one of those systems has a connection to the internet and has reversible fan motors. I hope that that is not the case, but I wouldn't bet against the AI in this scenario. – Henry Taylor Dec 22 '20 at 21:21