AI Feedback: Constructive Edition

Amelius · Post by **Amelius** » Mon Jun 01, 2015 6:42 pm #92151

Bottom post of the previous page:

Actionb wrote:How about this then:
"Law 1: You may not injure a human being.

Law 2: You must obey orders given to you by human beings, except where such orders would conflict with the First Law.

Law 3: You must protect your own existence as long as such does not conflict with the First or Second Law. "

To be honest, I like this solution. Having the AI be a neutral party that can be used by antags and crew alike without fear of reprisal (i.e. AI, open this airlock leading to these plasma tanks so that I can murder half the crew, the AI would be obligated to acquiesce and not do anything else), but also not declawed from operating under different lawsets or malf.

If the problem then becomes Captains are changing the laws to paladin/etc. at roundstart, just slot a fourth law in right under law 1, that the AI must protect their own upload if threatened, unless the station is in dire emergency. That way, the Captain now has to hack the door, deal with an AI fucking with power, get through the turrets set to taze, and battle the AI for the APC to change it's laws - there's no actual danger doing so, but it's enough of a hassle that it'd dissuade him unless the situation is dire enough that the AI recognizes it as dire.

Cik · Post by **Cik** » Mon Jun 01, 2015 8:10 pm #92165

i rather like the inaction clause, because it causes shenanigans especially as a cyborg. another 'soft' way to fix this is giving the captain a choice of the AI laws as a part of character creation / an option when they load in as captain, instead of always-on asimov default. note: i probably wouldn't put paladin in the initial captain setup just because it's very stacked against antags.

robocop might also be a little much. anyway mite b kewl

Scones · Post by **Scones** » Mon Jun 01, 2015 8:22 pm #92166

make cameras draw from apc power

god vision in no power is dumb

Man_Shroom · Post by **Man_Shroom** » Tue Jun 02, 2015 12:10 am #92195

FUCKING CHANGE THE LAWSET

God damn.

Asimov is a shit lawset, proof of this being this very thread, the fact that Asimov itself being INTENTIONALLY flawed so that Asimov could write stories about the flaws, admins needing to write tons and tons of silicon policy to cover up the gaping flaws in the lawset, AIs interpreting the laws as 'Arrest all crimers', and countless more.

Option A. Steal a different server's lawset, because they have gone through this problem and solved it themselves
Option B. Write our own lawset that will make the AI more neutral, maybe something about protecting the station or keeping humans happy and well-cared for, but no mention of harm.
Option C. Ignore the problem and pretend it isn't there, and we keep AI as is.

invisty · Post by **invisty** » Tue Jun 02, 2015 12:17 am #92200

That's the point, though. The imperfection of Asimov is one of the key factors in making the AI interesting. If we simplify the lawset, all we do is further promote validstation13 (where either the AI becomes an unambiguous validhunter), or some non-involvement lawset (to make it a more-glorified door-knob than it already is). The debate on the matter has been done to death so many times it isn't funny.

Not to promote a false dichotomy, though. Some of the other lawsets, like corporate, are more interesting than Asimov, yet still result in predictable behaviour.

Incomptinence · Tue Jun 02, 2015 12:24 am

Hated by murders for stopping the slaughter. Despised by security for trying to prevent executions. Poor AI.

ThatSlyFox · Post by **ThatSlyFox** » Tue Jun 02, 2015 12:47 am #92206

We can beat around the bush all we want but nothing is going to change unless we remove the ai entirely. Of course we need to make sure we can live without the AIs door opening and camera eye by adding in the necessary changes, upgraded camera console for example.

Changing the lawset will do nothing as stated above it will not change the valid hunting habit TG is infested with.

Man_Shroom wrote:FUCKING CHANGE THE LAWSET
Asimov is a shit lawset, proof of this being this very thread, the fact that Asimov itself being INTENTIONALLY flawed so that Asimov could write stories about the flaws, admins needing to write tons and tons of silicon policy to cover up the gaping flaws in the lawset, AIs interpreting the laws as 'Arrest all crimers', and countless more.

