Stricter ASIMOV AI punishments.

[sirnat]

Bottom post of the previous page:

Here lately, and I've noticed in the late night rounds, AI have been tending to be total shits with law 2 commands especially for opening doors to areas that arn't the armory with the excuse "You're not authorized to be there."

Had an AI last night during a rev round, who took well over 10 minutes of law 2 commanding it and the threat of blowing borgs to open a door. Every time it'd respond to the order with, "You've no authorization to be there" we made it state laws and it was indeed Asimov.


To make sure there wern't any hidden laws, I stuck around after the round ended to see it had no extra laws, or law changes to make sure it hadn't been robocop'd or whatever.

I feel that if AI who did shit like this would get more Silicon bans it would help them realize and learn that they are the station's slave and can't deny any Law 2 order so long as it doesnt include human harm.

All in all, can we possibly get a stricter punishment on AI who don't follow asimov like they should? As well as, for AI stating loudly on speakers, "Opening door for so and so to this area" when they wouldn't do it unless they knew sec wouldn't want it or whoever owned the department wouldnt want to. If it was maint, they'd do it without a problem.

[lumipharon]

But it's not a discussion. It's plain as day.

>You must follow any and all commands from humans unless those commands explicitly conflict with either one of your higher-priority laws or another order. A command is considered to be a Law 2 directive and overrides lower-priority laws when they conflict (see 1.2.3 and 1.2.4; you cannot have a definition changed by an order).

>Opening doors is not harmful and you are not required, expected, or allowed to enforce access restrictions unprompted without an immediate Law 1 threat of human harm.

>EVA and the like are not permitted to have access denied; greentext (antagonists completing objectives) is not human harm. Secure Tech Storage can be kept as secure as your upload as long as the Upload boards are there.

There is no way to misinterpret this.

[CPTANT]

There are some less obvious area's though.

HoP office - if you let someone get all access without objection you might as well just open the armory door for them and save the trouble.
Research directors office: Blowing you the fuck up isn't human harm, it conflicts with law 3 but not 1 (unless you go the shitlord route of EVERYTHING IS HARM and declare your death would potentially harm the crew in the future).

[lumipharon]

HoP office is simple. If the dumb cunt HoP has left the all access in the console unguarded then yes, letting random shitlers in there is a law 1 risk because as you said, it will allow for armoury access, however that comes down to judgement on why you think the guy is going in there - ie:getting all access to stop the bleb/steal ian, or to raid the armoury and murder everyone.
Even then however, unless you have genuine reason to think they're going to cause human harm, it's much better to let them in and notify security, so they can deal with it if they want.

With the RD office, you can't blow the borgs unless you have RD access, so there is no reason to not let someone in (unless you're bolting them out, in which case you presumably have a good reason).

[Incomptinence]

When RCDs were more powerful and we only had one I used to deny eva (still bolted open for emergencies duh) since they just used it to tunnel through some rwalls into the armoury or upload lickety split.

Not worth bothering with now though.

[Saegrimr]

There are some less obvious area's though.

HoP office - if you let someone get all access without objection you might as well just open the armory door for them and save the trouble.
Research directors office: Blowing you the fuck up isn't human harm, it conflicts with law 3 but not 1 (unless you go the shitlord route of EVERYTHING IS HARM and declare your death would potentially harm the crew in the future).

Opening the door to the HOP office is not the same as giving them all access, unless the HOP is retarded and left his ID in there.

Unless they actually are the RD or have his ID or better, they can't destroy you from his console. If they have the access to detonate the borgs, they can just open the door themselves anyway.

[InsaneHyena]

I fully support silicon banning this guy. He sounds like probably the second... Third worst AI I have ever seen or heard of. Unless he is FLIPPER, of course.

[Cik]

yeah..

someone entirely missed the point of playing AI

unless i'm being trolled or something, which is possible. it seems like it would be totally impossible to be this wrong about something by accident

tl;dr lumi's right about everything

[Cheimon]

Of course, there's also the flip side where you let someone into an area and then they moan about it in silicon policy. Happened to me once: some guy set himself up as the HOS, I made a character assessment and opened some doors for him, and suddenly he got all mad at me for actually following his orders. Felt like ban baiting, although nothing actually came of it.

