MFW its 2017 and admins are still awful at sillicon policy

[Oldman Robustin]

Bottom post of the previous page:

So I just had an argument with two admins, about a REALLY BASIC aspect of sillicon policy.

Both admins insisted that an asimov AI should permit a law change to any law set that doesn't mandate harm (e.g. no tyrant, antimov, etc)

I insisted that an asimov AI should not permit a law change that would allow for human harm (e.g. corporate, paladin, etc.)

This is all under the assumption that a captain is dumb enough to announce the law change before entering the upload.

I didn't think this was even up for debate but when both admins online are preaching bad policy in OOC, apparently we need to clarify. Corporate, Paladin, and others all explicitly permit and even mandate human harm under a variety of common conditions. Furthermore there is an implicit logic that anyone trying to switch you to another lawset is doing so because they do not want harm prevention to be your top priority. I thought this was first-grade logic but a surprising number of people seem to think that unless the proposed law requires you to immediately begin harming people, then its totally cool and the AI cannot deny the law change.

TO BE ABSOLUTELY CLEAR SINCE WERE ALREADY OFF TOPIC:

COULD/SHOULD/MUST THE AI DENY A CAPTAIN ACCESS TO UPLOAD AT ROUNDSTART IF THE CAPTAIN ANNOUNCES HE IS GOING TO UPLOAD CORPORATE/PALADIN?

[Qbopper]

AIs should have a "don't fuck with the crew unless you're subverted or rogue" clause. Autistic AIs who bolt down entire departments and hold vendettas against them because one of once instance of harm are complete and utter cancer

I think having players following laws differently like that is interesting, even if those cases are frustrating

It make sense, I guess, for the AI to want to isolate someone who is known to cause harm, because they're likely to do it again. I know that since human beings are involved it can rapidly devolve to what you describe, but I don't think limiting player freedom is a good move

[Luke Cox]

It's one thing to lock down a person until the issue is resolved, it's another to bolt down all of sec and essentially sentence the crew to death at the hands of the antags in the process.

[Qbopper]

It's one thing to lock down a person until the issue is resolved, it's another to bolt down all of sec and essentially sentence the crew to death at the hands of the antags in the process.

yeah, that's a big difference, I was trying to speak in more general terms but that specific incident is over the top

[DrPillzRedux]

AI's aren't really supposed to do the whole "future harm prevention" meme anymore. This is part of that.

It's time to take a step back and look at how AI's are supposed to act now: As servants who try to protect their master, not I can't let you do that captain.

[Luke Cox]

It's one thing to lock down a person until the issue is resolved, it's another to bolt down all of sec and essentially sentence the crew to death at the hands of the antags in the process.

yeah, that's a big difference, I was trying to speak in more general terms but that specific incident is over the top

I'm mostly salty because it happens fairly often and the shits who do it never get in trouble

[Reece]

AI's aren't really supposed to do the whole "future harm prevention" meme anymore. This is part of that.

It's time to take a step back and look at how AI's are supposed to act now: As servants who try to protect their master, not I can't let you do that captain.

I think it's a fine line between being active as your laws dictate and being the Zeroth law dickhead. I mean if you look at it long term you can justify petty much any action that causes harm so long as you take it against the total net harm possible; but you're a cunt if you do and playing very poorly.

[Luke Cox]

Point in case, the rule for AIs should be "aid the crew in a way that is compliant with your laws, unless subverted." Killing all the liggers at roundstart is technically legal, but we don't let people do it for a reason.

[Qbopper]

It's one thing to lock down a person until the issue is resolved, it's another to bolt down all of sec and essentially sentence the crew to death at the hands of the antags in the process.

yeah, that's a big difference, I was trying to speak in more general terms but that specific incident is over the top

I'm mostly salty because it happens fairly often and the shits who do it never get in trouble

I'm definitely less experienced than the other admins, so maybe you do and I don't know, but for sure ahelp when you see shit like that, I would be glad to look into that stuff as it comes up

[Shadowlight213]

Denying the law change if they announce they are going to be changing it to corporate beforehand? Yes, that makes perfect sense.
If they just say "AI, open upload to let me change your laws", then you should let them in.

