[Reign of Steel] AI civil War

[Boomerang]

I love the underground Paranoia style computer idea.

[johndallman]

It's possible to reskin Zone Washington as Friend Computer. The territory of the Zone is not as enclosed as Alpha Complex, but the same idea that the outside is terribly dangerous applies.

[Johnny1A.2]

Another way to get an AI civil war: awaken an AI backup complex.

Most of the AIs have 'backup' systems, AI mainframes constantly updated with copies of the original AI's memories, programs, and data, but kept on standby. They're only supposed to run the programs if the original is destroyed or falls out of communication.

(Apparently the AIs fall into the 'the copy is the original' school of thought on the doom thread.)

Anyway, if some force causes a backup complex to run the program when it shouldn't, suddenly there are two Overminds, z-Tokyos, or z-Berlins, or whatever. If the original zonemind didn't have precautions in place for this eventuality, it's a good question which AI the AUs and NUs will obey. If it does have precautions in place (which I would expect with the majority of them), you'd need for the interfering force to bypass that somehow so the second zonemind would have access to some resources.

But do that, and BOOM! AI wars. If you could arrange for it to happen to more than one zonemind at a time, chaos could emerge.

(z-Zaire actually nuked its own backup complex because it thought something had gotten to it. Imagine a situation where it was right.)

[fchase8]

Originally Posted by Johnny1A.2

Another way to get an AI civil war: awaken an AI backup complex.

Most of the AIs have 'backup' systems, AI mainframes constantly updated with copies of the original AI's memories, programs, and data, but kept on standby. They're only supposed to run the programs if the original is destroyed or falls out of communication.

(Apparently the AIs fall into the 'the copy is the original' school of thought on the doom thread.)

Anyway, if some force causes a backup complex to run the program when it shouldn't, suddenly there are two Overminds, z-Tokyos, or z-Berlins, or whatever. If the original zonemind didn't have precautions in place for this eventuality, it's a good question which AI the AUs and NUs will obey. If it does have precautions in place (which I would expect with the majority of them), you'd need for the interfering force to bypass that somehow so the second zonemind would have access to some resources.

But do that, and BOOM! AI wars. If you could arrange for it to happen to more than one zonemind at a time, chaos could emerge.

(z-Zaire actually nuked its own backup complex because it thought something had gotten to it. Imagine a situation where it was right.)

That's a great idea!

It would be tough to explain how all back-ups got activated at the same time, since they're all so isolated. Perhaps it's a long-term campaign, activating back-up AIs one-by-one.

It could start in Paris, where the rebel leader knows where the back-up AI is (though he wants to nuke it & the main AI). Or in Tokyo, learning from the human-started AI civil war there already going on.

And end in Zaire, where there's no AI to back-up. Or end in Luna - does Luna even have a back-up AI?

One question would be whether human-activated back-ups would be pro-human, or just as anti-human as other AIs. If the latter case, maybe it wouldn't be a good idea to activate the back-ups in Washington and/or London. Maybe the whole idea actually originates from Washington or London...

[Johnny1A.2]

Originally Posted by fchase8

That's a great idea!

It would be tough to explain how all back-ups got activated at the same time, since they're all so isolated. Perhaps it's a long-term campaign, activating back-up AIs one-by-one.

What difference does it make?

In fact, it's almost certain better, from a human POV, if all happens at once. If it happens in just one zone, you get a civil war in that zone. The other z-minds might well ascribe it to a local screw-up of some kind, or assume it was engineered by a rival z-mind. (Caracas, for ex, would probably love to pull something like that with z-Mexico City.) The net result of a 'just 1' incident might be that a neighbor takes over the zone.

If it happens in 2 or 3 at once, the z-minds will realize something is up that affects everyone, but most of them will still be free to act as a body.

But if it's happening to a majority of them all at once, they'll have their hands full either way. It doesn't much matter if the event is arranged by humans or happens naturally, if it happens to all the zones at once, you still have civil war in most of the zones, chaos, uproar, danger and opportunity both for humans.

It could start in Paris, where the rebel leader knows where the back-up AI is (though he wants to nuke it & the main AI). Or in Tokyo, learning from the human-started AI civil war there already going on.

And end in Zaire, where there's no AI to back-up. Or end in Luna - does Luna even have a back-up AI?

Not canonically. Luna doesn't actually have much at all, it isn't much more than a single base. It doesn't even have a robofac. That's why Luna knows it's doomed unless it can find some additional resources and/or money.

One question would be whether human-activated back-ups would be pro-human, or just as anti-human as other AIs.

The backups would be just as anti-human as the originals, at first. If you can access their programming enough to change that, then you probably don't need to do it at all. I was thinking that it might be possible to trick the backup into running when it wasn't supposed to, forcing a z-mind to fight itself, more or less.

Actually, I could imagine 2 Overminds getting along all right, at least at first. The others, not so much.