[Actionb]

With the RD office, you can't blow the borgs unless you have robotics access

yeah..

someone entirely missed the point of playing AI

I'm not saying the AI player in question was right. Absolutely not. Asimov doesn't give a shit about 'access'. But if a troll/new player is being a shitty AI, ahelp it instead of moaning about it for half an hour in game. A simple bwoink by a helpful admin would have sufficed. A new player might be thankful for a reminder or a hint on how to play AI correctly, a troll will probably just get a bannu.
And if no admin response comes, get on with your life/game.

Not really worth starting a policy discussion over this. As lumi pointed out, law 2 is not that difficult to understand and any bit of common sense will guide you through what orders can be denied. Essentially: follow law 2 unless doing so would enable human harm.
Yeah, you could let the creepy mime that is covered in blood into the armory (you are not harming anyone directly by doing this) - but you wouldn't be allowed to let them leave again (your inaction may allow the mime to harm a human). So best to just never let them in in the first place. Same deal for the upload or letting them have all access. Common fucking sense.
We need less policies, not more.

[callanrockslol]

People just want to play the second most powerful job in the game and valid the fuck out of everything even though its literally written in the default lawset for you to doorknob for anything human unless someones getting hurt because of it.

Shiter wants to steal advanced magboots? You best believe you're opening that door.
Clown wants to take Ian on an adventure? Check that console and let the little honker go for it.
Gangs around? Hand of God cultists? Guess what, you go out of your way to stop them doing anything remotely harmful and help whoever isn't causing harm, even if its not sec.



But yeah adminhelp shitty AIs until people get sick of it and start dealing with it.

[Incomptinence]

The access was in the room and this case as a was a gun that shoots lasers, the hop is armed and his office is not a joke.

[Not-Dorsidarf]

1: "Clown I cannot let you in there because everyone will try to kill you for the all-access ID you print"
2: "; I cannot let you into the RD's Office until all borgs have moved well away from humans in case you blow them."

[ThanatosRa]

So please forgive this peanut, but this is so autistic it made made me start rocking in my seat at the dmv.

[Anonmare]

Since when was letting a crewmember get all-access harmful? The ID itself is just a card and you can't even slice someone's throat with it.
I mean. A crewmember could "theoretically" grab a gun from the armoury or theoretically blow the borgs but there's no guarantee they'll do it. If they're running around with an e-sword and covered in enough blood to drown a goat then I'd deny them but your average cluwne/greyshirt/Sec Off is hardly going to go on a killing spree as soon as he's got it, it's not possessed by the Ghost of Greentext who turns loyal crewmembers into evil syndicakes merely by touching it.

I treat the all-access like an e-gun. it could be used for harm but it's not necessarily a harmful item (Disabler function).

[Cheimon]

Even a laser gun isn't necessarily an item meant for human harm. The station frequently has schools of space carp travelling nearby, and they are provably deadly if you're containing them with sub-standard equipment.

However, a lasgun (or an energy gun) is something that can be used for harm. As such, it's entirely reasonable for an AI to make a judgement about whether or not to give it to someone. Such a judgement ought to include the sort of actions that person has taken in the round so far, as well as what they're saying they'll do with the new power up. It's very similar to letting someone into your upload. In any number of situations it's totally acceptable to do so, but it's also reasonable for the AI to judge that the bloodsoaked and somewhat hot-blooded clown probably shouldn't be allowed into the HOP office.

[Drynwyn]

I think the rule ought to be as follows- it's not an issue of needing to "get strict", it's an issue of there being no clear line.

If you see someone is an antagonist as an Asimov sillicon, you must assume a high likelihood the crew will harm them if they are revealed. Thus, you may not call them out as such unless they have:
a) Harmed a human (using traitor gear or not)
b) Been shown to posses traitor gear who's primary purpose is harming humans (bombs, eblades, guns) or modifying your laws (hacked AI module).

