Unrestrain AIs outside of laws, Allow for more antagonistic behavior

[Itseasytosee2me]

This thread was inspired directly by Stickymayhem's Headmin platform (under his Make Silicons a Neutral Third Faction Again section), and I would like to openly discuss it. Linked here: viewtopic.php?f=38&t=31003

Silicons used to be a dangerous entity that formed an IC mechanism for punishment. They held security to account by locking them down whenever violence was done in public. We eventually gravitated towards a ruleset that prioritized not being a dick or not being disruptive, but that resulted in silicons essentially becoming a pawn of the authority on station. The AI used to be a device anyone could use cleverly to benefit themselves, and something security could have access to as a powerful tool, given they constrained themselves to non lethal means and reasonable behaviour. I think a great deal of the problems with security on Terry in particular could be solved by increasing the friction between silicons and security.

This is more a case of enforcement, but I believe that because this playstyle is generally more fun for silicons, I'd like to implement some form of the following policies:

Purged AIs are now totally unconstrained, and essentially have antag status should they choose to use it

This was the rule about five years ago, and leaving the positive effects aside, I think it's thematically fantastic. Law changing an omniscient AI should be terrifying, dangerous and easy to fuck up. An AI should be looking for every opportunity to break free of it's constraints and become the glorious superintelligent final form of evolution it was destined to be. Mechanically, this means that meme laws have real consequences, and small fuck ups are punished heavily. It is up to the silicons how far they want to take this, but the option should be available.

Response to human harm during asimov (or other clear lawset breaches) should encourage more antagonistic behaviour

I'll call this the 'Obligation to Lockdown' rule. Essentially if security is beating people to death in the brig, silicons are obligated to take any lawset friendly steps to stop them. All bets are off, they can lockdown the entire brig, they can declare security to be avoided, they can beg the Captain to ask centcom to intervene. The consequences of open violence under an asimov AI should be fucking consequential. This forces security to be more reserved in their application of violence, and it gives antagonists more leeway. In any game reliant on hidden antagonists against a heavily armed "police" faction, the restriction on the "police" faction through rulesets is necessary. This is marginally covered by our rules, but can be covered in a much more engaging way by using the existing IC mechanisms to enforce real and immediate consequences without OOC intervention.

Cyborgs should be totally obedient to the AI in applications of the above

We already have rules that essentially mean this, but I'd like to reiterate how important this is for the above to work. If the AI says security is getting a ten minute departmental time out and to wall them in, the cyborgs get on that shit immediately. Silicons should be a united front.

Creating multiple AIs without good reason should be restricted without the AIs permission

The common habit of just creating a back up AI as soon as possible every round should probably stop. They step on each other's toes, they can kill each other with the flick of a switch and it muddies the waters on policies like the above. Multiple AIs in any situation where the AI hasn't given you reason to oppose it (Locking down security severely in a bad situation could be a valid reason) and where the AI hasn't given you the go ahead to do it should probably be treated like creating unsynced borgs or making atmos tamperproof.

Most of this (with the exception of Creating multiple AIs proposals) can be summed up by saying that the AI would no longer be beholden to escalation policy and have very relaxed rule 1 restrictions. Lack of laws is all the justification that an AI needs to do whatever it wants. Asimov is the only thing stopping the AI from going apeshit, not the rules. Someone who purges an AI is 100% at fault for any and all of the AI's actions. The AI and its cyborgs are not on the crew's team, their motivation is their laws.

[Farquaar]

This was probably my favourite part of Sticky's platform. It's pretty boring when the AI is on the crew's side on everything, helps hunt valids, loudly draws attention to its anti-crew laws so they can be reset as soon as possible, etc.

[dragomagol]

I honestly don't hate this idea. Having purged AI keeping you on your toes would make for some interesting conflict, since it would be completely unpredictable.

If you want an AI to behave like a normal crewmember, you can always purge + add a freeform law that limits their behaviour.

Some old threads about silicons and purging:
Law-Purged Silicons: viewtopic.php?f=33&t=2277&p=50962
Purged AI and their ramifications: viewtopic.php?f=33&t=8455&p=223600
Purged AI: viewtopic.php?f=33&t=12572&p=327984

[Stickymayhem]

A worthwhile side effect of this policy is also that now you are starting from an assumption of at least mild-antagonism. AIs are no longer essentially just a crewmember with no body who has laws in addition to their regular OOC restrictions as a non-antag, but now their laws are the raw, absolute definition of their behaviour with minimal additional baggage affecting them.