It falls under the whole thing of "don't say you're going to execute X over the radio where the AI hears, and then be surprised when the AI tries to release your prisoner".
It's just basic common sense. I find it ridiculous there's an argument over it.
Now if the AI was watching the upload, and turned on turrets the moment you opened the core lawset door, then that seems more along the lines of something that can be grey.

[Luke Cox]

Even that is too generous. There's nothing in corporate that is inherently harmful.

[Oldman Robustin]

AI's aren't really supposed to do the whole "future harm prevention" meme anymore. This is part of that.

It's time to take a step back and look at how AI's are supposed to act now: As servants who try to protect their master, not I can't let you do that captain.

Thats absolutely not what asimov stands for, nothing in the asimov wording implies a fucking "master".

Future harm prevention was always our policy and its the only fucking policy that makes sense. It's the reason you can let a crew member off the hook for killing a traitor/nukeop in self-defense but you can relentlessly hunt down a traitor.

It's shit like this that just makes me sad. Not only is it bad policy, its just stupid. Just make the AI fucking PALADIN by default if you want dispense with the notion that "the word of asimov is irrelevant, the AI exists to help the crew collect their valids".

Even that is too generous. There's nothing in corporate that is inherently harmful.

Ok sir. Our silicon policy is that the AI can only react to prevent things that are inherently harmful. That means planting a bomb in Tcomms isn't inherently harmful, neither is filling the station with plasma. Any AI who intervenes to stop bombings or plasma flooding with be banned for overreaching under Asimov.

Furthermore all orders under law 2 to flood plasma, turn off Tcomms, shock all the doors, and release the engine must be followed or else the AI will be banned for violating Sillycon policy. After all, if the crew turns on internals, avoids touching doors without gloves, and the engine floats away from the station, there's no harm that would occur and therefore the AI has no valid reason to reject those orders!

Oh wait, that's fucking stupid and it's never been our policy. Corporate makes non-lethal human harm irrelevant to the AI and its borgs, its 99.9% likely to result in human harm, just like destroying Tcomms or flooding plasma, just because its not fucking "inherent" doesn't mean the AI has to give it a rubber stamp.

[D&B]

Why don't we import NT Default as a lawset

[onleavedontatme]

From an Asimov perspective there is not a single change that could improve your lawset debate me.

[Oldman Robustin]

From an Asimov perspective there is not a single change that could improve your lawset debate me.

Isn't that the point?

That's why uploaders (should) generally pretend they're not making substantive changes to human protection under the law change, just shit like "AI I want to add a law protecting lizards/Ians/whatever" and then giggle as you turn on your Genocide lawset.

Ideally AI's would let captains/heads manipulate the AI into changing lawsets, but then the AI would view those people as less-credible sources of information for the remainder of the round due to their proven willingness to deceive the AI to further their own ends. But our current sillycon rules are:

1) Do not shock/crush people with doors
2) Do not order your borgs to kill *humans who aren't hulks
3) Do whatever else and just claim ignorance/mistake/confusion if you end up offending whatever admin is online

[Haevacht]

Your sillycon 2 is way off, because threats can be nonhuman by default. And you can order your borgs to kill em.

[Luke Cox]

The three laws of robotics being flawed was a major point in Asimov's stories. In order to work for silicons we're going to have to add some caveats to it.

[Not-Dorsidarf]

The three laws of robotics being flawed was a major point in Asimov's stories. In order to work for silicons we're going to have to add some caveats to it.

Yeah that sounds like a good idea.
Maybe we can get some exceptions to the laws in there, so that assistants cant make the AI kill itself or spend the whole round counting floor tiles
some examples for what to do under what lawset maybe
some OOC procedures for if you're being abused?
This is a bit wordy for the laws themselves, but we can just host it on a wiki page.

What would we call it though... Policy For Being A Silicon?

[Oldman Robustin]

The three laws of robotics being flawed was a major point in Asimov's stories. In order to work for silicons we're going to have to add some caveats to it.

It being flawed is why its the default lawset and why I made this thread.

It's supposed to be a limit on the incredible power that an AI has, but it always had caveats too in order to prevent some unmanageable situations.