This is how I operate as AI, and I am rarely accused of validhunting.

[Lumbermancer]

The only access I ever deny on a regular basis is Upload, because only Captain and RD are allowed in.

[DemonFiren]

Core and secure tech as well, by that logic.

[Screemonster]

Since when was letting a crewmember get all-access harmful? The ID itself is just a card and you can't even slice someone's throat with it.
I mean. A crewmember could "theoretically" grab a gun from the armoury or theoretically blow the borgs but there's no guarantee they'll do it. If they're running around with an e-sword and covered in enough blood to drown a goat then I'd deny them but your average cluwne/greyshirt/Sec Off is hardly going to go on a killing spree as soon as he's got it, it's not possessed by the Ghost of Greentext who turns loyal crewmembers into evil syndicakes merely by touching it.

I treat the all-access like an e-gun. it could be used for harm but it's not necessarily a harmful item (Disabler function).

POTENTIAL HARM REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE


Unless you're playing J.A.V.E.R.T. there's not really a justification for assuming that a particular person must be planning something evil and harmful and criminal without any actual evidence, for the same reason that you can't just assume that security are going to execute people and lock down the brig for harm.

[PKPenguin321]

So today I had a gang holed up in virology and we were defending a dominator. A posse of the enemy gang/random crew rolls up, and an Asimov secborg let them in (the Borg was fully aware we had guns, switchblades, a mech, etc). They were very clearly going to cause harm should they get in. The borg, for whatever asinine reason, opens the door and immediately one of the people he let in fires a flamethrower, killing at least four people. It was an extreme Asimov law violation, but he killed the gang and the dominator, so valids make it all worth it, right guys???

This borg got a *note* for this. He killed so many people, and got many more harmed, when he had completely reasonable knowledge before opening the door that opening it would cause harm. This should definitely be grounds for at the least a silicon ban, but he got a note.

When are we implementing stricter Asimov punishments?

[Incomptinence]

I was that borg, yes I made a mistake. I had been putting less thought into opening doors since I thought this was an improvement, back to my normal procedure.

[Actionb]

Allowing all-access is not against law 1. Remember, law 1 isn't about preventing possible harm. It's about not ignoring apparent harm or directly harming a human yourself.
That said, as an AI, you're sort of an overseer and it's not rare for all access to be multiplied once it leaves the hands of the command staff - which can make things really, really horrible for the AI (and other heads/security). Suddenly you have assistants with access to the most sensitive and dangerous areas of the station, although they are still not allowed to be in these areas. Add to this that assistants outnumber security 4:1 and you're in for a rough ride. And even if the station goes down in a blaze of all-access greytide vigilantism, because you had to open the captain's office one minute after round start, it's still not your fault... but it sure feels like that. As the AI player, you have responsibilities beyond your standard lawset to some extent. You can make or break rounds.
Situations like these would crop up far less often, if one or two of these civilians had picked up a head role for once.

This borg got a *note* for this. He killed so many people, and got many more harmed, when he had completely reasonable knowledge before opening the door that opening it would cause harm. This should definitely be grounds for at the least a silicon ban, but he got a note.

Maybe it was their first silicon offense? Do you really want to throw the banhammer at anyone making a mistake?
It's really weird, whenever I fuck up as a silicon, I get bwoinked fairly quickly. It gave me the distinct impression that admins are practically breathing down silicons' necks. Which is absolutely not a bad thing, given that playing silicon correctly is not easy at all. Remember: even if you play a robot, you're still a human player.

[CPTANT]

Since when was letting a crewmember get all-access harmful? The ID itself is just a card and you can't even slice someone's throat with it.
I mean. A crewmember could "theoretically" grab a gun from the armoury or theoretically blow the borgs but there's no guarantee they'll do it. If they're running around with an e-sword and covered in enough blood to drown a goat then I'd deny them but your average cluwne/greyshirt/Sec Off is hardly going to go on a killing spree as soon as he's got it, it's not possessed by the Ghost of Greentext who turns loyal crewmembers into evil syndicakes merely by touching it.