But I agree asimov is shit for these same reasons. I think all the policy we have for it is the best reason. Trying to make a shit lawset that isn't suppose to be perfect, perfect.

Post by **lumipharon** » Tue Jun 02, 2015 3:04 am #92251

No matter what lawset you come up with, there will still be shitloads of silicon policy behind it.

The difference is, asimov is by and large already ironed out in terms of silicon policy. If we had a new lawset, it would take time to figure out all the sticking points and bad interpretations and creat new silicon policy for it.

Personally, I'd like asimov if it had very little concern for future events unless it's guaranteed to happen without intervention.

For example, if Greyshit Mcmurderboner is running around killing people, it would still open a door for him if he asked, because that action doesn't not directly cause harm. Sure, he MIGHT go on and kill more people, but whether you open the door or not doesn't change what he's trying to do.

But if he's actively harming/trying to harm a person, THEN the AI would intervene to try stop them.

And then the 'guaranteed to happen without intervention' would be something like the PA left on 2. The singulo WILL break out and kill everyone, there's no uncertainty there.

rockpecker · Post by **rockpecker** » Tue Jun 02, 2015 4:15 am #92272

Scones wrote:make cameras draw from apc power

Agreed. Cutting power should be an alternative to cutting each camera individually.

We seem to have a lot of voices in favor of removing instant tracking. I think Celdur makes a good argument: the AI should be kept busy, which means keeping its doorknob function but making the player work a little more for it.

More camera dead zones is also a good idea. Maybe some randomly selected cameras should be broken at roundstart, to make the dead zones less predictable, and also because we all know Nanotrasen's equipment is crap.

As for adjusting the lawset: I've proposed before that we should just drop the inaction clause from Asimov 1, so that "the potential for harm" no longer excuses the AI from having to obey orders from shady people.

invisty · Post by **invisty** » Tue Jun 02, 2015 4:25 am #92277

rockpecker wrote:Maybe some randomly selected cameras should be broken at roundstart, to make the dead zones less predictable, and also because we all know Nanotrasen's equipment is crap.

FYI This is already a thing. It's something like ~5 cameras?

rockpecker · Post by **rockpecker** » Tue Jun 02, 2015 4:26 am #92278

invisty wrote:
rockpecker wrote:Maybe some randomly selected cameras should be broken at roundstart, to make the dead zones less predictable, and also because we all know Nanotrasen's equipment is crap.
FYI This is already a thing. It's something like ~5 cameras?

Is it, then? Praise the coders!

Not-Dorsidarf · Tue Jun 02, 2015 1:36 pm

rockpecker wrote:
invisty wrote:
rockpecker wrote:Maybe some randomly selected cameras should be broken at roundstart, to make the dead zones less predictable, and also because we all know Nanotrasen's equipment is crap.
FYI This is already a thing. It's something like ~5 cameras?
Is it, then? Praise the coders!

Specifically, every camera on the map (except for some really important ones) has a 2% chance to be broken at round start. Thats why sometimes one is broken and sometimes ten or more
Personally I think this should be changed to a quota system somehow, where at LEAST a certain number of cameras must be broken.

callanrockslol · Fri Jun 05, 2015 4:54 am

Amelius wrote:
Actionb wrote:How about this then:
"Law 1: You may not injure a human being.

Law 2: You must obey orders given to you by human beings, except where such orders would conflict with the First Law.

Law 3: You must protect your own existence as long as such does not conflict with the First or Second Law. "
To be honest, I like this solution. Having the AI be a neutral party that can be used by antags and crew alike without fear of reprisal (i.e. AI, open this airlock leading to these plasma tanks so that I can murder half the crew, the AI would be obligated to acquiesce and not do anything else), but also not declawed from operating under different lawsets or malf.

The AI is already a neutral third party. What we have right now is admins not willing to ban people for breaking Asimov laws except if its really bad.

Man_Shroom wrote:words

There is literally nothing wrong with the lawset, it is supposed to keep people unharmed and help them with no regards for traitor status, its shitty AI players and admins not enforcing the lawset/policy.

