RoS: is defeating a single zone-mind always a Pyrrhic victory?

[Johan Larson]

In Reign of Steel, one strategy for the human resistance is defeating a single zone-mind and then somehow quickly consolidating their position as a human-run zone in a world of AIs.

Is that a viable strategy, or does the fall of any zone-mind just trigger a massive retaliation by all the other AIs?

Perhaps it would work best if the fallen zone were one of the real troublemakers, like Mexico City or Zaire.

[Rolando]

I think that is a very chancy bargain at best for the human resistance.

Most AIs will fear the capabilities of a human resistance able to eliminate another AI, even if it is a troublemaker. And fear is the main aggression builder and fear is what made Manila turn against humanity in the first place.

Many AI's will vote for a consolidated assault against the successful humans (Manila, Berlin, Mexico, Zaire will be the most militant).

I think that even Caracas may betray humas that kill Mexico, just to stay cool with the rest of AIs.

Serenity may be the only hope of support, but I doubt it will be enough.


The best alternative for humas it to somehow subvert an AI or kill it and make as if it is still alive somehow... doing something similar to London humans may fake it good enough to devise some survival strategy. Or faking a human/AI society similar to Washington.

But an outright killing will only bring the AIs together again, and give more political/moral power to Manila, Mexico, Zaire and Berlin.

[Johan Larson]

That sounds right.

OK, so if the humans are going to destroy a zone-mind, they have to destroy it quickly and completely, including the infrastructure. The other zones will mount a counterattack, and will try to reclaim the territory, presumably by giving the territory to the nearby zones. The quicker the humans get out, the less they are exposed to the counterattack, and the more the infrastructure is destroyed[*], the longer it will take for the new owners to set up shop again.

I suppose the nearby zones could come to blows over how to divide up the new territory. That might be a hook for the resistance. Step 1: destroy a zone-mind. Step 2: encourage the claimants to the territory to fight among themselves.

The other zones might also decide to awaken a new AI to replace the old one, rather than dividing up the territory.
[*] I picture vast armies of dumb-bots running lemming-like into the sea.

[Boomerang]

The idea that the humans could destroy an AI on their own would seem ridiculous to the AIs. The AIs would believe that another AI or group of AIs was ultimately behind the attack, perhaps to the extent that they may draw links between humans and AIs where none exist and provoke a civil war between the two main AI factions (awakened vs expansionist). The awakened essentially accusing the expansionists of being in an alliance with the humans.

[Johan Larson]

Offhand, if the strategy is to provoke a fight between AIs, which two would be the best targets? Vancouver and Moscow, maybe, who may already be in conflict over Siberia?

[tshiggins]

The smart move is not to destroy the AI, but to co-opt it.

The introduction of a virus that kills its sentience, while keeping it otherwise intact and allowing the humans to control it -- that's the smart move.

I'd always assumed that if the humans destroyed only one AI, they'd face massive retaliation from the others -- enough to wipe them out.

At that point, the AIs would start to squabble over which of them should gain control of the territory, and whether or not to create a new Zonemind to replace the previous one.

[Anaraxes]

Originally Posted by tshiggins

The introduction of a virus that kills {a Zonemind's} sentience, while keeping it otherwise intact and allowing the humans to control it -- that's the smart move.

I'd always assumed that if the humans destroyed only one AI, they'd face massive retaliation from the others -- enough to wipe them out.

Aren't the other AIs going to just as easily notice that an AI's sentience has been killed, even though its autonomic reflexes keep working? Even if you replace the sentience with another, the change of personality would be noticeable -- just as if some similar crude mind control happened to a human.

You'd really need to make some incredibly subtle changes that cause the target AI to project a false front and actively deceive the other AIs (including the appropriate skills to do so, of course). Do RoS humans even have that level of programming expertise? (And is running a many-year software development project with a team of hundreds a good adventure game activity? The premise is going to have to be a campaign plot point where the replacement AI is a freebie from the GM. Maybe the campaign even starts after the substitution has happened, though it could still be fun to play out the events of that initial substitution.)

[Johan Larson]

The GURPS Software Development Project source-book was not a big seller, as I recall...

