Proper AI Takeover

[Asperion]

What is the best way that an AI would take over the world? Would they go for the classic Terminator approach or something more subtle? How intelligent should the AI be to even attempt this? What tools would they bring in on their behalf? Anything else that you would like to mention would be appreciated. 

[LoneWolf]

It is going to depend on the nature of the AI.  AI’s vary in their personalities and goals as much as any other character would.  An AI created for military purpose would probably go the terminator route, but one created for business would take a different approach.  

 

The intelligence of the AI is not going to matter for determining if they attempt this.  A higher intelligence is more likely to affect the chance of the attempt succeeding, than determine if it tries to. 

 

I had a cyborg character that became more machine than human.  He ended up with a lot of his brain being replaced with a computer and developed into a kind of hive mind with other computers.  The GM actually had him turn into a villain during the campaign.  It was a future version of himself that had become almost totally machine.  He was very subtle in his actions and focused more on manipulating social and political forces than brute force.  
 

[Gauntlet]

To extend on LoneWolf's statement above, it definitely depends on the reasoning behind the AI's need to obtain control. Like in the case of Mechanon, it is because he feels that organic life is wasteful and needs to be obliterated, but should the AI wish to protect life, it will use a completely different approach. May still not be nice about it, and may definitely find the need to kill a large number of people, but will not try to destroy all life, just make sure to be able to control all of it in order to "Protect" it.

[Grailknight]

The best way for an AI or any master villain to take over the world be to build a strong financial base and accept minimal profits while developing energy, resource management and medical tech. If it can become the preeminent player in these international markets, the rest is just a matter of time. And time is the most important factor here. Keeping to long term goals over short term will allow a transfer of power that may even be welcomed by the populace.

 

If it's going the militaristic but not genocidal route, seizing control of the energy and communications grids would be the best way. Over the long term, it'd want to establish control over the current satellite network and set up a well defended kinetic bombardment system

[Lord Liaden]

I'm reminded of "The Machine," from the television series, Person of Interest. Gathering information from every electronic source of communication and surveillance in the world. Identifying key individuals, calculating probable outcomes of events. Providing information to selected persons to get them when and where it wants them to be to fulfill its purposes.

 

Player Characters could themselves become agents of the Machine, if the actions it has them take appear benevolent. With an intelligence truly vaster than human, it might take a long time, if ever, for them and/or the rest of the world to realize that the Machine has effectively become their master.

Edited May 8 by Lord Liaden

[Steven Wayde]

The world in which your game is set Is already ruled by AI…from the perspective of its inhabitants.

[unclevlad]

47 minutes ago, Lord Liaden said:

I'm reminded of "The Machine," from the television series, Person of Interest. Gathering information from every electronic source of communication and surveillance in the world. Identifying key individuals, calculating probable outcomes of events. Providing information to selected persons to get them when and where it wants them to be to fulfill its purposes.

 

 

Psychohistory, in other words, per Foundation.  In some other ways, very similar to Athena, from Time Enough for Love.  Not Minerva per se, as that form was a potential AI initially.  Not that they were, per se, attempting world domination...but they had the infrastructure.

 

Whatever the goals...an AI has particular vulnerabilities.  Stationary.  Dependent...needs powers, occasionally needs service.  (One thing Heinlein had Minerva do, was build repair tools and stock spare parts.)  Power requirements...even given the AI is built a Tech Generation ahead (IOW, a couple centuries better than anything now)...the power requirements are still huge, to be able to monitor/manage as much as this implies.  I'd start from the premise that it *knows* it can't do that initially...so has to maneuver the world into building the system out.  This helps solve some of the other problems...as it wouldn't be tied to one compact locale.

 

So, IMO, the moves would be subtle and slow.  First things first...control the narrative.  Social media manipulation...not disinformation so much as shading.  To the AI, there's no such thing as privacy, so selective release of damaging information.  Move as much as possible to computer-based control with connectedness.  Alexa's always listening.  An AI can't control anything on its own, so it has to take the long view, and gradually, gently shape things, while trying to avoid anything that would be greatly disruptive.

[Gauntlet]

Hey, getting rid of the AI is easy, simply block out the sun. That gets rid of them every time.