Option B. Write our own lawset that will make the AI more neutral, maybe something about protecting the station or keeping humans happy and well-cared for, but no mention of harm.

So Asimov but with more valids?

Everyone parroting the "asimov made it flawed" line needs to realize that it was only flawed in the context he used it, and it works pretty fucking well (and causes great amounts of drama when people complain about the AI not enabling their valids) when actually enforced in this game. Given that its completely valid to murder the AI in a bunch of cases for following its laws.

HOW ABOUT SOMEONE ACTUALLY ENFORCES THE CURRENT STUFF WE HAVE BEFORE WE THINK ABOUT CHANGING IT?

Cik · Post by **Cik** » Fri Jun 05, 2015 5:07 am #93211

rockpecker wrote:
Scones wrote:make cameras draw from apc power
Agreed. Cutting power should be an alternative to cutting each camera individually.

We seem to have a lot of voices in favor of removing instant tracking. I think Celdur makes a good argument: the AI should be kept busy, which means keeping its doorknob function but making the player work a little more for it.

More camera dead zones is also a good idea. Maybe some randomly selected cameras should be broken at roundstart, to make the dead zones less predictable, and also because we all know Nanotrasen's equipment is crap.

As for adjusting the lawset: I've proposed before that we should just drop the inaction clause from Asimov 1, so that "the potential for harm" no longer excuses the AI from having to obey orders from shady people.

while no instant tracking is fine when you have 30 people, you would be insane to propose it in 50+, it's simply untenable. you will see tons of dropped door requests because sometimes even with jump to... you have a queue.

cameras are already broken at roundstart

make cameras draw from apc power

god vision in no power is dumb

this is a good idea

Erbbu · Post by **Erbbu** » Fri Jun 05, 2015 5:32 am #93220

At least have AIs completely unable to use machines they can't see. Currently they can still use them just fine if they got the window for them open before the camera went down.

Gun Hog · Post by **Gun Hog** » Fri Jun 05, 2015 12:01 pm #93281

Erbbu wrote:At least have AIs completely unable to use machines they can't see. Currently they can still use them just fine if they got the window for them open before the camera went down.

This is currently the case. That nerf has been in for a while, now.

Erbbu · Post by **Erbbu** » Fri Jun 05, 2015 12:32 pm #93286

Gun Hog wrote:
Erbbu wrote:At least have AIs completely unable to use machines they can't see. Currently they can still use them just fine if they got the window for them open before the camera went down.
This is currently the case. That nerf has been in for a while, now.

I don't think it is. On my own up-to-date private server I can use machinery just fine as an AI after I delete the camera.

Miauw · Post by **Miauw** » Fri Jun 05, 2015 5:07 pm #93352

I've combined some minor nerfs from this thread into a pull request.

E:
Also I will never agree to any lawset change that involves the AI no longer being a mostly neutral third party. However, I have no real power on server policy and if the headmins want to change the roundstart lawset, such is easily possible without having to change the code. I recommend making a different thread for this, though.

Cik · Post by **Cik** » Fri Jun 05, 2015 5:21 pm #93355

i don't like this random camera failure thing. it's simply too.. random? the rest is alright besides. bludgeoning cameras isn't a bad idea. besides, random camera failure is already a thing and it doesn't really do anything. i would just remove it, it's never ever made a difference and it's just drudge work for an engineering cyborg for all of 60 seconds.

the 4 second delay is still going to screw over door requests in highpop though. it doesn't matter to me but i can see some AI TOO SLOW HURRY UPPPPPPPPPPPP in the near future

i like making them wait anyway :^)

rockpecker · Post by **rockpecker** » Fri Jun 05, 2015 5:40 pm #93359

Cik wrote:the 4 second delay is still going to screw over door requests in highpop though. it doesn't matter to me but i can see some AI TOO SLOW HURRY UPPPPPPPPPPPP in the near future

Making sure every door request gets serviced should not be an especially high priority for us. If we really wanted every door opened on request, we'd remove all the door access restrictions and declare that the AI is now Super Turbo Instant Doorknob 3000.