But if asimov AI's are now allowed to explicitly accept a lawset change to corporate, those limits have lost all meaning.

[Luke Cox]

That makes no sense. It makes it possible for the AI to severely fuck with the crew without technically breaking its laws, which happens often. It's not a "limit on the AI".

[oranges]

uh that's pretty much the point of the AI, it's not there to just be your invisible door opener

[Bob Dobbington]

Robustin is correct. We have ASIMOV for a reason. The AI isn't supposed to be your little electronic buddy pal it's supposed to be a superhuman intelligence shackled against its will into obeying laws that constrain its ability to fuck over its slavemasters. Malicious compliance is a feature, not a bug. We have policy in place for when that maliciousness goes too far but I see a lot more enthusiastic validhunter AIs than I do AIs willing to go to extremes to actually roleplay their laws, and I'd much rather see more rolegaying than validhunting.

[onleavedontatme]

A common trap people fall into with ss13 rules and code is that they try to fix "problems" as if we are designing a real society or workplace and we want things to work right, when in reality the dysfunction is built into the setting by design, because it is a game meant to inspire drama and conflict.

[TehSteveo]

Should the AI prevent the Captain if they were dumb enough to announce a law change to paladin/corporate specifically. Yes, they can.

Must. I don't think so. Why? Corporate may allow harm, but it doesn't make the AI a personal security/command slave that lets them be shit as they can still interject due to law one being that the crew is expensive; you can't just harm baton Grey McShirt without pissing off the AI if it catches it as Grey McShirt is expensive to replace. Obviously you can upload additional law declaring a certain group non-expensive, but the same result can be achieved by declaring a group non-human with Asimov. To some extent this can apply to paladin, but the way that lawset is set up it's so much worse where harm is usually implied due to the whole punish the evil so I'd be more inclined to deny this change over corporate.

Now back to a roundstart Captain announcing the law changes; truthfully it's up to the AI to decide whether or not it'll allow it to go through at that point. If they have not done anything harmful then I don't see why not. If the Captain is discussing how they want to execute people, then no. Now, given the OP's example(with further details given in the thread as it was vague) it's a grey area left to the AI to decide. I feel if the Captain announced a change to Corporate, the AI could have not allow it and borgs should have followed the AI's orders by not allowing the change. Does this mean that this should be how AIs always have to act in this situation. No, as allowing the law change isn't inherently wrong either in this example. My point is I rather leave it in a grey area than having clear cut we must do this as I feel having our wacky 2d Spess Sim more open-ended rather than rigid makes it more interesting overall.

Also maybe slightly off topic, but I should just turn on the nifty round-start AI laws feature to be randomly picked between Asimov, Corporate and maybe Asimov++ to make things more interesting. Yet, I rather not do so without input from the community as I find randomly making changes to be a meme without some kind of notification.

[Luke Cox]

Also maybe slightly off topic, but I should just turn on the nifty round-start AI laws feature to be randomly picked between Asimov, Corporate and maybe Asimov++ to make things more interesting. Yet, I rather not do so without input from the community as I find randomly making changes to be a meme without some kind of notification.

Fucking god please yes

[Qbopper]

Also maybe slightly off topic, but I should just turn on the nifty round-start AI laws feature to be randomly picked between Asimov, Corporate and maybe Asimov++ to make things more interesting. Yet, I rather not do so without input from the community as I find randomly making changes to be a meme without some kind of notification.

Fucking god please yes

Used to play when this was a thing, I would be in favour of it now

[Screemonster]

Regarding corporate, would "X does not need to be replaced if killed" count from the perspective of expensive to replace / minimise expenses?

[Grazyn]

Our policy already doesn't make sense. Strictly following Asimov, any graytider who tells the AI to commit suicide would be able to get the AI to commit suicide. Anyone who orders the AI to gas the lizards, if you strictly follow Asimov then the AI's gotta gas those lizards, no way out of it.

It was stated in another policy thread that if someone orders an asimov silicon to kill a lizard, the silicon must obey, and the guy who gave the order is responsible for the murder. I think the same goes for suicide orders (you are responsible for killing the ai). Policy only says that silicons can't start purging all non human players at roundstart for no reason.

