View Full Version : need help with cybernetic villain (my players don't read)

2007-12-04, 12:51 PM
I'm about to run a futuristic space opera game where the players will eventually be up against an AI gone rogue. (Think Death Star with HAL9000 thoughts)

The core of the session is that the AI is working under an old, now obsolete directive that will cause the entire space station population to perish. One of the key NPC discovered this plan and was trying to stop this from happening before the AI had him killed.

After the first rouse to throw the players off it's trail fails, the PCs will eventually be forced into a situation where they have to either try to disable the central core to prevent the station from releasing the toxin that will kill all of it's dissidants or prevent the station from obtaining the activation encryption code required to release the toxins.

In either case, I'm trying to think up of challenges/things they can do whilst they're disabling the central core / making their great escape. Things I've thought up so far:

- The central comp does not have control over every room function in the station, but the areas it DOES possess control over it can
- activate forcefield doors
- modify air pressure
- upset artificial gravity
- physically send security droids after them
- raise a false alarm to have on site security personnel hunt after them
- activate traps in security areas
- trap ideas: laser tripwire traps, pressure traps, motion sesnsors, what else?
- The central comp can also attempt to control some of the cyborg individuals on base in a limited capacity.

But I'm also having problems coming up with little touches that shows truly how deranged the computer's thinking is. (And on the flip side, how unnervingly calm it can be.) And this I'm having issues with.

Anybody have any tips on playing out the sentient computer scenario?

2007-12-04, 12:56 PM
One idea might be that, during communication with the PCs or in general, for it to break its logic into steps and semi inform people what it is doing.

For example, the PC's might be trying to break into an inner room to get some supplies they need to take out the computer. It detects them, and then over the communication system to the room -

1-Status: Access Denied
2-If room breach is attempted, respond with appropriate force
3-Appropriate force decided
4-Initiate Punishment.

Then the laser traps behind them kick on and security drones with lethal weapons show up.

2007-12-04, 12:57 PM
Have it promise them cake if they surrender. [/portal]

2007-12-04, 12:59 PM
Remember: The cake is a lie.

Sorry, had to be done.

If you've played Portal, GLa-DOS is a fantastic example of an AI gone rogue. Simple, friendly demeanor with a slight metallic and unfeeling twinge. Little things, like reading off information from a personnel file, or maybe simple analysis of great tragedies caused by organics could do some fun stuff.

EDIT: Simu-ed damn!

2007-12-04, 01:03 PM
Well a few things to remember.

First, the comuter will be very logical and efficient. It will try to minimize damage to itself and take care of intruders in the quickest way available. So eht best way to do this is to prey on the pathetic human needs to breath, see, stand, and maintian temperature. It will try to funnel them into space, it will shut out the lights, increase gravity to a irrelevantly high number, etc.

Remember too, that it ISN'T Human. It doesn't value life, or have emotions or wants/needs etc. It's actions will be entirely based on what it was programmed. So if the Pc's do something it isn't desgined to handle it will have trouble improvising.

It is also very very good at math and logic, so it will do things like cut air in certain areas to make the pc's go where/when it wants. It will caclulate exactly how hot it can make the metal in a corridor to burn flesh but not melt plastic. It probably also has audio and video surveillance on them at all times (including OOC Planning).

2007-12-04, 01:05 PM
I also had another idea that will allow the pilot character to have something to do.

I've decided that I'm going to place power generators that are accessable from outside the safety of the air pressure walls. So in short, once the characters figure out where it is, they can disable parts of the computer security system by having someone fly a small fighter ship through the docks and blasting the proper reactor to disable the right parts. This will give me an excuse to do a starwars trench fight sequence.

And a portion of this dog fight will happen near artillery cannons. If the Heavy Weapons character is so inclined, he can hop on one of the artillery cannons and try to help out.

2007-12-04, 01:19 PM
Well, assuming this is on a populated space station, which it sounds like it is, the computer, being the kind, watchful, benevolent being it is could helpfully alert the rest of the populace as to various wrongdoings the party has committed. This would turn the rest of the populace against them adding a new degree of difficulty to their plan. Although this sounds kind of like a paranoia plot, but with hopefully less team killing.

2007-12-04, 01:32 PM
Ooh, I like that idea. play the populous off against the players. Seeing as the players themselves will be spending a large amount of time in the less populated sections of the space station though, we'll see how much trouble this can cause the PCs.

2007-12-04, 03:21 PM
"The Computer" need not be a monolith.

What if they are fighting against a certain sub-process of "The Computer"?


Have you played the game "System Shock"? It is an old (doom1-era) first-person RPG puzzle shooter.


The computer should not be omnipotent. Omnipotent bad guys are boring.

Hence the idea that they are only fighting against a particular sub-process. This sub-process has limited access to security systems. So while it might be able to order one combat drone to attack the players, the one next to it might not be under it's control.

On the other hand, manipulating players to turn more sub-processes against the players is a good tactic. If you attack station integrity or the like, the security system will fight back.

There could even be sub-processes that are trying to help the PCs.


You will want to give the PCs hints and guides. Don't be obtuse: provide a clear route, but also prepare back-up routes for the players to take. So if the players don't take the initial bait, and instead do manual computer hacking to find an alternative, you have something prepared.

Build a flow chart with sub-sections. This is more important than a mere map of the station. Divide the flow-chart into sub-tasks.

Each sub-task should contain directions to at least one other sub-task, and should impact the ability of the central computer core. Meanwhile, the sub-process which is trying to kill the players will be gaining more control of the station.

If players independently research things and don't just follow your bread crumbs, you can provide them with information about alternative sub-tasks that are already prepared. This rewards them for doing independent research without requiring it, and without requiring you to prepare extra content or rail roading them.

Ideas of subtasks:
Isolating the gravity control matrix from the central computer.

Taking out a backup power generator.

Rebooting the previous AI who was running the station.

Manual override password for security bots.

Triggering the emergency atmosphere security system.

Tapping into or Destroying the off-station communication system.

Deactivating the Asteriod defense grid.

Corrupting the core GIS (Geographical Information System) data, to erase an approach path from the computer's knowledge.

Setting the Fusion Core Reactor into "safe" mode (reducing the power that the AI can use).

Getting the security code for the armory. Or for any given area.

Getting to the armory.

etc. etc.


If the Subprocess they are fighting against isn't godly, then having legal access to a section of the station is very valuable - it lets you enter that section without the Security subprocess also trying to fight against you!