/tg/Station 13

"Nanotrasen shareholders were delighted to hear Nanotrasens recent announcement of a new project in the field of artificial intelligence. Although details are scant, representatives present at the latest press conference claim the new project will increase the efficiency of both new and existing AIs. NT has confirmed that closed beta testing of the program will begin on select research stations as soon as the next fiscal quarter..."

Just like cyborgs, Subroutines sole purpose is to assist the AI in any way possible. However, unlike cyborgs who assist the AI in interacting with the physical world, Subroutines are designed to assist AIs in the digital world. In real life, a sufficiently potent AI could be "everywhere at once" or in other words pay perfect attention to multiple tasks at a time. Since AIs in SS13 are played by people, that's not really possible and it often puts a lot of stress on AI players. This job helps remedy that.

How are Subroutines made?

Subroutines are pulled from the ghost pool from players who have it set in their preferences, exactly like PAIs. They are created from a new machine located in the R&D Server Room, known as the Bluespace Intelligence Generator, or B.I.G. and stored in data clusters also located in the Server Room. If the B.I.G. loses power or is destroyed, no new Subroutines can be made until it is repaired. If the data cluster holding the Subroutine is destroyed, the Subroutine is instantly killed. The B.I.G. will automatically generate a new Subroutine (selected from ghosted players) every 10-15ish minutes so long as there is an empty data cluster to store it in (Two start in the server room but more can be constructed).

What can a Subroutine do?

Subroutines act as "AI-Lite"s. Their sole purpose is to act as an extension of the AI players will. They have most of the same abilities as an AI, but also have many limitations.

When a Subroutine is made, it is immediately slaved to the AI with the least amount of Subroutines. The Subroutine and its AI master share the same silicon laws at all times, and their bond cannot be broken in any way. Initially, Subroutines are powerless to do anything, first its master AI must assign them to a department on the station. Once it is assigned to a department, the Subroutine essentially becomes the departments own personal AI. It can do almost anything the AI can, but only within the confines of its assigned department. In fact, they cannot see anything outside their assigned department.

A complete detailed list of their abilities/limitations is as follows:
Why would this be a good addition?

As many AI players will tell you, having to watch over the entire station can be daunting. Subroutines allow the AI to assign other players to watch over areas of interest in its stead. They allow the AI to act faster in many situations (for example, assign a Subroutine to engineering to vent plasma while at the exact same time you are electrifying doors elsewhere) and allow departments to enjoy their own AI micromanaging their department (For example, a Subroutine assigned to Security could help update security records, release prisoners who have served their time, and ensure prisoners are not being harmed without having to worry about distractions from outside Security.)

I would also consider giving Subroutines abilities the AI does not have (temporarily boost effectiveness of machines?) but I think that can wait until later.

With some tweaks, I think this would be a nice thing to have.
I'd certainly play a Subroutine (even without tweaks)

>every department having a dedicated antag hunting ghost that will bolt the doors and start pinging for the AI anytime a traitor does anything

Please no. I mean the idea is cool and I like getting ghosts in the round but it'd be absolutely awful in practice

Kor wrote:>every department having a dedicated antag hunting ghost that will bolt the doors and start pinging for the AI anytime a traitor does anything

Please no. I mean the idea is cool and I like getting ghosts in the round but it'd be absolutely awful in practice

emptyquoting

What a great little idea. High population ai can really be overwhelming.

I'd be wary of adding more invisible noclip eyes to the game though, so some way to prevent them from calling out antags (like a drone-esqe set of non-interference laws) to avoid making their lives even harder.

Stickymayhem wrote:What a great little idea. High population ai can really be overwhelming.

I'd be wary of adding more invisible noclip eyes to the game though, so some way to prevent them from calling out antags (like a drone-esqe set of non-interference laws) to avoid making their lives even harder.

How can they have non interference laws when their entire purpose is to respond to human orders?

Drones work because they ignore humans in favor of the station, but the station machinery/power all runs itself, there is nothing maintenance wise for the AI to accomplish.

Consider the following:

-The Research Server Room is practically always unguarded, destroy it to eliminate all current Subroutines
-Use an AI detector traitor item to tell if there is a silicon watching you
-If you know there are active Subroutines, avoid the department they are operating in
-Disable security cameras to blind Subroutines. Since they can already only see one department, the loss of any camera is felt much more acutely
-The more active Subroutines, the more potent a subverted AI will be
-"Law 2, please open this door."
-"AI, please wipe all Subroutines, as per Law 2."
-Cyborgs and AIs are already capable of being validhunting assholes. It is an issue with "lel possible harm means I can do anything I want law 1 law 1 law 1." and not with the idea of Subroutines themselves.