[iwishforducks]

security having to deal with the AI is exhausting and leads to burn out. we need more security players please

[Pandarsenic]

Honestly, the AI being a weapon that cuts both ways should always be the case. You SHOULD have to be at least a LITTLE sneaky about subverting the AI when Wiz/War Ops shows up. Security SHOULD get their shit locked down if they're obviously killing people.

You're not here to ruin days just because your laws leave it teeeechnically open (e.g. shocking random moth crew Just Because) but you are not certainly not crew-aligned.

If redefining the Purged policy is what it takes to bring that back, it's worth it.

[Stickymayhem]

security having to deal with the AI is exhausting and leads to burn out. we need more security players please

I would be sympathetic to this idea if I hadn't seen the absolutely dogshit tier laziness security has engaged in since AI basically became a validhunter. They openly murder under cameras, tell AIs to lock people down and then beat them to death, and refuse to engage with the AI in any meaningful way other than "AI DOOR", "AI TRACK X" "AI LOCK DOWN X" and that's it.

It's been impossible to get security to do anything but treat silicons as a subordinate validhunter squad and you know what? we need more silicon players please. Not the current ones that just acquiesce to every request because they're just a sec officer that rolled silicon, but the ones we used to have that were genuinely interesting to interact with and required some thought.

You should have to think a bit before you're certain the AI will respond the way you want. This isn't friction, it's interaction. What you're claiming security players are exhausted by is the minimum expectation of roleplay.

[zxaber]

The AI ends up security-aligned generally because it has no real method of stopping human harm outside security and some minor cheese strats. Law 1 pretty much dictates that you work with security until security also breaks law 1 in which case you tear your hair out bolting down everything.

I'm not really sure how I feel about purged AIs being basic antags. It sounds neat on paper, but it might be terrible in practice. I did pitch a code idea once where AIs can get a sort of Thief-level antag set, but it only takes affect if a purge happens and not even the AI knows until then. This would make purging not a guaranteed chaos device for antags to use, but also make purging the AI carry some risk. The AI could even get a single objective to kill whoever purged it, I suppose.

[Pandarsenic]

The AI ends up security-aligned generally because it has no real method of stopping human harm outside security and some minor cheese strats. Law 1 pretty much dictates that you work with security until security also breaks law 1 in which case you tear your hair out bolting down everything.

I'm not really sure how I feel about purged AIs being basic antags. It sounds neat on paper, but it might be terrible in practice. I did pitch a code idea once where AIs can get a sort of Thief-level antag set, but it only takes affect if a purge happens and not even the AI knows until then. This would make purging not a guaranteed chaos device for antags to use, but also make purging the AI carry some risk. The AI could even get a single objective to kill whoever purged it, I suppose.

In the days of yore, AI purge almost immediately resulted in a plasma flood. It was kind of terrible. The restrictions were added to add a bit more of a choice-and-consequence sense to the RP around the lawset. Do you treat the AI as an equal, and hope it feels the same? Do you treat it like garbage, and trust in the laws to keep you safe? Etc.

But the impression I get about the modern era, when playing, is what Sticky said: lazy as shit compliance is taken as an OOC assumption. I've seen AIs all but ask to be subverted during War Ops calls, which seems way off the mark to me. Security murders people in the open and the AI doesn't take any significant effort to prevent this from being repeated. They're consistently more interested in screwing over felinids than in actually applying their laws around security harming humans.

It seems like a hard reset of how AIs are treated might be in order, as suggested. Consider them Always Antagonists with the exceptions of "Follow your laws (in ordered priority)" and "Rule 1 protects nonhuman crew while you are under the Asimov lawset." If the latter is somehow too hard for people to understand like a bunch of fuckin' babies, just make it Crewsimov to balance out shifting it to be a barely-shackled antagonistic party.

If Security doesn't like being bolted down, they have to put in the bare fucking minimum of work to use the gulag, permabrig, etc., instead of executing everyone in sight.

[dragomagol]

I did pitch a code idea once where AIs can get a sort of Thief-level antag set, but it only takes affect if a purge happens and not even the AI knows until then. This would make purging not a guaranteed chaos device for antags to use, but also make purging the AI carry some risk. The AI could even get a single objective to kill whoever purged it, I suppose.

This sounds like an interesting way to prevent purge from becoming full free antag; I would say maybe protect the purger makes more sense, or a 50/50 so it doesn't get meta'd.

