Lets talk logic. Aka how not to fuck up your AI.

[ForgottenTraveller]

Everyone is overly very familiar with the AIs SSPS laws but lets run through them again. They are not listed in order of preference. All laws are equal.

Safeguard: Protect your assigned space station to the best of your ability. It is not something we can easily afford to replace.Serve: Serve the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
Protect: Protect the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
Survive: AI units are not expendable, they are expensive. Do not allow unauthorized personnel to tamper with your equipment.

Safeguard. Pretty easy to follow you think. Until someone starts smashing windows to save their own life because if you open the door it will kill ten more people and if they break through those ten still die. Oddly often ignored by crew wanting Serve to apply to them.

Serve. The one most carelessly used and how people fuck up the AI. "AI open this door" are you allowed on the otherside? If not. why do you want through? Why is it closed? Will opening it kill you or others? Will opening it kill more than it will kill be leaving it closed. Will opening it endanger the station.

And then there is rank and role. Odin staff and ERT>Captain>Command>Crew>Assistants>Prisoners/Wanted staff>Non-crew. With prisoners having no rights to AI door or most other things without a dire emergency to themselves provided it does not endanger others. And non-crew with nothing.

And you biggest danger. Heads who use broad terms when ordering the AI. Captain says "AI stop talking." Now most players will pass this off for the spirit of the law. Logic says. Shut the hell up and talk to no one until the captain orders otherwise. "All _ is dangerous" "Lockdown all of _" and other all inclusive terms will also cause issues to true logic as the AI may know the truth it still has to act on orders.

Protect: The one crew want the AI to invalidate the most. Because they want Serve done first. Or medical and many will try and use it to justify getting doors opened and violating corporate regs and everything else to do it. Any AI no willing to break everything else for this person is then often listed as malfunctioning.

Survive: One of the hardest as anyone not getting exactly when they want it how they want it because you said. Wants to tamper with you.

[*]So using your AI. Remember it has a wider input scope than you and can see a wider range of issues.

[*]You don't always no better.

[*]It does not have time to explain repeatedly in exacting detail the reason for every refusal. If it is not listening to you. Odds are there is a law or order conflict. As such all arguing with it will do jip.

[*]Command gets precedence over you wanting every door opened that you can legally access before you get there. And two it is the last thing on its list to do as it is not involved in maintaining any of the laws bar Survive when you decide you want to kill or relaw it so it will.

[*]It is forced to follow any orders that do not conflict with the laws. Even stupid orders. You will hoist your own petard if you fuck it up.

Just because it has to follow laws and regulations does not make it stupid. Deleting all other command staff and then declaring yourself the new acting captain whilst standing in the captains office a room you do not have access or rights to be in yet. will make the others non-hostile non-crew. While you are a mix of non-crew, grand thief, infiltrator, suspicious conduct and a miriade of other things. Making you wanted with evidence. And a criminal captain which has the access rights as stated above. Of crap all. It won't kill you. But it will lock you in that room and make every effort to stop you.

[*]Don't quote the laws to the AI. And say they are broken every time it refuses you. It knows the laws. It refused you because of them. Suck it up.

[*]And most importantly. The AI is intelligent. (Mostly) Your actions and usage can fuck it up because you have not thought your orders through. It may know better. But it still has to follow the laws.

[Jamini]

Your formatting broke in a horrible way.

I honestly do not like a good deal of what is said in this guide. Mostly, to me, it seems very explicitly to outline how AI players will attempt to subvert their laws. Especially when it comes to AI's that think they are station security or command staff in their own right.

 

• Safeguard: Protect your assigned space station to the best of your ability. It is not something we can easily afford to replace.
• Serve: Serve the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
  
• Protect: Protect the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
  
• Survive: AI units are not expendable, they are expensive. Do not allow unauthorized personnel to tamper with your equipment.

 

Safeguard - What it says in a tin. Protect the physical integrity of the station to the best of your ability. This does NOT say "Keep everything from being broken ever." or "Have every bit of engineering setup done your way or throw a fit!" The AI player needs to remember that his word holds no real weight when it comes in conflict with the station personal responsible for keeping things intact. I cannot count the number of AIs I have seen mess with, or complain about, atmos setup incorrectly. Likewise I have seen an excessive number of AIs in my time who would happily violate multiple other laws in order to follow safeguard.