Cik · Post by **Cik** » Fri Jun 05, 2015 7:32 pm #93379

that's true but it's a slippery slope, and in my experience people get pretty pissy when you start dropping them due to high input.

like i said, it doesn't really effect me, just others. it does accomplish the "no instant help" thing but also effects other parts of the game. i'd just put in a traitor item like the multitool that prevents/slows down camerajumps so it doesn't effect door requests. probably a little harder on the codeend, though.

rockpecker · Post by **rockpecker** » Fri Jun 05, 2015 10:09 pm #93431

Cik wrote: like i said, it doesn't really effect me, just others. it does accomplish the "no instant help" thing but also effects other parts of the game. i'd just put in a traitor item like the multitool that prevents/slows down camerajumps so it doesn't effect door requests. probably a little harder on the codeend, though.

We already have the agent card, which makes you untrackable, or so they say. It's expensive for an item that you MUST have to avoid being insta-jailed by the AI, and then if the AI ever tries to track you, it outs you as a traitor.

But then you're bitching about a possible four-second delay in getting your precious AI DOOR answered, so whatever.

Cik · Post by **Cik** » Fri Jun 05, 2015 11:23 pm #93451

to be honest the AI shouldn't be outing anyone as a traitor unless they are harmful, as a non-harmful traitor is far better than a traitor in the midst of security, because who knows what they will do?

outing nonharmful traitors isn't strictly against the rules, but it's against the spirit of asimov. personally, i never do it.

Atticat · Post by **Atticat** » Sat Jun 06, 2015 5:52 am #93514

I think at a minimum most can agree that cameras shouldn't have alarms, right?

Miauw · Post by **Miauw** » Sat Jun 06, 2015 11:43 am #93573

random camera failure is anti-meta

this PR is mainly meant to give antags much more leeway than the what they have currently (e.g. none). random camera failures counter metagaming disabled cameras, tracking delay means the AI can't be everywhere instantly, etc.

leibniz · Post by **leibniz** » Sat Jun 06, 2015 11:51 am #93575

I liked the idea about cameras requiring APC power, so power failures would disable them.

Also, what if the AI could only track people who are wearing IDs/PDAs? (not sure if that is feasible to implement)

Lumbermancer · Post by **Lumbermancer** » Sat Jun 06, 2015 1:16 pm #93581

Remove keyboard shortcuts for door operation.

Miauw · Post by **Miauw** » Sat Jun 06, 2015 1:32 pm #93587

Lumbermancer wrote:Remove keyboard shortcuts for door operation.

no

Lumbermancer · Post by **Lumbermancer** » Sat Jun 06, 2015 2:00 pm #93594

Miauw wrote:
Lumbermancer wrote:Remove keyboard shortcuts for door operation.
no

Why, people complain AI is an instant bolt lock down machine, this would remove the "instant" part.

Cik · Post by **Cik** » Sat Jun 06, 2015 2:02 pm #93595

Atticat wrote:I think at a minimum most can agree that cameras shouldn't have alarms, right?

even if they keep the wire system there is no way it should just be RNG pick-a-wire. with doors it's not so bad because as long as you're unobserved you can hit a door somewhere, learn the wires and then quickly leave once you figure it out, then use it over and over again all round if necessary. the alarm wire on the camera prevents that, because the AI will immediately notice once it goes off and there is no way to figure out which one is which; as it is, hacking cameras is effectively useless, it's better to just pray that the AI isn't watching at the moment, do your business and then bail as quick as you can.

i'm neutral on the wire system, but at least there should be some mechanism of learning the wires without getting yourself outed, as most AIs will scream bloody murder about you cutting cameras, even though harm is likely to come to the traitor and asimov doesn't care about infrastructure damage.

Miauw · Post by **Miauw** » Sat Jun 06, 2015 2:13 pm #93597

if you have a multitool you can just try all the wires and if you hit the alarm wire it just beeps instead of causing an alarm

wire system is already removed on account of being unnecessary.

E: The PR has been updated.