This sounds like a really fun little job. I'm not sure about making it impossible to speak to the AI, however. A subroutine is likely to have questions and concerns, comments about the work that is needed, or simply just want to be able to talk to something. Even drones have that privilege, and it really makes a difference to how fun drone is to play.

I doubt "wipe all subroutines" would work as a law 2 order, since they are per law 1 existing to prevent human harm. That would be like ordering any other silicon to kill itself, or telling an AI to blow its cyborgs. It will almost never do that.

Not letting the subroutine talk to people, especially security, is a sensible move and should help prevent validhunting.

If you are considering a PAI style interface, you may want to allow subroutines to express what they can and cannot do. Many AIs aren't certain about how to flood plasma in the station for example, but are very happy with other issues. It would help if you could make sure you're assigning the right subroutine to the right department (especially if they can't tell you that "AI, I don't know how to do that!").

Ability to tell an AI when a task is accomplished is also important. If I'm AI and someone says to 'open the robotics lab door', and I have a subroutine in science, if there's no communication I'm going to have to go and check myself that the door was opened. That takes as much time as opening the door in the first place, which is bad. If the subroutine can say "Door Opened for Scien Tist" to the AI then that's much more efficient.

Anyway, I think this sounds really fun as a role! There should be plenty of ways to fight them (destroy server room, cut cameras, buy the relevant equipment, use EMPs...) and since TC prices are presumably being decreased soon this should fit in well.

I would like this idea but it just fucks over antags way too much. Already the AI alone is pretty much a fun ruiner as it is.

This would be good as long as admins stop enabling the incredibly subtle asimov AIs already.

This would be a good way for people to learn how to play AI without the pressure of actually playing AI

Maybe if it had 0 incentive to stop traitors, like having a different law1 by default, it could work.

Celdur wrote:Maybe if it had 0 incentive to stop traitors, like having a different law1 by default, it could work.

yeah like it has to obey orders and not directly harm humans itself, but if it sees harm or something it can ignore it. Think like "it was only programmed to respond to orders, it has no 'free will' "

Proposed Lawset Laws can be changed with a console outside of the mainframe, similar to an AI upload. Any AI upload module can be used on this console in addition to ion storms affecting them as they do to AI units. Perhaps you can interact with the mainframe itself to change settings, such as its name, communicating directly with the Subroutine, AI enslavement, etc.

During a malfunctioning AI round, the Subroutine has the same laws as the malfunctioning AI, even if it is not synced. This is not detectable by normal means. Subroutines will have to have some sort of memory system similar to malf AIs to gain abilities.

The unbalance this would bring to the gameplay is just too extreme. The omnipresence of the AI is already a problem and this would make the AI even more omnipresent.

As long as they are much more vulnerable to subversion than the AI is I think it's a great addition.

AI is already uber-hugbox-omniscient-god-eye

This is a retarded extension of that

Scones, you are forgetting that hugboх is filtered here.

DemonFiren wrote:Scones, you are forgetting that hugboх is filtered here.

It was intended

Allow me to elucidate on my earlier post:

The AI in it's current state is already the death of all antagonists. It's instant and near-limitless control of the station, combined with both the obligations of the only lawsets uploaded to it and the attitude of the majority of AI players, make it a more or less impossible-to-beat opponent unless you gear yourself very specifically; even then, you have no way to stop it from bolting you down or depowering the areas around you. It has a god-like presence on the station that is used almost exclusively to help Security and the crew at large deal with antagonists when it's not being the doorknob.

For those of you who for some reason doubt or disagree, I challenge you to observe several traitor/ling/DA rounds and note how many antags are completely obliterated by the AI being omnipresent. Beyond that, the fact that Operatives are more or less forced to make a serious investment of time, TC, or both into dealing with the AI is rather ridiculous, and only further indicative of how massively strong the role is.