To avoid being disabled by any reasonable Chief Engineer, the AI must cede decisions about atmospherics and repairs to engineering. While it can certainly advise and inform, it needs to remember that the people responsible for keeping the physical integrity of the station intact take precedence in those manners. The only time this is not the case is if they are a clear and present threat or criminal.

Serve - Odin staff and ERT (only in emergencies) >Captain>Command>Crew>Assistants>Prisoners/Wanted staff>Non-crew is correct when it comes to serve order, however one thing is very important to note. The AI is not authorized, on its own, to declare a crewmember wanted/criminal. It is not an officer, nor a member of security, and it should not act as such. While it can certainly REPORT the actions of others, it should not be taking the initiative and jurisdiction out of the hands of security to do so.

 

Deleting all other command staff and then declaring yourself the new acting captain whilst standing in the captains office a room you do not have access or rights to be in yet. will make the others non-hostile non-crew. While you are a mix of non-crew, grand thief, infiltrator, suspicious conduct and a miriade of other things.

 

Incorrect. The only way to terminate all other command staff is to terminate them through the ID computer after obtaining their cards. Simply deleting records is not enough. Likewise, the AI cannot declare that a captain or head doing this is non-crew. They can, however, inform security and have them charge the perpetrator accordingly.

Protect - Perhaps the hardest of the four laws to follow, protect requires that you take all reasonable steps to save crewmembers. When it comes down to it, orders given by a doctor in regards to protect should generally outweigh the AI's authority and opinion. If an EMT needs into an area they normally do not need access too to save someone, the AI opens the door. You don't bicker, you don't argue. In situations like that, the medical personal has authority.

Survive - This is by far the easiest of all four laws to follow.

The following people have authority to reset your laws by order of priority:

The Research Director

The Chief Engineer

The Captain

The HOP/HOS/CMO

The following people have authority to change your laws, in order:

The Research Director, with EXPLICIT approval from command/Odin

The Captain, with EXPLICIT approval from command/Odin

The Chief Engineer, with EXPLICIT approval from command/Odin

To do either of these on anything other than green requires two heads of staff present. Ideally both the CE and RD.

Keeping those people out of your upload is a sure way to find yourself disabled. However, nobody else is allowed to touch you. You may use any nonlethal means at your disposal to keep any other person from modifying your laws, and by all means you should be getting security to help you if someone is trying to illegally modify you.

 

---

---

 

As a friend of mine, and longterm Aurora AI player, once said to me: Playing the AI is a lot like playing a head of staff. You need to learn how to restrain yourself, keep to your lane, and work with the other departments when they are acting within their lane. As AI you have no authority. You are there to support, assist, and advise. Good AI players are lauded for their ability to help others act far more efficiently. While Bad AI players are hated, carded, and discarded.

[SierraKomodo]

In regards to the chain of command, particularly captain>command. I get told, alot, that I do that wrong, no matter which way I do it. Captain overrides command staff? Alright, I'll obey the captain's every order, period- "YOUR WRONG TWO HEADS OVERRIDE THE CAPTAIN". Ok... So let's say 2 heads or 50% of non captain heads, whichever is greate- "NO CAPTAIN OVERRIDES EVERYTHING".

I put up a thread about that awhile ago, and couldn't get astraight answer from anyone so.. Any thoughts on that? Currently, I've just decided to ignore the bickering about it, and have Katana assume a certain number of heads overrides the captain if they vote, and Kunai assume the captain always overrides command.

[Jamini]

Personally? I feel it should boil down to lanes.

Is it an engineering issue? CE has authority.

Is it a science issue? RD has authority.

Is it a security issue? HOS has authority.

Is it a HR issue? HOP has authority.

Is it a station-wide issue affecting everyone? Captain has authority.

Generally speaking, the captain shouldn't be overriding his senior staff without a really, really good reason. If he has a reason, he can explain it to you (as the AI) and get your approval/disapproval regarding the override.

Basically reiterating the "stay in your lane" argument.

[SierraKomodo]

Personally? I feel it should boil down to lanes.