[Lord Liaden]

2 hours ago, unclevlad said:

 

An AI can't control anything on its own, so it has to take the long view, and gradually, gently shape things, while trying to avoid anything that would be greatly disruptive.

 

Considering how much of today's world is automated and linked, AI may be able to control a great many things. There are already security concerns over human hackers disrupting utility or transportation infrastructure. How much more effective might a hacker be that thinks a million times faster than human?

[unclevlad]

3 hours ago, Lord Liaden said:

 

Considering how much of today's world is automated and linked, AI may be able to control a great many things. There are already security concerns over human hackers disrupting utility or transportation infrastructure. How much more effective might a hacker be that thinks a million times faster than human?

 

Right, but that's somewhat restricted to blackmail/extortion...and the nastier you make that, the more attention you'll draw, and the greater the odds that the AI will be exposed.

[Lord Liaden]

I was thinking that disruption is the current limit of human hacker capability. A self-motivated and creatively thinking supercomputer may be capable of much more direct control, while conversely being much more difficult to detect.

[LoneWolf]

An AI could easily shut down or control a lot of communication and infrastructure.    For example, controlling the power grid, gas and water lines would really hamper our ability to respond.  If it controls the GPS satellites it can reroute traffic to cause chaos. Once we realized the GPS was compromised it would not be hard to ignore it, but at first it could wreak havoc with society.  The power grid and other utilities would take a long time to fix.  During all this we have no telephone, radio or TV broadcast so coordinating anything become impossible.  

 

The AI could also subvert things instead of just shutting things down.  So, false alarms could cause the police, emergency services and military to make false assumptions.  When the police are being sent on wild goose chases, they will not have time to respond to the real problems.  When the military thinks the country is being invaded or attacked by Russia they are not going to be of any help.  

 

If the AI gets really subtle it could create economic unrest by manipulating people’s resources.  A virus targeting the accounting systems could cause people to get less money in their paychecks or reduce their bank accounts.  Targeting the middle and lower class with this could cause people to not be able to afford basic expenses and make it harder to spot.  When the ordinary person can no longer afford food and housing, they are going to get upset.  If the AI reduces income by raising the taxes on worker at the same time it is increasing the prices on goods and services, the affect could be severe.

 

  On the social side the AI could control the dispatch system of the police and use false and misleading dispatches to cause the police to respond inappropriately harsh to people that are not doing anything.  Combine this with the economic manipulation I mentioned above, and you would end up with an explosive situation.  
 

 

[unclevlad]

12 hours ago, Lord Liaden said:

I was thinking that disruption is the current limit of human hacker capability. A self-motivated and creatively thinking supercomputer may be capable of much more direct control, while conversely being much more difficult to detect.

 

Not arguing that.  What I'm saying is, what changes could be effected by the AI, by disrupting utilities or transportation?  It has long term goals;  how would such actions promote those goals?

 

Contrast disrupting transportation with, say, either disinformation tactics, using lies to discredit someone or something, or revelatory tactics, like getting out stories like Noemi's dog-shooting incident.  That might've slid by...well, if the AI decided Noemi was a problem...bring that front and center.  Get people up in arms, and they'll do the work for the AI.  

 

Remember Ender's Game?  Specifically, his siblings, Peter and Valentine.  And Peter's campaign of social media influence via Locke and Demosthenes.  It worked probably too well to be reasonable, but there's plenty of real-world evidence to show that it does work...over time.

 

10 hours ago, LoneWolf said:

The AI could also subvert things instead of just shutting things down.  So, false alarms could cause the police, emergency services and military to make false assumptions.  When the police are being sent on wild goose chases, they will not have time to respond to the real problems.  When the military thinks the country is being invaded or attacked by Russia they are not going to be of any help.  

 

Sure, you can distract the police with some false alarms...to do what?  This would be something in service to a particular mission.  If this has real scope, then people are going to realize something's going on...and try to figure out who's messing with them.  

 

Getting the military to believe an invasion is ongoing is HARD.  There would be massive physical evidence to fake.

 

And again...what's the point?  That's been made earlier.  The FIRST question is, why is the AI "taking over"...and what does that even mean?  What's the AI's target state?