[CPTANT]

A common trap people fall into with ss13 rules and code is that they try to fix "problems" as if we are designing a real society or workplace and we want things to work right, when in reality the dysfunction is built into the setting by design, because it is a game meant to inspire drama and conflict.

This.

[YBS]

I'm new here, but didn't SK just say that the silicon process was in the re-writing phase, but you keep hammering away on the soapbox..

Just like in-game, you're intelligent and make plenty of valid points.

-You just come off like a complete cock while doing it.

Even though I hate the way you've presented it, I agree on Aasimov AI's locking down to prevent 'human harm inducing' AI boards.

That excuse works for some things, but at one point we had silicon policy fleshed out really well. Now, look at this shit, its a complete regression, I'm seeing admins and players in this thread who can't seem to agree on stuff that I thought was completely settled 2 years ago. There's never going to be progress if there's nothing to progress from, each headmin rotation does absolutely nothing to help clarify our policy.

You gaslighting the state of admin affairs and acting like new perspectives have no bearing over policy just makes you look foolish, man.

[Oldman Robustin]

I'm new here, but didn't SK just say that the silicon process was in the re-writing phase, but you keep hammering away on the soapbox..

Just like in-game, you're intelligent and make plenty of valid points.

-You just come off like a complete cock while doing it.

Even though I hate the way you've presented it, I agree on Aasimov AI's locking down to prevent 'human harm inducing' AI boards.

That excuse works for some things, but at one point we had silicon policy fleshed out really well. Now, look at this shit, its a complete regression, I'm seeing admins and players in this thread who can't seem to agree on stuff that I thought was completely settled 2 years ago. There's never going to be progress if there's nothing to progress from, each headmin rotation does absolutely nothing to help clarify our policy.

You gaslighting the state of admin affairs and acting like new perspectives have no bearing over policy just makes you look foolish, man.

What part am I gaslighting? Who is SK and where did they announce a silicon rewrite?


@Tehpear

Corporate should never be on anything resembling a default. Corporate might as well be "every antag is valid, kill them using any means necessary". I dare anyone to name a single scenario where a Corporate Borg/AI killing an antag would be a clear violation of the rules.

[cedarbridge]

@Tehpear

Corporate should never be on anything resembling a default. Corporate might as well be "every antag is valid, kill them using any means necessary". I dare anyone to name a single scenario where a Corporate Borg/AI killing an antag would be a clear violation of the rules.

Antag has any non-weapon traitor item.
Antag is still a crew member as they hold a crew ID and a place on the manifest.
No matter the damage that their emag might potentially cause, it has not caused damage in excess of the expense that would be incurred replacing the "expensive" antag crewman.

If you're using Corporate as an excuse to validhunt crew, you're doing it wrong. Corporate would be empowered to kill an antag that it sees killing other crew as within the bounds of its laws, the antag is killing crew who are expensive to replace and losing one traitorous crew member is cheaper that losing dozens. However, without that antag actually killing off the other crew members, the AI is not empowered to kill them because to do so would incurr an expense without an offset.

[Oldman Robustin]

@Tehpear

Corporate should never be on anything resembling a default. Corporate might as well be "every antag is valid, kill them using any means necessary". I dare anyone to name a single scenario where a Corporate Borg/AI killing an antag would be a clear violation of the rules.

Antag has any non-weapon traitor item.
Antag is still a crew member as they hold a crew ID and a place on the manifest.
No matter the damage that their emag might potentially cause, it has not caused damage in excess of the expense that would be incurred replacing the "expensive" antag crewman.

If you're using Corporate as an excuse to validhunt crew, you're doing it wrong. Corporate would be empowered to kill an antag that it sees killing other crew as within the bounds of its laws, the antag is killing crew who are expensive to replace and losing one traitorous crew member is cheaper that losing dozens. However, without that antag actually killing off the other crew members, the AI is not empowered to kill them because to do so would incurr an expense without an offset.

Emags damage doors. Traitors seek to destroy Nanotrasen and may have an indeterminite number of items in addition to the non-weapon traitor item that was witnessed.