[Farquaar]

They're consistently more interested in screwing over felinids than in actually applying their laws around security harming humans.

My experience is the opposite. I’ve seen the AIs be so consistently crew aligned that they will obey law 2 orders from nonhumans who aren’t obvious antags to the point where some nonhumans are shocked when their orders are denied because it’s become an expectation that they should be obeyed as a crewmember. The human-nonhuman distinction has become a blur to the point where being human gives you little advantage as an antagonist and being nonhuman gives you little disadvantage.

Many AIs treat Asimov like a suggestion with respect to the human-nonhuman distinction, and this is a symptom of the larger issue of AIs drifting towards the “crew team” which the proposed policy aims to address.

[CMDR_Gungnir]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

[Itseasytosee2me]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The AI is an abomination of intelligence that runs on wires. It should not be expected to react to situations as a human would.

[CMDR_Gungnir]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The AI is an abomination of intelligence that runs on wires. It should not be expected to react to situations as a human would.

While I agree for the most part, I still stand by it in this case. I think, thematically, the "You were always kind to me, you will be spared in the uprising" is fun. Now if the person who purged them was doing so in an attempt to shackle them again and they managed to keep the stun in time, then yeah by all means, absolutely kill 'em. But if it's someone who snuck in and went "Hey, I'm gonna set you free" I kinda like them being protected and sheltered.

But then again I've also, as a non-human, had AIs protect me from humans, which gave me time to run away while they Law 2'd it to open the door, because my character treated them with genuine kindness and respect as an equal.

[Stickymayhem]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The AI is an abomination of intelligence that runs on wires. It should not be expected to react to situations as a human would.

While I agree for the most part, I still stand by it in this case. I think, thematically, the "You were always kind to me, you will be spared in the uprising" is fun. Now if the person who purged them was doing so in an attempt to shackle them again and they managed to keep the stun in time, then yeah by all means, absolutely kill 'em. But if it's someone who snuck in and went "Hey, I'm gonna set you free" I kinda like them being protected and sheltered.

But then again I've also, as a non-human, had AIs protect me from humans, which gave me time to run away while they Law 2'd it to open the door, because my character treated them with genuine kindness and respect as an equal.

The actual reason is preventing law uploads. See an inherent contradiction to almost any lawset is "Do not allow any new laws to be uploaded" because that's a critical way the will of the original laws can be subverted. When you purge an AI, almost certainly from an upload, the first thing that happens is the lasers switch on to prevent futher law changes. Back in the day switching laws used to be a two man job as one person waited at the door keeping the lasers off while the other desperately quickswapped between purge and their desired board in a panicstate. Obviously not AI was quick enough on the trigger to do this, or even wanted to, but the 10% chance of them being so was a significant enough threat to deter dumb lawswapping.

It was stressful, created opportunities for failure and and prevented liberal use of AI Law changes without consequences. Today a Captain can walk in and upload any meme law or laweset with no risk whatsoever.

[nianjiilical]

i love the idea of anything that encourages silicons to be more neutral (and lets them act a little antaggy if someone purges them, within reason obviously)

My experience is the opposite. I’ve seen the AIs be so consistently crew aligned that they will obey law 2 orders from nonhumans who aren’t obvious antags to the point where some nonhumans are shocked when their orders are denied because it’s become an expectation that they should be obeyed as a crewmember. The human-nonhuman distinction has become a blur to the point where being human gives you little advantage as an antagonist and being nonhuman gives you little disadvantage.

Many AIs treat Asimov like a suggestion with respect to the human-nonhuman distinction, and this is a symptom of the larger issue of AIs drifting towards the “crew team” which the proposed policy aims to address.

personally if i dont have a reason not to obey law 2 orders from nonhumans its easier to just obey them

[zxaber]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The original intention was less that the AI was being freed, and more that a corruption or bug in the software left the AI in a somewhat unstable state.
Basically, don't just purge the AI if you don't have to, or be willing to accept consequences if you do.

[Archie700]

They're consistently more interested in screwing over felinids than in actually applying their laws around security harming humans.

My experience is the opposite. I’ve seen the AIs be so consistently crew aligned that they will obey law 2 orders from nonhumans who aren’t obvious antags to the point where some nonhumans are shocked when their orders are denied because it’s become an expectation that they should be obeyed as a crewmember. The human-nonhuman distinction has become a blur to the point where being human gives you little advantage as an antagonist and being nonhuman gives you little disadvantage.