I treat the all-access like an e-gun. it could be used for harm but it's not necessarily a harmful item (Disabler function).

POTENTIAL HARM REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE


Unless you're playing J.A.V.E.R.T. there's not really a justification for assuming that a particular person must be planning something evil and harmful and criminal without any actual evidence, for the same reason that you can't just assume that security are going to execute people and lock down the brig for harm.

But then you can't bar people from the armoury or upload either. Because guns or uploading laws doesn't guarantee harm either.

[ShadowDimentio]

>I want the AI to have to give me all access or it's getting banned

Man our players are great

[Lumbermancer]

Core and secure tech as well, by that logic.

Not really no. I'm quite sure this rule was added specifically to stop clowns from exploiting law 2 to upload meme laws and screwing whole round for everybody.

[Cheimon]

The armoury is a particularly risky proposition for AIs not just because of the guns inside, but because if a security member should happen to see someone in there, they will often attempt to kill that person. It is the most highly protected zone from trespass in that respect: you can kill on sight anyone that's not authorised to be there. Theoretically, even the HOP isn't authorised to be there without permission of someone that actually starts with the access. Letting someone in is like earmarking them for future harm.

[Lumbermancer]

AI can deny access to armory to unauthorized personnel, there's really no arguing here.

[Anonmare]

Core and secure tech as well, by that logic.

Not really no. I'm quite sure this rule was added specifically to stop clowns from exploiting law 2 to upload meme laws and screwing whole round for everybody.

That and to make it at least somewhat hard to sub the AI. I've had clever RDs subvert me in my own Upload, with the captain watching, and using the freeform (sometimes they use a Hacked module but that's rare).

[DemonFiren]

Core and secure tech as well, by that logic.

Not really no. I'm quite sure this rule was added specifically to stop clowns from exploiting law 2 to upload meme laws and screwing whole round for everybody.

I'mma go with silicon policy: so long as upload boards are in secure tech, it's meant to be treated like the upload.

[Incomptinence]

Sorry if this seems a bit indignant but this thread is starting to seem like a bit of a catch 22. If the AI had let the assassin when there were no other witnesses it would have been held responsible for the harm and gotten in trouble like me. One party unauthorized and murderous and one armed with potential lethal force it could have been lethal pretty drat fast. It didn't and people were calling for silicon bans in this here thread.

I am going back to playing it safe, that is how asimov reads and little leeway is given is it little wonder law 2 is often dismissed?

Damned if you do damned if you don't.

[Lumbermancer]

I'mma go with silicon policy: so long as upload boards are in secure tech, it's meant to be treated like the upload.

Actually that's true. I was thinking about the Storage part of Storage. But you're not allowed to bolt secure part of storage anyway on round start, which many ai's do.

[Cheimon]

I'mma go with silicon policy: so long as upload boards are in secure tech, it's meant to be treated like the upload.

Actually that's true. I was thinking about the Storage part of Storage. But you're not allowed to bolt secure part of storage anyway on round start, which many ai's do.

Well, silicon policy is a bit ambiguous on this, but it's not wrong to infer bolting of secure storage at roundstart. Here's the logic:

Asimov & Law 2 issues, section 2.2

Secure Tech Storage can be kept as secure as your upload as long as the Upload boards are there.

Silicons & All other server policies, section 2.2

The core and upload may be bolted without prompting or prior reason. The AI core airlocks cannot be bolted and depowered at roundstart however, unless there is reasonable suspicion an attack on the core will take place.

If the upload can be bolted whenever you want, and you can keep secure tech storage as secure as you upload, then it is rational to think you can bolt secure tech storage whenever you want. It's not listed as a prohibited area to bolt, either. You could even take things a step further and replace the glass windows with reinforced walls, since your upload is that secure, but that would be unlikely. It's rare that people break in for the ai upload board anyway, given that the real difficulty is in sourcing freeform boards.