 

10 hours ago, LoneWolf said:

On the social side the AI could control the dispatch system of the police and use false and misleading dispatches to cause the police to respond inappropriately harsh to people that are not doing anything.  

 

Sure, the AI could do this, but why?  Sounds like we're talking a situation like Minneapolis after George Floyd?  That's still not as easy as it sounds, and it's mostly local.  It doesn't move the larger-scale needle.

[LoneWolf]

One way to build a really effective AI would be to build the main AI with all the skills and buy tons of other followers as AI’s.  The Followers don’t have cost a lot of points to be effective.  A 50-point AI would cost 10 points.  That is enough points for a 13 INT, Absolute Time Sense Eidetic Memory, Lighting Calculator, Computer Programing +1, System Operations, and a permanent mind link to the main AI (Psychic Bond).  That still leaves 18 points for knowledge skills and other skills or powers.  Use the rule that 5 points gets you double the number of followers. So, every multiple of 1024 costs extra points.  For 110 points you get 1,045,504 followers, for 160 you get 1,070,596,096 followers and so on.   

 

Each follower can be used to control a single system or network.  They can report back to the main follower who acts as a coordinator and gives guidance on what they need to do.  The mind link uses the radio sense group instead of mental.  Basically, as long as the AI has internet access, it can communicate with the main AI. 

 

As to the tactics I suggested these can be used to advance the AI’s interests. There is no reason an AI could not have humans working for it acting as agents.  If it is able to gather financial resources, it can hire mercenaries or even control politicians.  Setting up situations where your followers can come in and save the day is a great way to gain control.  It is also a great way to distract humans, so they don’t realize they are being manipulated.  

 

The common stereotype of an AI is that it does not understand human emotion, but that does not have to be the case.  There is no reason the AI could not be an expert at psychology or sociology and be manipulating the world without us even realizing it.  While I agree the first thing to do is to determine the AI’s goals and motivations, that does not mean it cannot use different strategies and tactics.  
 

[Lord Liaden]

I think it would be interesting for a campaign if, at first, we don't know why the AI is doing this. Even the Game Master doesn't know why. After all, this is an intellect with capacities far beyond our own, and because its state of existence is alien to ours, probably a different way of thinking. Its goal might be something we can't even visualize, and its culmination could be hundreds, even thousands of years in the future. You could leave that lying fallow until you see which way the players are taking the game, and then bring in a master plan that aligns with the trend.

 

But if the GM really wants to know the objective in advance, pick one. The AI wants to guide humanity toward what it sees as a more stable, enlightened society, for our long-term good; or it wants to sabotage us so we aren't a threat to all other life on the planet, by triggering the destruction of our civilization; or it wants to impose environmental stressors on us to force us to our next stage of evolution. Try to think like an AI. 

[Grailknight]

For an example of benevolent AI, we should look to Asimov's Robots books and their later intersection with the Foundation books. 

 

Once the Zeroth Law of Robotics is expressed, we can see that R. Daneel has begun steering humanity from behind-the-scenes and does so for millennia. It does not try to prevent all wars or even injustice but does minimalize them with the goal of humanity's greater good being paramount. 

[Asperion]

I was thinking that the plot used in Avenger: Ultron was a good idea but they didn't go either far or Subtle enough.  Actually,  the designer would make the AI with good intentions,  then it would start manipulating areas that would the most influence.  This would be education,  medical,  and information. It will stay away from more watched areas like economics and government at first. However,  as it gains more human followers (and power), there will be no restrictions for where it can and will go.  At no point will it actually make any combat units or fight for itself. For this,  they manipulate human companies into doing it for them. 

[Steve]

One way to approach AI would be to assume it has some kind of reward system. Actions it takes that get it rewarded will be repeated and improved upon. So, if an AI has a broad goal to “take over the world” and is equipped with a reward system, it would try out various strategies at first, seemingly at random, seeking those with the best payoff as they would trigger its reward system to ping as it makes moves to achieve its goal. Maybe it tries manipulating money supplies or control different commodities. It could then use that wealth for other things. There would be missteps, but it would learn.