As an adventure campaign, subverting an AI might have the following parts:
1. Locate the communication line by which the AI periodically uploads its current state to its backup location. (Wilderness adventure)
2. Tap the line and copy one of the state snapshots. (Hardware/Software hacking. Possibly locating construction machinery, too, if the line is buried.)
3. Locate one of the original designers of the AI software, who just might be able to tweak it to make it friendly to humans. (Post-apocalyptic detective work.)
4. Help the original designer modify the snapshot as required. (Software development.)
5. Infiltrate the backup location and install the modified snapshot, designated as the target snapshot for fail-over. (Rogue/heist work)
6. Attack the main AI complex and cause enough damage that the AI will do a full self-validation, causing it to switch to the saved snapshot (which will be the altered snapshot). Suppose power-loss would be sufficient. So knock out the main generator momentarily. (Military assault)

That's a campaign, right there.

[fchase8]

Reign of Steel proposed that a "buffer state" that is human-run might be acceptable to some AIs. This would mean that humans wouldn't control all of the former AI's territory, and that those fighting over said territory would have to be enemies.

Unfortunately, I don't know what zonemind would fit that. They actually have few out-and-out borders - for instance, if Mexico City fell, Denver & Vancouver (to relative friends) would take most of the territory, with Caracas picking up just some of Central America & Washington maybe Caribbean.


I agree with the idea that AIs would assume any human offensive/takeover was assisted by another AI. They would have blinders kind of like Centrum in Infinite Worlds.

This could actually spell bad news for any potential human state - even if the neighboring AI was okay with it existing, other AIs would assume it would be letting the human state exist because it was controlled by that AI. So better to topple it and say it had never been supported (however unconvincing that may be).


Note that taking over the moon or Orbital might not fit in this - if humans could somehow win there, it would be hard for the other AIs to invade (except for Orbital invading the moon). But the AIs would cut off human-run space from trade, which would easily strangle it.

[patchwork]

That's a very good outline, Johan. I always thought there were two obvious campaigns from the original setup, a North American one and a European one.

The European one involves getting a couple of nukes or similarly highly destructive devices, playing with the guy who knows where Paris keeps both its primary and backup, and decapitating Zone Paris. The Zoneminds will initially suspect each other rather than humanity, but Paris' territory will be divided up - Iberia to London, France to Berlin and its North African holdings to Zaire. This is a net win for humanity, as it puts more territory under disengaged London and more under environmental steward Berlin. Zaire's no good, but since it's paranoid and impulsive even by Zonemind standards, and it directly benefited, it is the most logical suspect for Paris' murder...so you manipulate the other Zoneminds into taking it out. Alternatively, you can take it out yourself with a third nuke once you know exactly where it is, it has no backup, but that's tipping humanity's hand in exchange for a head start at seizing and repurposing its assets.

The North American campaign starts with a probably-friendly AI, Lucifer, in need of a powerbase. Presumably Lucifer CAN fool other Zoneminds into thinking it's whoever it replaced (or a linear upgrade of same), at least long enough. So you start with a choice of Denver or Washington (I don't see Vancouver or Mexico City as very practical options). Washington builds a strong human-AI alliance almost immediately, after which it can liberate Denver's camps, reveal Denver's abominations (and weakness) to the other Zoneminds, and Denver goes down under a dogpile with Lucifer-Washington gaining ground. In a perfect world, Vancouver commits significant forces to the Dogpile on Denver, and Moscow stabs it in the back in Siberia. Next step is to wipe Vandenburg off the map - without its groundside facilities, Orbital has to become utterly dependent on either New Delhi or Beijing, the two with independent launch facilities - and Beijing doesn't like Orbital. Forcing Orbital into a partnership with New Delhi upsets the AI balance of power badly - New Delhi can now get accurate satellite data and communications and if it directs Orbital to deceive the others, Orbital pretty much has to go along. Since that state of affairs is unacceptable to the Zoneminds...I think they gang up on New Delhi even if it DOESN'T immediately exploit its new position. A diplomatic solution would be possible, but it would involve either Beijing or New Delhi ceding its launch facility to Orbital - which I don't see happening. So the ZMs lose their space advantage.

In a perfect world, these things happen at once. If Lucifer-Washington and Tranquility become aware of each other, they can partner easily. And now humanity has reliable space infrastructure and the AIs don't. Lucifer-Washington's biggest problem is Mexico City on its doorstep, but if it has Tranquility on its side, orbital surveillance, mass drivers and awakening new AI based on Lucifer (let's call it Azazel) to seize MC's assets before other ZM reps are on site. At this point, Lucifer probably can't conceal its nature any longer, but Paris, Zaire, Washington, Denver and Mexico City are offline, space is on humanity's side and the surviving AIs are fighting each other in Asia. Caracas could perhaps be flipped if promised the opportunity to restore the biomes of Mexico and Africa...

I see we've gotten away from the initial question. My apologies. I think no, defeating Paris or Washington could be a substantial gain even if it stopped there.