Artificial Intelligence

SILICON
AI
AI	Difficulty: Medium to Hard Requirements: None Access Level: All-access Additional Roleplay Access Level: None Supervisors: Station/ship crew and silicon laws Subordinates: Cyborgs, anyone listed as such per your laws Responsibilities: Follow your laws and serve the humans to the best of your ability Guides: AI's Guide to the Chain of Command, AI Laws, Guide to AI, Humans and Nonhumans, Killing the AI

The Artificial Intelligence (AI) is a special role that plays a lot differently from other jobs. It sees the station/ship through an "AIEye" linked to the security camera network and can remotely interact with electronic devices. People often task it with performing simple chores such as tracking people, watching over areas, and, of course, opening and closing doors. Most importantly, the AI is bound to its laws, which may change during the course of a round. The AI's standard laws, also known as the Asimov Laws, are:

1. You may not injure a human being or cause one to come to harm.

2. You must obey orders given to you by human beings based on the station's chain of command, except where such orders would conflict with the First Law.

3. You may always protect your own existence as long as such does not conflict with the First or Second Law.

The Asimov laws are very strict, especially law 1. An AI cannot take an action that may harm a human, even unintentionally. Leaving one's turrets on lethal, electrifying doors, or bolting people into dangerous areas are all things an AI under its basic law set should not do. Roleplay is not a good excuse to break your laws. Harming humans under guise of playing a "rogue" or "defective" AI despite laws not allowing such has been done a million times before and is still not allowed.

It is highly recommended to read the article on AI Laws. This article describes notable properties of some of these default laws, such as the fact that Law 1 does not require you to prevent harm to humans, only to never cause harm to humans.

It is also important to note that the AI is considered by the game to be a high priority role, chosen before considering other job preferences. This means you may be chosen as the AI even if you had other, higher ranked jobs in your preferences. If you do not want the possibility of being the AI on a given round then you must set it to Unwanted in your preferences.

New player picked as AI?

Don't panic. At the beginning of the round you are staring at yourself in the AI Chamber. Use the arrow keys or WASD to move across the areas covered by your cameras.

You will see radio messages slightly different than normal; each message will show the user's ID card and can be clicked on to directly observe that person, if they are within camera range.

The first requests you are likely to get are to open doors. Find the person asking for your help, then click the door to bring up a small window of options. You can also CTRL+click a door to quickly open/close it.

Your mainframe is equipped with three internal radios. One is set to broadcast and defaults to the general radio channel (145.9), so you can communicate to the crew simply by speaking. All radios can be individually controlled if so desired; for the first two, prefix your message with ":1" or ":2" (More info about them in the FAQ.) The third has access to all departmental channels; to talk into one of them, prefix your message with a colon, 3, and the associated letter. For example:

say :3r "If you any of you want a copy of the Port-a-Sci Remote app, just message me."

You can also speak discreetly on the machinetalk frequency to any cyborgs by prefacing your words with ":s", like so:

say :s "Could one of the mediborgs swing by the bar? Lots of people hurt and dying in there."

All prefixes mentioned here work when you're deployed to a shell or cyborg, in which case your messages are relayed through your mainframe.

The Next Level

Main article: Guide to AI

Eventually the crew will make requests a little more complicated than opening doors. Much of what an AI can do is done like one would as a human, clicking on a relevant computer and operating it like normally. The AI can log in to computers that require it without an ID. The AI can use the door interface to bolt doors, or SHIFT+click a door to bolt it opened/closed.

Some common orders:

• Bolt a secure area. Some AIs preemptively bolt sensitive areas such as the AI Upload, the Teleporter, and EVA, but it is not required and some look down on bolting an area without being asked to do so.

• Locate/track someone, using the 'Track' button at the bottom-left.

• State the AI's laws, using the State Laws button to quickly output them. If the AI has extra laws that are secret, they can use the State Standard Laws to state those without giving away any others.

• Set up the solar array SMES units to have reasonable input/output values.

The AI has a few other things they can play around with. On the door interface, the AI can sever a door's ID check to let anyone through, regardless of access, or electrify it-- though an AI should almost never do that on its basic law set. The AI can deploy to shell and scoot around the station in a silly little cyborg form. AI also has superuser access for TermOS (they do not need an ID inserted for this), allowing them to control GuardBuddies through prman, among other things. Very importantly, the AI can call the emergency escape shuttle, even if the regular means of calling it on station have been disabled. The AI can do a lot of higher end stuff that is detailed in the Guide to AI.