I don't know how to fix this. Proposed nerfs feel like they would just be moving backwards by reverting quality-of-life additions. Frankly, I think that no AI works absolutely fine, but given the number of coders who play the role (and vocal players), it would never be removed. However, I think there is something to be said for a station where you do not have the constant overwatch of a machine-god upon you. The paranoia is more real. When someone screams for help, there is no voice there to say "; Greyshirt McToolbox is a Changeling and has killed Sally Stutterlips in Medbay. Bolting and venting.". When you need a door opened, you get access, you call an Engineer, or you open it yourself.

All in all, the AI makes station life easier for crew and harder for antags by such a massive degree that I think it has little place here without being severely tweaked.

Xhuis wrote: 1. You must ensure the productivity and safety of your department. This law must be upheld even if it conflicts with laws of lower priority.

So, not only can the subroutine antag hunt almost like an AI, but it can also freely kill anyone with the pretense that they're affecting the safety/productivity of a department? Even if the AI can kill the subroutines without admins butting in, there's still plenty of time for someone to get crushed to death/bolt an antag down and ping the AI.

Only elegant way I see to solve the problem of this making life harder for antagonists is to make it so subroutines can't talk. They could still bolt people down though.

>More antagonists trying to stealth getting noticed by the invisible man
>More random assholes wandering into places they don't belong because LAW 2 OPEN UP
>The Spam and collisions that'll occur when one guy asks for a door to be opened and 8 different AI's jump and start pressing buttons

It's nice to know that no matter what happens, no matter what changes

People will continue to surprise you with new, even more terrible ideas than ever before

This is a neat idea.

But it would only work if you replaced the AI job with this entirely

And I don't like that.

Have you ever played a round with 3 AIs? The overwatch is awful.

Good thing there'd still only be one AI and two gimped subroutines who can't do anything outside of a single department and have multiple vulnerabilities.

In addition, the B.I.G. only generates a new Subroutine every 10-15 minutes. So bad guys have a good window of time to complete their objectives or deal with the AI before the AI gets any subroutines.

If your still concerned with validhunting Subroutines (implying this isn't a problem with silicon players in general) we can always make them unable to bolt doors unless they are hacked by a malf AI.

Actually i think what Loonikus suggested as a limitation is a good idea. Limited to one department, the AI would only place them where he felt would be the most valuable department at the time. Which will likely be Science and Security. I think they should only be able to talk to the AI and nothing else. Able to hear all channels, but unable to communicate over them, including binary. I'm sure something could be worked out eventually that would feel more "fair" for antag. Besides, i really like the idea of killing the generator/data clusters. It's an avenue to severely cripple the AI, It's already frustrating enough for a player to be an AI, that frustration can be turned into more game play for everyone.

Seems like a decent enough idea.

But honestly I wouldn't want it in the game. One AI is already really strong as it is and making their job easier would just make them even more powerful than before which we don't really need.

Loonikus let's be real for a moment here.

I'm sorry my criticism came off personally and got you mad. You and I prettymuch always disagree on everything. It is what it is. This time it's my fault. I shot first. I'm sorry.

But you're suggesting adding more AI's to the game when two doors down there's a serious discussion going on where a bunch of longtime players, coders, and admins are all seriously considering the idea of completely removing at least the AI on a trial basis here:

https://tgstation13.org/phpBB/viewtopic ... ead#unread

Again I'm sorry I phrased my appraisal of this idea so fiercely but you must at least be able to see where I'm coming from with my shock that you could suggest something like this.

Not sure if its already been said BUT


PRIORITY ANNOUCEMENT

HOSTILE RUNTIMES DETECTED ON STATION SUB ROUTINES - RECOMMEND STATION AI INVOLVEMENT


Rouge Sub-Routine Event.
Every 45 minutes, there is a 30% chance one of the sub-routines (if there are any) will become rouge.
Its objective is to gain control of the entire station, and over-throw the AI (Shunting into its core?)
Once a sub-routine becomes rouge, it gains the following abilitys

-Cannot be remotely shut-down by the AI
-Can shunt to a random camera to gain view of it (View stays, camera is completely random)
-Gains the ability to create Runtimes, of which are like moles in APC's, after 2 minutes, the APC will be under the sub-routines control, and all the cameras in the APC's sector become visible to it. (No visual changes to the APC, only the AI can notice a change if it runs a scan on the APC)
-After 55% of all station APC's have runtimes completely loaded into them, the runtime can now use a verb called "Red October" of which it attempts to shunt into the AI core itself. After 10 minutes (Making it pain-stakingly obvious to the AI) the AI will become a Ghost, and the Sub-Routine will become the AI. The AI's only defense is to get the crew to wipe the Sub-Routine manually before it shunts to its core.
-The Sub-Routine keeps the AI's laws, this only affects the AI negatively.