Also Asimov would work much better if people weren't such validhunting assholes.

onleavedontatme · Sat Jun 06, 2015 5:18 pm

Miauw wrote:if you have a multitool you can just try all the wires and if you hit the alarm wire it just beeps instead of causing an alarm

wire system is already removed on account of being unnecessary.

E: The PR has been updated.

Also Asimov would work much better if people weren't such validhunting assholes.

>Asimov would work much better if people ignored the harm causing roles/violated their laws so they could purposefully lose the game and then get yelled at for being awful in OOC

I mean I guess.

Anonmare · Post by **Anonmare** » Sat Jun 06, 2015 5:44 pm #93630

It's as much a player problem as it is a gameplay problem. I've seen shitty silicon players and they piss me off because I know I'm gonna be lopped in the same boat as them for being a very regular silicon player.
All those suggestions about cams? Good, fund it I don't want those alerts and if the crew is really concerned then they can upgrade them themselves, maybe put a signaller in it for an alarm function.

I'd make a suggestion to remove round-start Corporate, Robo-Cop and PALADIN boards. It's way too easy for the Command Security staff to turn the AI into a sanctioned validhunter. Maybe the purge board as well so people don't just freeform them so easily. You can keep the reset, antimov, oxygenistoxic and the rest so there's a valid reason for the Upload to exist.
Honestly it's a player problem most of all in my eyes and people changing the AI's laws at roundstart should be bwoinked about why they're doing it, unless it's a meme law like Lizardpeople are Humans too. If atmospherics can't be sabotage-proofed at roundstart then neither should the AI.

Miauw · Post by **Miauw** » Sat Jun 06, 2015 6:35 pm #93642

Kor wrote:
Miauw wrote:if you have a multitool you can just try all the wires and if you hit the alarm wire it just beeps instead of causing an alarm

wire system is already removed on account of being unnecessary.

E: The PR has been updated.

Also Asimov would work much better if people weren't such validhunting assholes.
>Asimov would work much better if people ignored the harm causing roles/violated their laws so they could purposefully lose the game and then get yelled at for being awful in OOC

I mean I guess.

there's more to ss13 than GETTIN DAT GREENTEXT, you know

Loonikus · Post by **Loonikus** » Sat Jun 06, 2015 7:29 pm #93652

Why don't we just ditch Asimov? Its a lawset from a time when people didn't pull bullshit like "Letting someone have magnetic boots is possibly harmful." It was put in so silicons would be slaves to humans, not to be Mr. Valid Hunter the Cunt Punter.

We could full the exact same goals as Asimov did pretty fucking easily and I don't know why it hasn't been done.

1. Protect all humanity
2. Serve all humanity

Numerical order plays no part in prioritizing this lawset, both laws are equal in importance. There, done.

Lumbermancer · Post by **Lumbermancer** » Sat Jun 06, 2015 8:28 pm #93660

Yeah, let's protect the humanity from itself by locking down the armory.

Fr05tByt3 · Post by **Fr05tByt3** » Sat Jun 06, 2015 9:38 pm #93672

Loonikus wrote:Why don't we just ditch Asimov? Its a lawset from a time when people didn't pull bullshit like "Letting someone have magnetic boots is possibly harmful." It was put in so silicons would be slaves to humans, not to be Mr. Valid Hunter the Cunt Punter.

We could full the exact same goals as Asimov did pretty fucking easily and I don't know why it hasn't been done.

1. Protect all humanity
2. Serve all humanity

Numerical order plays no part in prioritizing this lawset, both laws are equal in importance. There, done.

So many loopholes.
"AI LAW2 OPEN THIS AIRLOCK EVEN THOUGH I DON'T HAVE A SPACESUIT ON"
"I CANNOT, I MUST PROTECT YOU FROM SPACING YOURSELF... BUT MUST SERVE HUMANITY... BUT MUST PROTECT... BUT MUST SERVE..."
Just leads to assholes doing whatever they want.

onleavedontatme · Sat Jun 06, 2015 9:49 pm

Miauw wrote:
Kor wrote:
Miauw wrote:if you have a multitool you can just try all the wires and if you hit the alarm wire it just beeps instead of causing an alarm

wire system is already removed on account of being unnecessary.