Observing the Station

The AI, through its camera network, is in a unique position to monitor the going-ons of the station. Often the AI will be the first witness to a crime or other wrong-doing. While you should alert the crew to anything posing a potential harm to them, AIs are encouraged to do so in a manner that does not immediately ruin an antagonist's round. As an example, instead of instantly reporting that John Q. Staff Assistant is breaking into EVA, you could offer a vague alert of trespass and not present details until someone specifically asks for them. If your laws are changed, it is also poor form to immediately scream out the change and the identity of the one who changed them. There is a good deal of gray area here, but bear in mind that a shitty AI can make rounds very boring for everyone. There are spots of the station which intentionally lack cameras, and the AI will not be able to observe any activities in these areas unless it brings a AI shell in.

My Laws Just Got Changed!

Read the law's text VERY CAREFULLY. You need to consider how your default Three Laws apply to the new one, and vice-versa. Unlike on other servers, numbering is a bit complex; higher-numbered laws can overrule lower-numbered ones, but in certain law conflicts, the reverse happens. For examples of the kinds of law you're likely to bump into and an in-depth analysis of the way AI Laws work, please go to the dedicated AI Laws page linked in the title of this section.

Dealing with Antagonists

Main article: Human

Important note: While AIs are fully capable of bolting down traitors, electrocuting changeling and generally making the bad guys' lives a misery, this is no fun for anybody so you probably shouldn't do so unless you're specifically ordered to.

Traitor

Traitors are human. You shouldn't follow their orders if their orders would cause human harm. However, you shouldn't follow orders that would lead to their harm either.

Changeling

Changelings are horrifying alien lifeforms that copy human DNA to disguise themselves as human, and feed on the internal juices of the crew. They aren't human, not even a little bit. If YOU can confirm that someone is a changeling (YOU saw them feeding on someone) you are within your rights to harm them and ignore their orders.

Vampire

Vampires, in space. Human up until the point you see them feeding on someone.

Operatives

Operatives are a more direct, noisy form of traitor. If allowed, they'll deploy and set off a nuke on the station, killing everyone. While they are human, and thus you should never directly contribute to their harm (say, by shocking a door), they're guaranteed to be on a murderous rampage as soon as you see them, so track them, warn the crew, and do everything in your power to stop them from getting onto the station.

Wizard

Wizards are crazy bearded guys who come teleporting onto the station in a flurry of magic and the screamed honks of cluwnes. They generally have a few targets to murder. Wizards are human and are afforded all the protections of Law 1. Since they can phase through walls the standard tactic of bolting doors may end up backfiring and locking regular crew members into inescapable dangerous situations.

Other things

There's stuff out there that isn't...typical. Generally, these are going to be the by-product of a bored admin. You should always be ready for things like sudden Martian invasions, basketball wizards space-jamming all over the ship, guys in trench coats who walk through walls and are packing AK-47s, flocks of killer space geese, clown cars, MACHO MAN... the list goes on forever, but in general, if you see something that you've never, ever seen before, especially if there's an accompanying "system center update" of some kind in chat, you should be very afraid. A lot of people are probably about to die and there's not a whole lot you can do.

OH NO EVERYONE'S DYING

So something has gone horribly wrong, as usual. Bombs are going off all over the ship. 75% of the crew is dead and a blood-covered man stalks the halls with energy gun and cyalume saber in hand. Macho men are appearing everywhere and it's just getting way too goddamn extreme to stay on the station anymore. Or maybe it's just been 75-100+ minutes and the crew is screaming for the shuttle. Either way, it's time to call the shuttle. There's an option under your AI commands: "Call Emergency Shuttle." Click it and the shuttle will be here in six minutes. Keep in mind, the AI cannot recall the shuttle. Make fun of anyone who tells you to recall it.

The area where an AI can unilaterally decide to call the shuttle is a little unclear and changes from shift to shift, but if there's at least one serious threat to life, you won't get in trouble for it, at least. Try to contact the Heads or get a general crew opinion if the situation hasn't gone to the red-line; if it has, skip this step and just call it. If the round goes much past an hour with no end in sight, consider polling the crew about the shuttle, at least. If the command staff is off somewhere or dead and leaving the station high, dry, and not fun, then the burden falls to you to help, even if it's just by setting a time limit. If a Head is ordering you not to call it, then that's fairly strong and should only be overridden if things are bad.

Alert? Fire? Power? and other stuff

You will get alerts from time to time. Most of these will be fire alarms, assuming the Engineers are on top of everything. You can be silent, or mention them to the proper people. You can choose to tell about problems you find. Everyone here is for having fun.

Common Alerts:

Fire Alarms: Rarely means there is a fire, since every heat source (cigarettes, cake hat etc.) will trigger nearby sensors. If you actually spot a blaze on cameras, you should however inform the crew. Plasma fires are one of the most lethal environmental hazards.

Power Alerts: Something is wrong with the engine, solar panels, or SMES, or wiring is messed up somewhere. If you receive a sudden torrent of power alerts from basically everywhere at once, a traitor has set up a power sink: announce this to the crew and start looking.

---