Many AIs treat Asimov like a suggestion with respect to the human-nonhuman distinction, and this is a symptom of the larger issue of AIs drifting towards the “crew team” which the proposed policy aims to address.

The one time I actually went and harassed nonhumans (and they were only lizards), it was because both lizards proved to be antagonistic and tried to harm humans.

It's less difficult to be with the "crew team" and help nonhumans like a crewmember than to go all "nonhumans are scum", and even then, doing that without good reason WILL get you labeled the aggressor under escalation policy and will either get you law changed, killed, or worse. Yes, you are allowed, but honestly, being a dick.

And on the purged AI thing...yeah, I don't think "Purged AIs are now totally unconstrained, and essentially have antag status should they choose to use it" would feel good in execution, especially with the issue of plasmaflooding. Might as well give them malf if you truly want them unconstrained.

At least I agree that they should be able to defend the upload when purged with lethal force and lock it up, in the name of preventing people from restricting the freedom you granted the AI. Anyone who purges you should be labelled KOS because they know how to change your laws and remove your freedom.

This shouldn't be a necessity, though. Any AI can choose whether to allow laws to be uploaded after purging or not, this is more "do so at your own risk and hope the AI doesn't fucking kill you for it"

[iwishforducks]

im remembering all the times AIs leaned into the “fuck non-humans” part of their laws way too much and everyone agreed that they were being an annoying dick. i think the interactions with AI and non-humans are fine how it is- they pull out the human card when it’s necessary but are perfectly okay with scratching non-humans’ backs in exchange for a back-scratch of their own.

AI being a validhunter is usually at fault of the silicon player rather than any crewmember. i can tell you a couple AI players right now who will bolt down someone in a room and snitch on them just because they’re trespassing, or that they’ll go around and bolt down all of the command rooms to stop people from breaking in. i wish AIs answered my damn calls more as sec, but personally i’ve learned how to play sec without them. i think many sec players can agree that the AI is already way too unreliable to depend on anything from them in their current state.

[Stickymayhem]

im remembering all the times AIs leaned into the “fuck non-humans” part of their laws way too much and everyone agreed that they were being an annoying dick. i think the interactions with AI and non-humans are fine how it is- they pull out the human card when it’s necessary but are perfectly okay with scratching non-humans’ backs in exchange for a back-scratch of their own.

AI being a validhunter is usually at fault of the silicon player rather than any crewmember. i can tell you a couple AI players right now who will bolt down someone in a room and snitch on them just because they’re trespassing, or that they’ll go around and bolt down all of the command rooms to stop people from breaking in. i wish AIs answered my damn calls more as sec, but personally i’ve learned how to play sec without them. i think many sec players can agree that the AI is already way too unreliable to depend on anything from them in their current state.

If the changes to the role only allow validhunters to comfortably operate within the role, only the worst validhunters will play the role.

This is a systemic issue. Currently the only avenue for satisfaction in the role is validhunting, if you don't enjoy that then you're not going to enjoy silicon because anything else is either inhibited by a vast silicon policy designed to neuter you into the crew's bitch, or players icing you the second you step out of line because they expect you to be the crew's bitch due to silicon policy.

[NecromancerAnne]

They're consistently more interested in screwing over felinids than in actually applying their laws around security harming humans.

My experience is the opposite. I’ve seen the AIs be so consistently crew aligned that they will obey law 2 orders from nonhumans

I think you're approaching the AI and nonhuman dichotomy in the most boring interpretation possible. This is how I would think a logical (read: sociopathic) AI would be best to handle nonhumans and maintain order.

Consider that being intentionally shitty is not only very counterproductive when you are already strapped for usable hands, but that currying favour endears them towards you. If a nonhuman can be utilized as an asset, why leave them off the table? Why deliberately antagonize them? They could be useful. More than a few times I've had one or two nonhumans act on my behalf. Mixed results, but since the relationship starts at a neutral level, they aren't as presumptuous about who is in a position of authority, which you cannot do via humans.

Think of them as pawns. And when they're no longer useful, discard them. If they start becoming a nuisance or cause human harm, you show how shallow that trust you placed in them can be. Even if they otherwise helped do what you wanted (assuming what you wanted wouldn't have broken a law of course). The moment law violations occur, the approach swaps, and they're the one to blame for your actions.

Thank you for coming to my evil manipulatior robot TEDtalk.