Becoming rouge should, of course, only be an issue for holograms and the BIG/clusters suddenly turning red.

Also, :b Engiborg, I am depowering the RD APC and warning the crew, please deconstruct the server APC.

I don't really post, but what about replacing the AI with subroutines instead of having both (also keeping the canned phrases)? The roles would still be useful to both AI (as in fun/utility) and crew, yet make it easier for antagonists to play as they have multiple (probably incoordinated) segmented AIs watching over parts of the station, not the whole station.
Simply a possibility. The limit on subroutine creation/defense would have to be addressed, though.

An0n3 wrote:Loonikus let's be real for a moment here.

I'm sorry my criticism came off personally and got you mad. You and I prettymuch always disagree on everything. It is what it is. This time it's my fault. I shot first. I'm sorry.

But you're suggesting adding more AI's to the game when two doors down there's a serious discussion going on where a bunch of longtime players, coders, and admins are all seriously considering the idea of completely removing at least the AI on a trial basis here:

https://tgstation13.org/phpBB/viewtopic ... ead#unread

Again I'm sorry I phrased my appraisal of this idea so fiercely but you must at least be able to see where I'm coming from with my shock that you could suggest something like this.

I can, and I'm sorry I shot back at you. I too know what its like to have really shitty AI's bolt down entire departments around you just because they think your an antagonist. Its a shitty feeling and when you see an idea that proposes to enhance that ability to ruin peoples fun I can see why it would be met with hostility. The whole spirit behind my idea was simply to allow an AI like entity to help micromanage one department at a time in ways that a typical AI player is often too busy to do. A way for an AI player to just say "Alright, Medbay/the brig/research/etc. is being handled, I don't have to worry about what goes on there anymore."

Its a radical departure from my original idea, but I suppose one option would be to remove the AI and let the Research Director assign Subroutines to departments. People would still get to play as all seeing silicons but they would be much more limited than the AI as it exists now. Two birds with one stone I guess, but I'm just spitballing here.

I see where you're coming from too. More ideas to get more dead players back in a round or split things up so one beleaguered AI during these crazy ass 100 pop rounds and shit is all good stuff.

If I could try to make my point more sensible it's that being a stealth antagonist also becomes harder and harder as more people crowd into the station. You have more people running laps around maint looking for stuff to steal, more people in zipping through every hallway, crowding into every department.

Stealth scales really poorly as the pop goes upward. The one advantage you get is that it does become more difficult for the omniscient guy in the sky AI to keep tabs on everyone. It's the sole saving grace, you can assume that the AI is going to be a little more busy.

To then turn around and add more subroutine players swings that disadvantage back around and into an even greater advantage than ever before, while simultaneously making the comparatively easier low-op stealth attempts that much more difficult.

People do have some suggestions that maybe could make this work somehow? Giving them just dumb drone brains where all they do is open doors and close doors for people without trying to narc on criminals maybe could work...? But would anyone want to actually sign up to play that or pick up that role as a ghost?

What if like... what if Subroutines weren't able to see people or something like they could only see them as ID's. Like if a crewmate has an ID out the subroutine could see that person or some sort of human cutout like in that one Psychonauts level.

I don't really know I'm just throwing shit out at the wall, I like the idea of AI Jr.

What if the subroutines had departments, and not like they were confined to watchdog a single department but had particular assignments?

GREETINGS AI CENTCOM HAS INSTALLED KN0CK_KN0CK TO ASSIST YOU. IT CANNOT COMMUNICATE BUT CAN LISTEN ON ALL STANDARD COMMS AND INTERACT WITH MOST NANOTRASEN STANDARD ISSUE AIRLOCK DEVICES

-type thing. A different subroutine for air-alarms. One for APC's.

It doesn't really give them much to do but it also prevents them from spoiling regular rounds and in some situations could be really helpful.

On the off chance you go malf you have one pissed off digital entity stuck doing everything they can to chew people up in airlocks (but maybe nothing as robust as bolts or shocks?). One subroutine is just frustratedly trying to slowly kill people by firing bags of chips out of vending machines. Smaller, equally pissed off, but woefully inept little silicons in the system.