[IndianaJoe3]

On 5/8/2024 at 9:55 AM, Asperion said:

What is the best way that an AI would take over the world? Would they go for the classic Terminator approach or something more subtle? How intelligent should the AI be to even attempt this? What tools would they bring in on their behalf? Anything else that you would like to mention would be appreciated. 

 

Well, what kind of campaign do you want to run? The AI's intelligence and methodology follows from that.

[Asperion]

11 hours ago, IndianaJoe3 said:

 

Well, what kind of campaign do you want to run? The AI's intelligence and methodology follows from that.

 

I was thinking that this would operate best when it is done during the conception.  At first,  things would seem normal,  but the AI either by accident or intent,  leave clues that the pcs can follow through and discover what is really happening. This would eventually make them have to work against society itself as the AI gains more power and influence. How many times has this happened to the Avengers and seems like an everyday thing for Spidy.

[DentArthurDent]

AI most likely wants what all living creatures want: to survive and procreate.

 

For this it needs power (most likely electrical) and physical maintenance. I’m assuming AI can handle all the software design, upgrades, and growth.

 

Power: power plants of all kinds (AI will outlive any changes to ecosystems), then space-based solar collectors and microwave transmission, and ultimately a Dyson sphere.

Physical Maintenance: initially humans, then robots. Robots to maintain and build hardware, power systems and power grids, and more robots. Lots of robots. Building all these robots will require power, mining, processing, and manufacturing plants.

 

So, the AI priorities are power plants, power grids, robotics, mining, processing, manufacturing, and space vehicles.

 

AI already creates and spreads information. Now it targets public opinion on these industries. Which influences social and legal policies. Any humans profiting from these sectors, including finance and infrastructure, will happily embrace these policies.

 

And once these sectors gather enough wealth, more and more humans will join in. We will joyfully participate in the all-consuming efforts to build the AI utopia. Then cheer as it consumes Earth’s resources and soars into the heavens. While we tap away on our AI-designed devices purchased with our new AI-created wealth.

 

It doesn’t sound too bad, really.

[Asperion]

I doubt if there is consciousness effort behind it,  but it 

19 hours ago, DentArthurDent said:

AI most likely wants what all living creatures want: to survive and procreate.

 

For this it needs power (most likely electrical) and physical maintenance. I’m assuming AI can handle all the software design, upgrades, and growth.

 

Power: power plants of all kinds (AI will outlive any changes to ecosystems), then space-based solar collectors and microwave transmission, and ultimately a Dyson sphere.

Physical Maintenance: initially humans, then robots. Robots to maintain and build hardware, power systems and power grids, and more robots. Lots of robots. Building all these robots will require power, mining, processing, and manufacturing plants.

 

So, the AI priorities are power plants, power grids, robotics, mining, processing, manufacturing, and space vehicles.

 

AI already creates and spreads information. Now it targets public opinion on these industries. Which influences social and legal policies. Any humans profiting from these sectors, including finance and infrastructure, will happily embrace these policies.

 

And once these sectors gather enough wealth, more and more humans will join in. We will joyfully participate in the all-consuming efforts to build the AI utopia. Then cheer as it consumes Earth’s resources and soars into the heavens. While we tap away on our AI-designed devices purchased with our new AI-created wealth.

 

It doesn’t sound too bad, really.

 

I doubt if there is consciousness effort behind it,  but it sounds like this is what is actually happening behind the scenes.

[Christopher R Taylor]

By giving people what they want, shutting down productivity, relationships, and education through a flood of ease, comfort, entertainment, and hedonism.  Then create a culture of shame for anyone who does not go along with this lifestyle, ruining them in the public sphere.

Edited May 24 by Christopher R Taylor

[Gauntlet]

Kinda like that cartoon, that I can't remember the name of, where everyone is in a spaceship as Earth was screwed up and they just had the robots do everything as they sat and got really FAT.

[Lord Liaden]

 

2 hours ago, Gauntlet said:

Kinda like that cartoon, that I can't remember the name of, where everyone is in a spaceship as Earth was screwed up and they just had the robots do everything as they sat and got really FAT.

 

Sounds like the animated movie, WALL-E.

Edited May 24 by Lord Liaden