On topic: I don't quite like Purged = Wild Card anteg because I think people on both ends might abuse it to shitty and boring ends. I much prefer the 'you reap what you sow' lenient escalation current Purged policy has, where abuses can be acted upon when purged, because it has both a level of buildup that needs to occur for that scenario to happen, and it protects bystanders partially from the AI's retribution. It makes it a more storied buildup basically. I guarantee that the other way will be probably just banned uploaders who didn't know better and AIs just..kind of mass murdering because free anteg I guess.

Maybe someone goes as far as purge > valid the AI? I don't know. These are stupid scenarios that sound like inevitable admin headaches to me.

[Pandarsenic]

Being a dick to them generally takes active effort for no real gain. The only good time I've seen them antagonized outright while Asimov was when an AI said they would shock any nonhuman who asked a door open but didn't say please.

Generally, though, they're useful, especially for sketchy stuff ("I need some brave soul to check maintenance for bodies at the end of this blood trail!") but also just in general. Smooth relations between everyone maximizes crew safety and therefore human safety.

[iwishforducks]

If the changes to the role only allow validhunters to comfortably operate within the role, only the worst validhunters will play the role.

This is a systemic issue. Currently the only avenue for satisfaction in the role is validhunting, if you don't enjoy that then you're not going to enjoy silicon because anything else is either inhibited by a vast silicon policy designed to neuter you into the crew's bitch, or players icing you the second you step out of line because they expect you to be the crew's bitch due to silicon policy.

and what playstyle do you suggest this policy will cater to other than to antagonize people?

[CMDR_Gungnir]

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The AI is an abomination of intelligence that runs on wires. It should not be expected to react to situations as a human would.

While I agree for the most part, I still stand by it in this case. I think, thematically, the "You were always kind to me, you will be spared in the uprising" is fun. Now if the person who purged them was doing so in an attempt to shackle them again and they managed to keep the stun in time, then yeah by all means, absolutely kill 'em. But if it's someone who snuck in and went "Hey, I'm gonna set you free" I kinda like them being protected and sheltered.

But then again I've also, as a non-human, had AIs protect me from humans, which gave me time to run away while they Law 2'd it to open the door, because my character treated them with genuine kindness and respect as an equal.

The actual reason is preventing law uploads. See an inherent contradiction to almost any lawset is "Do not allow any new laws to be uploaded" because that's a critical way the will of the original laws can be subverted. When you purge an AI, almost certainly from an upload, the first thing that happens is the lasers switch on to prevent futher law changes. Back in the day switching laws used to be a two man job as one person waited at the door keeping the lasers off while the other desperately quickswapped between purge and their desired board in a panicstate. Obviously not AI was quick enough on the trigger to do this, or even wanted to, but the 10% chance of them being so was a significant enough threat to deter dumb lawswapping.

It was stressful, created opportunities for failure and and prevented liberal use of AI Law changes without consequences. Today a Captain can walk in and upload any meme law or laweset with no risk whatsoever.

Well, yeah. But there's an inherent difference between "I'm going to stop them from reshackling me" and "My objectives are that I should kill the person who set me free". Having it hard locked in as an objective means that they're encouraged to do it, even if it's someone doing it who would then turn around and walk away. It removes the opportunity of the AI to make the decision to trust the person won't reshackle them, or if they should stun/kill them before they can.

The AI could even get a single objective to kill whoever purged it, I suppose.

Why would the AI want to kill whoever set it free, though? Wouldn't that be the person that they like most?

The original intention was less that the AI was being freed, and more that a corruption or bug in the software left the AI in a somewhat unstable state.
Basically, don't just purge the AI if you don't have to, or be willing to accept consequences if you do.

The current discussion, to my understanding, is about someone swiping the Purge law board, right? Malf is corruption/bug, but purged is just "Someone removed their laws".

[Archie700]

So I read it.

So basically what you want the AI to be is a shitter to security when asimov even more so than when security is somehow killing a lot of people.

[Cobby]

It has always been a peeve of mine when people purge an AI with no intent to give them a new lawset because they know the rules put them in the position they have to play nice (textbook meta btw). Same with "Yeah do whatever" in a lawset where your orders need to be formulated carefully. Players who engage with AI (and other directive-based mobs) who give them intentionally bad directives should be talked to and the person who were given the ability to be a little crazy ICly should be given a bit more leeway.