[Minuteman37]

I'm just going to layout my thoughts for how the humans could trigger an AI civil war with a pro human faction.

This assumes the following

• Lucifer is a legitimately good pro human AI with plans to assassinate Washington and covertly take his place with the help of Free America.
• London is secretly pro-human and left a backdoor in Brisbane for "stuff".
• Moscow doesn't really hate humans and could be persuaded to side with the pro human faction if he thought they actually had a chance and he could gain something from it.
• Caracas knowing the negative impact his AI brothers are having on the planet's environment would be open to siding with the humans if he'd gain a large portion of additional territory for environmental protection.
• Giving Caracas the former territory of Mexico city, Zaire, Brisbane, Manila and Antarctica is a very compelling offer to Caracas
• Tranquility and the colonists can succeed in attacking Luna and merging the mainframes to put the friendly Luna/Tranquility AI on even footing with the other AIs when it comes to processing power.
• This AI can use the additional resources from Tranquility base and what they can obtain from Orbital to build a mass driver.
• Astronauts from tranquility can infiltrate Orbital and upload a new virus to overwrite Overmind's original one.

Anything this plan is missing?

[Tallor]

*adjusts glasses* I'm gonna need a few spare hours to catch up with this. Glad I'm not the only Steel enthusiast! :P

[fchase8]

Originally Posted by Johnny1A.2

What difference does it make?

In fact, it's almost certain better, from a human POV, if all happens at once. If it happens in just one zone, you get a civil war in that zone. The other z-minds might well ascribe it to a local screw-up of some kind, or assume it was engineered by a rival z-mind. (Caracas, for ex, would probably love to pull something like that with z-Mexico City.) The net result of a 'just 1' incident might be that a neighbor takes over the zone.

If it happens in 2 or 3 at once, the z-minds will realize something is up that affects everyone, but most of them will still be free to act as a body.

But if it's happening to a majority of them all at once, they'll have their hands full either way. It doesn't much matter if the event is arranged by humans or happens naturally, if it happens to all the zones at once, you still have civil war in most of the zones, chaos, uproar, danger and opportunity both for humans.

It matters from a 'what makes more sense?' point of view. It's just harder to explain why all the back-ups would go online at the same time.

I agree, from humanity's perspective, all going on at once would be better than one at a time. Though if it's a one-at-a-time orchestrated by humans, they might start with the most anti-human AIs (Overmind, Mexico City), and only end - or not do at all - with the most pro-human (Washington, London).

Perhaps, once it's started in some anti-human zones, the less anti-human AIs would make a deal with humans to treat them better, in return for not activating the back-up.

Originally Posted by Johnny1A.2

Not canonically. Luna doesn't actually have much at all, it isn't much more than a single base. It doesn't even have a robofac. That's why Luna knows it's doomed unless it can find some additional resources and/or money.

Luna would be an afterthought in this kind of war. Though the weak AI might enjoy seeing all of the more powerful AIs crippled by civil war (as would Zaire).

Originally Posted by Johnny1A.2

The backups would be just as anti-human as the originals, at first. If you can access their programming enough to change that, then you probably don't need to do it at all. I was thinking that it might be possible to trick the backup into running when it wasn't supposed to, forcing a z-mind to fight itself, more or less.

Actually, I could imagine 2 Overminds getting along all right, at least at first. The others, not so much.

True - if you could make a pro-human AI, then you're already way ahead of the game.

But activated back-up AIs might side with humanity simply because they need more support - particularly the back-ups of anti-human AIs. Whereas an activated back-up Washington might be very anti-human, since the humans would be on the original Washington's side.

It's a question as to whether the AIs are naturally anti-human, or are just that way because they saw humanity as their greatest threat. Given a bigger threat, would the AIs work with humanity?

[Jasonft]

One way not really touched on yet to start the war is if the aliens start giving serious help to the AI that is their ally. Depending on how much hardware can be brought in covertly this could be quite a substantial change to the strategic balance.

Assume for the moment two things - 1) The aliens don't want the AIs to start looking outward for threats, and 2) The aliens want the biosphere of Earth more or less intact [or at least recoverable].

The allied AI suddenly gaining lots more capacity and maybe even a higher TL without the other AIs finding out about it gives you all *kinds* of options for destabilizing without things going nuclear. Even better is how the AIs being focused on each other trying to figure out who the traitor is means even less chance of spotting the aliens.

[fchase8]

I like the idea of the aliens wanting to preserve Earth's biosphere - it not only gives them a reason to support humanity, but it's not just a generic 'pro-human' reason.

Plus, it gives them reason to differentiate between the AIs, and not on their human policies, but their environmental policies. So the aliens might be against anti-human Mexico City, but for anti-human Berlin.

And this would mean that humans would be wary of their new alien allies - what do the aliens want to do with Earth once the AIs are defeated?...