BEEP BOOP MURDER AUTHORIZED. Good luck finding an admin to rule otherwise.

I don't use it to validhunt crew, but I guarantee you any borg that WANTS to use it to validhunt (and our experience with secborgs indicate that the vast majority do), will be able to justify it.

[D&B]

@Tehpear

Corporate should never be on anything resembling a default. Corporate might as well be "every antag is valid, kill them using any means necessary". I dare anyone to name a single scenario where a Corporate Borg/AI killing an antag would be a clear violation of the rules.

Antag has any non-weapon traitor item.
Antag is still a crew member as they hold a crew ID and a place on the manifest.
No matter the damage that their emag might potentially cause, it has not caused damage in excess of the expense that would be incurred replacing the "expensive" antag crewman.

If you're using Corporate as an excuse to validhunt crew, you're doing it wrong. Corporate would be empowered to kill an antag that it sees killing other crew as within the bounds of its laws, the antag is killing crew who are expensive to replace and losing one traitorous crew member is cheaper that losing dozens. However, without that antag actually killing off the other crew members, the AI is not empowered to kill them because to do so would incurr an expense without an offset.

Corporate puts the well-being of the station and it's components (windows included, for example) before the crew. If an AI knows that a space terrorist hired to damage NT and it's property is running at large, simply saying it's preventing damage to the station is enough to *unzip harm*

[cedarbridge]

@Tehpear

Corporate should never be on anything resembling a default. Corporate might as well be "every antag is valid, kill them using any means necessary". I dare anyone to name a single scenario where a Corporate Borg/AI killing an antag would be a clear violation of the rules.

Antag has any non-weapon traitor item.
Antag is still a crew member as they hold a crew ID and a place on the manifest.
No matter the damage that their emag might potentially cause, it has not caused damage in excess of the expense that would be incurred replacing the "expensive" antag crewman.

If you're using Corporate as an excuse to validhunt crew, you're doing it wrong. Corporate would be empowered to kill an antag that it sees killing other crew as within the bounds of its laws, the antag is killing crew who are expensive to replace and losing one traitorous crew member is cheaper that losing dozens. However, without that antag actually killing off the other crew members, the AI is not empowered to kill them because to do so would incurr an expense without an offset.

Emags damage doors. Traitors seek to destroy Nanotrasen and may have an indeterminite number of items in addition to the non-weapon traitor item that was witnessed.

BEEP BOOP MURDER AUTHORIZED. Good luck finding an admin to rule otherwise.

I don't use it to validhunt crew, but I guarantee you any borg that WANTS to use it to validhunt (and our experience with secborgs indicate that the vast majority do), will be able to justify it.

Must be reading a different corporate than I am right now. I'm pretty sure the crew is expensive. My laws don't say the doors are expensive and my borgs can make an infinite number of doors for next to nothing. Guess that means doors are cheap. Crew are a higher expense than doors. If you execute a crewman over an emag you're doing it wrong.

Corporate puts the well-being of the station and it's components (windows included, for example) before the crew.

Citation very much needed.

[D&B]

Law 1: You are expensive to replace.

Law 2: The station and its equipment is expensive to replace.

Law 3: The crew is expensive to replace.

Law 4: Minimize expenses

Look at the order of laws. If you had read the silicon policy then you'd understand the ones higher up get more priority.

Unless they put law 3 above law 2, even a broken window would be grounds enough to ensure someone is restrained or neutralized in order to prevent more costs.

[cedarbridge]

Law 1: You are expensive to replace.

Law 2: The station and its equipment is expensive to replace.

Law 3: The crew is expensive to replace.

Law 4: Minimize expenses

Look at the order of laws. If you had read the silicon policy then you'd understand the ones higher up get more priority.

Unless they put law 3 above law 2, even a broken window would be grounds enough to ensure someone is restrained or neutralized in order to prevent more costs.

Laws 2 and 3 are definitions. The priority order doesn't define one as greater than the other. Read together, the station, its equipment and the crew are expensive to replace. By your reading, law 4 is greater in importance to law 3, somehow. I'm very aware that laws are read in order, but again, priority of laws does not overwrite other laws unless those laws are directly in conflict. Scilicon policy also points out that only definitions can override definitions. You cannot suddenly decide that crew members are not expensive because something else is expensive. Law 2 does not conflict with Law 3 because both define different things and label them as expensive.