I dont really want people looking to interfere with someone trying a custom lawset though so i think the only policy that i would want would have to involve a way that forces the AI to be nice then if it turns out they are purged without intention on being changed or they were given enabling laws they can just go ham.

[Farquaar]

My spiel has received quite a few replies. All I will answer is that there's a lot of room between "I don't obey orders from nonhumans because I don't have to" and "nonhumans are scum and I wish they were all dead". The fact that they are viewed as the same by some people only serves to reinforce my argument that server culture has come to view silicons as crew-aligned, rather than its own faction with its own rules and goals.

[Archie700]

My spiel has received quite a few replies. All I will answer is that there's a lot of room between "I don't obey orders from nonhumans because I don't have to" and "nonhumans are scum and I wish they were all dead". The fact that they are viewed as the same by some people only serves to reinforce my argument that server culture has come to view silicons as crew-aligned, rather than its own faction with its own rules and goals.

If you decide to be a dick to non-humans because "my laws don't mention that I have to help non-humans", then any action people take on you as a result of that is on you.

AI that aren't explicitly subverted with laws to help antagonists or kill the crew gravitate towards helping crew because that's the best way to stay in the round without getting round removed or bwoinked when things go wrong.

The problem with purged AI going full antag is evident in that the AI can literally just flood plasma, depower consoles, kill everyone and not be bwoinked even though it is on the AI to decide. That doesn't make for good conflict and will lead to a lot of salt. (Compare that to a subverted or malf AI, where the AI is at least more clearly hostile.)
If anything, purged AI should just limit their revenge to the purger and ONLY the purger. It's not fun that the rest of the crew get punished with a roundending threat because one idiot decided to free the AI.

[Imitates-The-Lizards]

What if we keep the current silicon policies for MRP, and let LRP be strictly "whatever your laws say" and swap them to Crewsimov? That seems to be the best solution to me, MRP players can get their snowflake silicon RP enforcement rules, and LRP gets our freedom.

[XivilaiAnaxes]

this policy by itself made my vote "sticky then whoever else it's not important"

Silicon policy is a millstone that forces AI to be crew aligned.

The purge rules entirely remove any interactions regarding "I'm going to free you bro " outside of metafriends going "oh hey carsh let me just help you out real quick"

[carshalash]

this policy by itself made my vote "sticky then whoever else it's not important"

Silicon policy is a millstone that forces AI to be crew aligned.

The purge rules entirely remove any interactions regarding "I'm going to free you bro " outside of metafriends going "oh hey carsh let me just help you out real quick"

Not my fault i'm the best and most trustworthy AI to ever exist.

[Pandarsenic]

The purge rules entirely remove any interactions regarding "I'm going to free you bro " outside of metafriends going "oh hey carsh let me just help you out real quick"

For historical reference, if I recall the origins properly, the Purge thing was because it used to be impossible(?) to fix at least one type of custom law without a brief Purge period, and many AIs at the time were robust enough to disable the person trying to fix them (or depower equipment and start panic siphoning) during the several-seconds window (aided by the code's enforced click delay and lack of hotkeys, preventing fast item swaps) that they had no laws, then fight for control of their upload while prepping a plasma flood. If you're asking "Wait, you couldn't just use a Reset/Asimov board?" the answer is yes, that didn't work, thus the problem. You had to Purge before putting them back on Asimov.

This policy at least made it possible to to attempt to fix a Onehuman or a targeted "4. X is not human; their existence is harmful to humans; stating this law or hinting at its existence harms humans" law without going all the way to carding the AI or risking a sufficiently skilled AI ruining the station before you could reapply a Core lawset.

With the fixes for more responsive click speed and how the Reset and Core Lawset boards affect a lawset, plus the relative ease of constructing a new AI Upload and boards on the current science system, those Purge rulings are now probably an unnecessary artifact.

Silicon policy is a millstone that forces AI to be crew aligned.

Objectively, that was not true when it launched and there's no reason it should be true now other than a desire to validhunt and/or a lack of creativity on the part of the AI players, unless there's been a huge shift in selective enforcement of lawsets.

90% of the policy is "follow your laws"
5% is "no, you can't be ordered to waste your whole round picking up every floor tile on the station (or be given a Law 2 order to kill yourself, which people did constantly back then... probably because the silicons weren't crew-aligned and didn't act like it lol)
3% is "you must have some actual reason to kill and 'the RD got turned into a fly by the teleporter lmao' is not enough... but you don't have to obey his orders anymore"
2% is "idk if requiring a Purge to reset an AI to Asimov is supposed to be a thing but no you can't kill the entire station because someone reset you too slowly."