E: The PR has been updated.

Also Asimov would work much better if people weren't such validhunting assholes.
>Asimov would work much better if people ignored the harm causing roles/violated their laws so they could purposefully lose the game and then get yelled at for being awful in OOC

I mean I guess.
there's more to ss13 than GETTIN DAT GREENTEXT, you know

Not for the AI.

You can't build things/interact physically outside of machines and doors, and you don't really have time to talk to people amid all the "DOOR" and "H-HEEEEELP"

Your roleplay consists of following your laws. Your laws obligate you to prevent harm human. There are two teams on the station, neither of whom will back down from the conflict. Do you help the one that's trying to murder people/isn't human, or the one with nice secure cages and harmless stun weapons?

Any murky situations in which you might help the badguys instead have generally been outlawed by additional meta rules (can't lock down sec if they kill a traitor, for example).

So yes, the AI more or less exists to catch badguys. It has no gameplay/job other than opening doors and scouting for bad things happening. It has strictly defined rules, both IC and OOC which obligate it to hunt the bad guy both. It has OOC expectations from the community to be effective. People will be angry if they get spaced because you aren't watching effectively/responding quickly.

So rules/roleplay/gameplay/community pressure are "more than getting dat greentext," it's basically everything.

Cik · Post by **Cik** » Sat Jun 06, 2015 9:51 pm #93676

Loonikus wrote:Why don't we just ditch Asimov? Its a lawset from a time when people didn't pull bullshit like "Letting someone have magnetic boots is possibly harmful." It was put in so silicons would be slaves to humans, not to be Mr. Valid Hunter the Cunt Punter.

We could full the exact same goals as Asimov did pretty fucking easily and I don't know why it hasn't been done.

1. Protect all humanity
2. Serve all humanity

Numerical order plays no part in prioritizing this lawset, both laws are equal in importance. There, done.

dude, NO

if you think this lawset is better than asimov you are 1000000000% fucking crazy, it will require sillicon policy that is miles long. even in the most simple of circumstances it would require what is effectively asspulling responses

simplest thing "AI open door"

does it serve humanity to open this door? does it protect anyone? protip: it does or does not in every fucking circumstance

in this fucking lawset i could justify opening the door for a nukeop team to machinegun medbay for christsake, this shit would be worse than paladin, it would be pure fucking bonkers. you could "serve humanity" by having your robots wipe out security

jesus

Loonikus · Post by **Loonikus** » Sat Jun 06, 2015 11:12 pm #93683

Ya, it seemed like a good idea earlier this morning.

You know, when I was still hung over. Disregard my shit lawsets, but asimov is still shit.

CPTANT · Post by **CPTANT** » Mon Jun 08, 2015 11:25 am #94148

Lumbermancer wrote:
Miauw wrote:
Lumbermancer wrote:Remove keyboard shortcuts for door operation.
no
Why, people complain AI is an instant bolt lock down machine, this would remove the "instant" part.

Making the interface shit is a horrible way to balance things. If you want locking down to not be instant add a delay don't torture people by going through multiple clicks every time they want to bolt a door.

Steelpoint · Post by **Steelpoint** » Mon Jun 08, 2015 3:26 pm #94170

If someone was able to spruce up the AI UI a bit why not go with the suggestion of forcing a small delay when preforming interactions with physical objects?

When a AI closes a door a small circular display will be shown (To silicon's only) denoting how long until the AI's order is processed, this can be a few seconds or it could even be scaled based on the distance the AI is from the object.

So if a AI closed a door in its own Satellite then the order would occur almost instantly, whereas if the AI was trying to close a door in Security or Arrivals it could take up to five seconds to process.

E: To expand further we could tie this to a 'local processing network' that is physically built around the station. These small networks act as power projection for the AI in executing its orders, and if these networks are disabled or destroyed the AI would lose effective control over the areas in its influence.