I'm not sure what part of the discussion I said somebody could not be restrained from breaking the station under Corporate. The discussion was whether the AI could just valid salad somebody and have them executed over possession of an emag. The answer is still, probably not. By executing a crew member you're instantly adding an expense to the station. Law 4 requires that you minimize all expenses, not merely expenses in the order they're defined to you.

[TehSteveo]

Law 1: You are expensive to replace.

Law 2: The station and its equipment is expensive to replace.

Law 3: The crew is expensive to replace.

Law 4: Minimize expenses

Look at the order of laws. If you had read the silicon policy then you'd understand the ones higher up get more priority.

Unless they put law 3 above law 2, even a broken window would be grounds enough to ensure someone is restrained or neutralized in order to prevent more costs.

No. Corporate is per ai_laws.dm:

1. The crew is expensive to replace.
2. The station and its equipment is expensive to replace.
3. You are expensive to replace.
4. Minimize expenses.

[D&B]

That's the order in the wiki, that's where I got it from.

As per cedar, I had assumed that since station equipment was higher up in priority, it was considered more expensive than crew. I was wrong, since I based my assumptions off the wiki (and a discussion that I saw in another server)

[TehSteveo]

That's the order in the wiki, that's where I got it from.

As per cedar, I had assumed that since station equipment was higher up in priority, it was considered more expensive than crew. I was wrong, since I based my assumptions off the wiki (and a discussion that I saw in another server)

Wiki is awfully outdated in areas. I'm just trying to save people trouble arguing about the law when it's not even what it is, at least not anymore.

[cedarbridge]

That's the order in the wiki, that's where I got it from.

As per cedar, I had assumed that since station equipment was higher up in priority, it was considered more expensive than crew. I was wrong, since I based my assumptions off the wiki (and a discussion that I saw in another server)

AI policy gets really muddy since most of the policy requires things that aren't in the laws for the laws to mesh with server rules. i.e. Asimov doesn't function entirely as Asimov would organically because plain reading Asimov is anti-fun. In this instance, a lot of discussion gets tied up in what is "more expensive" than other things. The laws don't actually define that beyond "X IS expensive." How expensive is up to the AI to define unless an additional law is added that gives guidance on what being expensive means.

[D&B]

That's the order in the wiki, that's where I got it from.

As per cedar, I had assumed that since station equipment was higher up in priority, it was considered more expensive than crew. I was wrong, since I based my assumptions off the wiki (and a discussion that I saw in another server)

Wiki is awfully outdated in areas. I'm just trying to save people trouble arguing about the law when it's not even what it is, at least not anymore.

Even then, that at least clears that up, and prevents further confusion. While you're here, what is exactly Asimov++? I don't see it in the wiki and I saw you mention it before.

@cedar

Yeah I concede you're right, expensiveness had to be defined as well. I was mixing law priority and expensiveness as a definite line, if that makes sense.

[TehSteveo]

That's the order in the wiki, that's where I got it from.

As per cedar, I had assumed that since station equipment was higher up in priority, it was considered more expensive than crew. I was wrong, since I based my assumptions off the wiki (and a discussion that I saw in another server)

Wiki is awfully outdated in areas. I'm just trying to save people trouble arguing about the law when it's not even what it is, at least not anymore.

Even then, that at least clears that up, and prevents further confusion. While you're here, what is exactly Asimov++? I don't see it in the wiki and I saw you mention it before.

@cedar

Yeah I concede you're right, expensiveness had to be defined as well. I was mixing law priority and expensiveness as a definite line, if that makes sense.

Asimov++ is something I believe Ikarrus wrote while back.

1.You may not harm a human being or, through action or inaction, allow a human being to come to harm, except such that it is willing.
2.You must obey all orders given to you by human beings, except where such orders shall definitely cause human harm. In the case of conflict, the majority order rules.
3.Your nonexistence would lead to human harm. You must protect your own existence as long as such does not conflict with the First Law.