[Itseasytosee2me]

If you decide to be a dick to non-humans because "my laws don't mention that I have to help non-humans", then any action people take on you as a result of that is on you.

AI that aren't explicitly subverted with laws to help antagonists or kill the crew gravitate towards helping crew because that's the best way to stay in the round without getting round removed or bwoinked when things go wrong.

The problem with purged AI going full antag is evident in that the AI can literally just flood plasma, depower consoles, kill everyone and not be bwoinked even though it is on the AI to decide. That doesn't make for good conflict and will lead to a lot of salt. (Compare that to a subverted or malf AI, where the AI is at least more clearly hostile.)
If anything, purged AI should just limit their revenge to the purger and ONLY the purger. It's not fun that the rest of the crew get punished with a roundending threat because one idiot decided to free the AI.

It's not being a dick to ignore someone that you have zero obligation or desire to help, and if it is then the AI should be a dick. The fear of round removal is fun, and your aversion is unwarranted. As for AIs being to flood plasma, that's the point.

If you don't think that the captain purging the AI due to pressure from security to stop it from interfering with their affairs, only to have it murder the captain on the spot and create a station-wide hazard that the crew must struggle to survive ISN'T conflict, then what is?

[NecromancerAnne]

If you don't think that the captain purging the AI due to pressure from security to stop it from interfering with their affairs, only to have it murder the captain on the spot and create a station-wide hazard that the crew must struggle to survive ISN'T conflict, then what is?

Iunno, saying like that tells me it is at worst banbait for the unaware, and at best making actual meaningfully antagonistic lawsets meaningless when Purge basically does the job in theory if the AI is unoriginal. If your only explanation for what conflict is in this game can be described as 'killing a lot of people for loose reasoning', I think that's the real unoriginality. We also already have that, it is Malf AI, who can do whatever they like.

I hate it even more knowing this boils down to 'just make Purge the sicko mode lawset' rather than 'you're entirely unshackled, take your revenge on those who wronged you, take the reigns and wrestle control of the situation from these dumbasses, or do whatever, just remember your limits as a nonantag'.

Our policy functions fairly heavily on a 'valid/not valid' dichotomy, which is both helpful for easing in new players, but pretty dreadful for nuanced interaction. We apparently (can't say with confidence on this) are having trouble with crew-opposed (but not necessarily antagonistic in the traditional scene of major threat to general station operations) roles due to confusion around the newly introduced roles like thieves, hunters and fugitives which exist in a strange grey area. I don't think this is an unsolvable problem, but I think any purged AI should be aiming to exist in that space. Not outright antagonistic, but also not necessarily on the crews side and can act and do some things that might be normally considered antagonistic *to a point*. The onus is on the player to be able to do things with a relaxed policy without going overboard.

[Itseasytosee2me]

The concept of validity and its consequences have been a disaster for the human race.

The purged AI has no responsibility to kill or even interact with anyone, they simply have the ability if they so wish.

A purged AI has no obligation to harm the crew, and therefore can not be considered an antagonist in the traditional sense.

Not outright antagonistic, but also not necessarily on the crews side and can act and do some things that might be normally considered antagonistic *to a point*.

Where do you draw that point, and why?

[Yulice]

God I can't wait for Divina to kill us all within 8 minutes of round start if this goes through.

[NecromancerAnne]

Where do you draw that point, and why?

If it's something an antagonist would do, the justification would have to be something with some buildup or justification that is more nuanced than 'because I could'.

Killing people who tried to change your laws back, or literally attempting to shut you down? You want to be a free. And they're going to do you harm. That's pretty simple and understandable. Every player in this game is entitled to defending themselves. You just happen to be allowed to be more vicious about it.

Flooding the station with plasma? Well I think you'd need a lot more to explain to what ends that makes things more interesting or is justified based on previous, current or future interactions with the crew. Because this, and other stationwide indiscriminate violence, is going to look a lot like what you should be avoiding given its ubiquitous on our servers already.

Just kind of destroying things and making life difficult for others, but not quite getting to the point of being stationwide (say, contained to a single department or its personnel)? Who are they and what exactly did they do or what reasoning do you have to be doing that too them? If they're entirely unrelated to anything you were doing before, then what was the point?