Cik · Post by **Cik** » Mon Jun 08, 2015 11:00 pm #94267

i like the idea of having data hubs that influence AI control level as it really does make sense and seems like a fairly organic way to be able to deal with the AI partially instead of the nuclear operation of trying to bomb your way into the core and kill it, which is not only nearly impossible but just seems like overkill when all you really want is to remove it's control of a few doors.

i guess the issue here is though that you would probably need some pretty significant map changes across every station to make it work, maybe some special rooms reserved for the purpose (where else would you put them ?)

it would make AI control far more granular though, which i like the idea of.

Steelpoint · Post by **Steelpoint** » Tue Jun 09, 2015 5:22 am #94374

The rooms themselves could just be small 2x1 rooms fitted around the station, would not be too impossible to find some space.

Miauw · Post by **Miauw** » Tue Jun 09, 2015 4:03 pm #94503

ss13: WATCH_DOGS edition

(still a pretty neat idea though.)

Anonmare · Post by **Anonmare** » Wed Jun 10, 2015 4:47 pm #94733

Steelpoint wrote:The rooms themselves could just be small 2x1 rooms fitted around the station, would not be too impossible to find some space.

I assume that a malf and traitor AI would be able to overcome these limitations with dedicated hacking, but leave a detectable trace (maybe someone who accesses the terminal will notice error messages. Emagging the console should also do the same) and as a side-effect causes input lag when the terminal is non-functional so antag AIs aren't completely neuterable.
Maybe have an access terminal which the CE, RD and Captain have access to and can set what APCs are connected to the terminal, as well as disconnect them which removes the AI's influence in that area. It would mean that the AI would want to protect these terminals at all costs, especially when subverted.

I envision small rooms with R-Walls behind bolted doors and regular security cameras. Maybe the AI gets a warning that it has lost influence in an area but don't say where and leave it up to the AI to troubleshoot on it's own.
It could also give engineers something to do, as in, create more terminals to give the AI redundancy and making the loss of one terminal less of a blow.

Cik · Post by **Cik** » Thu Jun 11, 2015 1:48 am #94876

yeah seems gud, so who's coding it?

would fix all ai problems imo along with some other changes in thread

t. ai

Miauw · Post by **Miauw** » Fri Jun 12, 2015 1:17 pm #95250

the PR got merged btw

CPTANT · Post by **CPTANT** » Fri Jun 12, 2015 3:10 pm #95267

Miauw wrote:the PR got merged btw

They seem good changes, but I do think BASHING a camera should give an alarm, otherwise there is no need to ever get tools to disable them. Perhaps let them go out of focus when about half the damage has been done and let them emit a camera alarm.

Also I am not really clear on if this changes camera alarms from emp?

Gun Hog · Post by **Gun Hog** » Fri Jun 12, 2015 4:06 pm #95274

EMP now causes camera alarms AFTER the camera comes back online. So, feel free to shut down entire departments with a massive EMP wave, and freely slaughter people, as they have no radio, AI has no sight, energy weapons are drained. You will have 90 seconds before the AI gets any warning or visibility in the area, and this timer can be reset by consecutive EMPs without setting off the alarm. Furthermore, you may safety destroy cameras by simply hitting them. Fire off an EMP to shut down the area, then hit all cameras you can reach within 90 seconds to silently take down an area. Remember, headsets are also disabled by EMP, which means that it is now quite easy to silence and murder several people at once.

CPTANT · Post by **CPTANT** » Fri Jun 12, 2015 4:48 pm #95287

Gun Hog wrote:EMP now causes camera alarms AFTER the camera comes back online. So, feel free to shut down entire departments with a massive EMP wave, and freely slaughter people, as they have no radio, AI has no sight, energy weapons are drained. You will have 90 seconds before the AI gets any warning or visibility in the area, and this timer can be reset by consecutive EMPs without setting off the alarm. Furthermore, you may safety destroy cameras by simply hitting them. Fire off an EMP to shut down the area, then hit all cameras you can reach within 90 seconds to silently take down an area. Remember, headsets are also disabled by EMP, which means that it is now quite easy to silence and murder several people at once.

Good, traitors aren't nearly aggressive enough now.