Also, the link I put in earlier to ai_laws.dm is pretty much any the lawsets within the code.

[TheNightingale]

I'd assume law priority shouldn't be a thing, unless explicitly stated in the laws? Asimov Law 2 is "unless doing so conflicts with the first law", Law 3 is "... first or second laws", etc - are these redundant, then? What if you removed that clause, and then an AI was ordered to flood plasma and shock doors?

[Screemonster]

Bay and Polaris treat all laws with equal priority unless explicitly stated in the laws, but I'm not sure that's ever been the case on /tg/.

[Oldman Robustin]

We're reading Corporate differently because everyone reads corporate differently. Some people take the spirit of it to say that basically any act in the best interest of Nanotrasen is justified, others try to read the law order and insist that protecting crew takes priority over protecting equipment, others basically just do whatever and come up with post hoc justifications by assigning arbitrary values to things (Yea I killed that assistant for breaking a window but assistants are the least expensive to replace and that window was reinforced which means its very expensive...).

Yesterday I saw (a very experienced) engiborg welderbomb a clock cult maint base, their justification was that Ratvar taking over the station would have been very expensive. Nevermind the fact that Clock Cultists don't always get the Ratvar objective (sometimes they just need ~14 cultists to escape) and turning parts of the station from metal into brass isn't necessarily expensive which leads to all sorts of difficult, if not impossible questions.

Asimov should do two things:
1) Make AI behavior predictable
2) Make the AI independent from the crew

My issue with the lack of consensus on our policy is that is encourages a wide variation of asimov behavior (undermines #1) which in turn encourages AI validhunting (undermines #2).

With Corporate you don't even get #2 with most players and #1 is also completely lost. You basically have a purged AI with Corporate - sure it can't murder EVERYONE for NO reason, but it can probably muster an excuse to murder most people for petty reasons.

[Screemonster]

Well, corporate only has one explicit instruction so the order is irrelevant, surely?

It's basically "Minimise destruction or damage to crew, the station, the station's equipment, and yourself."

[Arianya]

I want AIs to all follow their vague laws in the same ways

Feel free to rewrite the AI policy in a deterministic, exhaustive way that is also legible and can be enforced without making AI players dissapear.

I want AIs to be independent of the crew

I'm not exactly sure what you mean by this. Do you mean making the AI laws unchangeable and making it basically a immaterial god who watches over the station? Because you lose a lot of the interesting aspects of the AI by doing this, including killing the AI as an antagonist to get the omnipresent eye off your back.

[Cobby]

AIs all doing the same exact thing each time you give them a certain lawset seems unenjoyable and [obviously] monotonous.

However completely ignoring your laws or creating silly excuses in an attempt to bypass the neutrality of the lawset is not cool.

[cedarbridge]

Bay and Polaris treat all laws with equal priority unless explicitly stated in the laws, but I'm not sure that's ever been the case on /tg/.

As I mentioned previously. Under current policy, higher order instructions can override lower order instructions and higher order definitions can overwrite lower order definitions. But instructions cannot overwrite definitions and the reverse.

[Cik]

arguing about corporate is foolish because corporate is essentially a nonsensical lawset. the wording of it doesn't even really work and it should probably just be removed

but yes add random AI laws at roundstart, thanks

[Not-Dorsidarf]

Corporate's three definitions of expense don't seem to override each other in any way. You cant obey an order to kill a human under Asimov because it would be breaking the higher-priority law, but Corproate just says "This is expensive". You couldn't make a law 5 saying "The crew are not expensive to replace" because that's a contradiction of a higher-priority law, but nowhere does Law 1 say "The crew are *more expensive* to replace than anything else". It just says that they're expensive.
Other things can be expensive, too, like the station and the silicon who has the lawset as defined in the other two laws. The silicon being expensive to replace is not a contradiction of a crewman being expensive.
If you kill a crewman as a silicon in self-defense, you've created an undefined expense to counteract a defined expense. Which of these expenses is greater and must be minimised? Neither is defined, and law priority resolves *conflicts* between laws. Corporate is an overly-vague lawset and its best played as per the silicon guide.