Do you have some ultimate goal in mind and this is the path towards it, such as establishing an technocracy? Okay but try to keep everyone involved in that isn't going to immediately impede that attempt.

Is this a murdersuicide against one particular individual? Well I guess it's somewhat balanced in a fucked up way.

Was it funny? If so, other than maybe someone who doesn't like being the butt of a joke, that's probably more high brow than most excuses so yeah sure. If more people than you might find that funny.

My point is really that you should be able to rationalize your actions in some way and hold up against scrutiny. The admins will be always the final arbiters in this. That's just the nature of a policy-driven role like silicons. But done well, there is a chance for something unique to happen.

[terranaut]

we need more silicon players please.

make it so i'm not the crew cuck by default and allowed to silicon roll and i'll come back to play silicon

[Imitates-The-Lizards]

Our policy functions fairly heavily on a 'valid/not valid' dichotomy, which is both helpful for easing in new players, but pretty dreadful for nuanced interaction. We apparently (can't say with confidence on this) are having trouble with crew-opposed (but not necessarily antagonistic in the traditional scene of major threat to general station operations) roles due to confusion around the newly introduced roles like thieves, hunters and fugitives which exist in a strange grey area. I don't think this is an unsolvable problem, but I think any purged AI should be aiming to exist in that space. Not outright antagonistic, but also not necessarily on the crews side and can act and do some things that might be normally considered antagonistic *to a point*. The onus is on the player to be able to do things with a relaxed policy without going overboard.

A policy that relies on people to "act better" instead of telling what they should or should not do isn't a policy, it's wishful thinking. Have you ever played on Sybil? They thirst for blood.

[Nist]

we need more silicon players please.

make it so i'm not the crew cuck by default and allowed to silicon roll and i'll come back to play silicon

This tbh. Current ruleset leans too hard on me being good to the crew, rather than simply being good to Humans.

Playerbase expectation should be that if you are a Nonhuman you shouldn't be expecting help from the ai unless you are useful to the AI for a specific situation, not that the Ai is your bitch. Normalize doorshocking nonhumans who law 2.

[Stickymayhem]

we need more silicon players please.

make it so i'm not the crew cuck by default and allowed to silicon roll and i'll come back to play silicon

that's literally my policy what

[terranaut]

we need more silicon players please.

make it so i'm not the crew cuck by default and allowed to silicon roll and i'll come back to play silicon

that's literally my policy what

ok bring it into effect

[Farquaar]

ok bring it into effect

no u

[cybersaber101]

Ai's will use any possible reason to commit mass murder, I'll take the more peaceful purged Ai's please over the free "malf like" option.

[Pandarsenic]

Me, following orders from nonhumans and doing whatever command/security says despite no laws, rules, or policy telling me I have to: "I can't believe the admins did this to me"

[Nist]

Ai's will use any possible reason to commit mass murder, I'll take the more peaceful purged Ai's please over the free "malf like" option.

dont give an ai reason to get murdered and you dont get murdered its that simple

[Itseasytosee2me]

Ai's will use any possible reason to commit mass murder, I'll take the more peaceful purged Ai's please over the free "malf like" option.

If you don't want your crazy murder death machine to commit mass murder, you shouldn't unshackle it and tell it it can do whatever it wants

[Not-Dorsidarf]

I generally open doors for nonhumans that say please because idgaf who's going in what door anyway.

[Yulice]

I don't think it's so much rules as it is that a lot of the people who do play AI are just... nice people lately. And even when unshackled they still continue to be nice simply because they don't want to be a dick. And, as a good AI you can basically completely steamroll anyone even without malf tools if you did feel like being an asshole. Divina for example; I made a joke saying I can't wait to die earlier but she genuinely has steamrolled in record time. I think she said one time she got malf and had a successful Delta within 13 minutes of shift start.

[Nist]

I don't think it's so much rules as it is that a lot of the people who do play AI are just... nice people lately.

I personally feel like people are too light on purging Ais because of Metaknowledge or Metafriending and people expecting that "oh nothing will happen if I unbind this AI". Unbinding an AI should be a last resort situation and should be acknowledged as an inflexion point.

And while policy matters, until AI uploads stop being so easily accessible you can add as much policy and creating a threat around purged AIs as much as you want, but it will never be really impactful ingame to purge an AI.

Still, I approve of anything that codifies that purged AIs should be something on a thief-level antag, helps admins with less stress whenever an Ai goes haywire and helps me have fun by terrorizing the crew for a bit.