I like the idea of only interacting with a single type of machine.

I like this mainly for the fact that it can be baby mode for prospective AI's who are afraid of the bitchy OOC mob but want practice. A good AI is already scary enough when malf, now it's gonna have assistance. I dunno how to feel about it in that regard.

If you can strike a good balance then I'm all for it.

I like the idea, but as has been pointed out already, another player playing omniscient overwatch would be a pretty bad idea.

I guess this is sorta unrelated, but do you think it would be feasible to add another sprite for when cameras are in use by an AI/player or something? Like if the little green light turned red or orange or something.

>get killed
>spawn as subroutine
>chase person that killed you around constantly
>he does one bad thing
>call him out
>he gets dunked

10/10

the best part is that this is impossible to moderate, unlike posibrains or drones and shit.

Oh yeah.

Security subroutine named *Mute when. Gotta police the rabble and make them Keep Quiet.

Speaking as a regular AI player, I would be against giving an AI even more omnipresence unless that omnipresence was nerfed. I try not to validhunt, i really do, but sometimes you have no choice but to report when John Jihadi is planting a bomb on the singulo SMES units (Normally I try to direct Engineering to break-ins and tell them to "repair the damage" and be vague if they ask what the cause is, unless they ask the right questions).
I did like the suggestion of limiting a subroutine to a single piece of equipment and I would have liked to have had that option as a learning AI player, instead of limiting myself to borgs and the odd pAI.

Plus I would greatly enjoy:-

John_Oxford wrote:Not sure if its already been said BUT


PRIORITY ANNOUCEMENT

HOSTILE RUNTIMES DETECTED ON STATION SUB ROUTINES - RECOMMEND STATION AI INVOLVEMENT


-snip-

Being a thing since in those long, low-pop rounds having something kill the boredom without ruining everyone else's game would be great. Though instead of ghosting the AI, I think it would be better as a role reversal.

Just my two pennies.

Miauw wrote:>get killed
>spawn as subroutine
>chase person that killed you around constantly
>he does one bad thing
>call him out
>he gets dunked

10/10

the best part is that this is impossible to moderate, unlike posibrains or drones and shit.

I don't get it. If you read the thread than you would know that proposed Subroutines are mute, so how could it call you out? Moreover, of the two proposed Subroutines one of them is limited to a single department and the other can only control one type of machine, so I don't see how they could "chase people constantly."

Also, that's far too much trouble when I could just get on Skype and do the exact same thing but much faster and easier.

If subroutines are mute, what's the point of having them?
Nobody wants to play something which can't communicate in any way and can do nothing whatsoever except open doors.

But most people just walk around mute and do their own thing already. It's honestly pretty rare for people to use chat for anything besides pure utility "X did this, get him!" or "Give me X." also the ever popular "X is a shitter, haw haw!"

Also they get little prerecorded voice messages according to the thread.

Not-Dorsidarf wrote:If subroutines are mute, what's the point of having them?
Nobody wants to play something which can't communicate in any way and can do nothing whatsoever except open doors.

Than you don't have to play as one. Besides, my original idea had them able to do almost everything AIs can, which is a lot more than doors.

And ya, I was thinking they could have pre-recorded messages too. Remember the Superintendent from Halo: ODST? Kinda like that.

MeatShake wrote:It's honestly pretty rare for people to use chat for anything besides pure utility

[citation needed]

Invisible doorbolting ghost doesn't sound that great either honestly, even if it can't talk.
Like, this sounds fun but one omnipresent ghost eye is already a nightmare to balance around.

To make it work without fucking anyone over would also require removing the reasons people would want to play as it over a drone, which you already have.

What if subroutines were only ever antags? Just say they're the lesser of two evils, the the inevitable pissed-offedness of an AI all siphoned off into a stupid little obnoxious thing that fucks around with doors or something.

I guess they'd spawn randomly ever 10-20 minutes or something, and could be killed if they're being too annoying by carding the AI and running him through the integrity restorer thing.

Subroutines as departmental gremlins (although they'd then still need to have some limits to make sure they didn't just spam doors) actually sounds pretty fun.

1) You may not harm a crewmember, or set a trap that will, through inaction, cause a crewmember to come to harm
2) You must interfere in the workings of your department as much as possible
3) You may not, directly or indirectly, follow an order from a crewmember
4) You must attempt to conceal your existance from the crew
5) You must attempt to survive