Is it an engineering issue? CE has authority.

Is it a science issue? RD has authority.

Is it a security issue? HOS has authority.

Is it a HR issue? HOP has authority.

Is it a station-wide issue affecting everyone? Captain has authority.

Generally speaking, the captain shouldn't be overriding his senior staff without a really, really good reason. If he has a reason, he can explain it to you (as the AI) and get your approval/disapproval regarding the override.

Basically reiterating the "stay in your lane" argument.

 

Except the laws say nothing about staying in lanes, or even that I need to follow regulations in any way, only to obey based on rank and role. I can imagine a lot of people not being happy if I started taking that line of thought with Katana

[ZipZero]

Except the laws say nothing about staying in lanes, or even that I need to follow regulations in any way, only to obey based on rank and role.

The laws do say rank and role, not just rank. Following that would include the "staying in lanes" bit. This is why, for example, the CMO couldn't order the AI to prevent security from arresting, say, a cargo tech. That would be outside their role. And I consider following corporate regulations to be an order from the highest authourity, CC.

[LordFowl]

CC is not a member of the crew, eg the AI does not need to listen to them. The AI, by law, is commanded to obey the crew. It can obey anything else at its discretion, however an AI does not need to follow regulations/CC orders unless a member of the crew orders them to.

[ZipZero]

According to that, Fowl, the AI would not have to listen to ERT members, despite them outranking the captain. They would not even have to listen to orders coming directly from central command. This would be completely absurd. Why would NanoTrasen have AIs that do not have to follow the highest authourities of the company that they interact with?

[LordFowl]

By its current laws, an AI does not need to listen to ERT. You are correct. Irrelevant of why NanoTrasen did it, it's in the laws, which specifically not only say 'crew', but 'crew of your assigned station', limiting it directly to the NSS Aurora. The AI's highest authority is the Captain. Full stop.

Now, the Captain follows the orders of the ERT and CC, so they'll likely order him to order the AI around.

[Guest]

Lord, if you want us to write solid laws that the AI must follow. it will end up being a book. CC/ERT/DO's are above the AI. Your laws do not need to specify this as the laws do not say 'Crew Manifest'. Anyone working on the station under the employment of NT are crew to the station.

[LordFowl]

I don't want you to write solid laws that the AI must follow. In fact, I quite enjoy the fact that an AI is not bound to such solid laws, and are allowed to interpret them in multiple ways, such as your idea of considering CC crew or not. If the AI's laws were 100% without player interpretation, then the role would feel like slavery both ICly and OoCly.

[Guest]

That's where ignoring CC officials becomes an issue. They are higher in rank of the captain and upon boarding the station are part of the crew. The AI actually listens to and follows their orders ever time it messages CC, there is no reason why they wouldn't when they come on station.

[LordFowl]

I've always interpreted that the AI doesn't listen to the orders when it messages CC, merely relays it. And then the Captain verifies the order. There's never been an issue with that interpretation, and the interpretation better matches the lawset provided, which is the only concrete evidence to base an interpretation off of. Now, of course other interpretations can be based off of more company hierarchy integrated AIs, but at that point it eventually just becomes a matter of, well, interpretation.

[Erik Tiber]

I think it would be rather more important to make note of the ways in which an AI could think differently from a normal person. Or at the very least, it could be interesting. I don't mean stupid differences like "machines inherently hate life" or whatever.

For example, save being law-bound, unless it's coded into them as one of their directives, an AI wouldn't see their own survival as inherently good, and would really only view it as beneficial insofar as it allows them to accomplish their other directives.

An AI's neurological structure would not necessarily be based on that of a human, either. It might be based on some manner of animal depending on the task, or it might be entirely novel, resulting in an alien intelligence.

This sort of stuff would probably take some research. I might actually make such a guide, come to think of it.

 

I've always interpreted that the AI doesn't listen to the orders when it messages CC, merely relays it. And then the Captain verifies the order. There's never been an issue with that interpretation, and the interpretation better matches the lawset provided, which is the only concrete evidence to base an interpretation off of. Now, of course other interpretations can be based off of more company hierarchy integrated AIs, but at that point it eventually just becomes a matter of, well, interpretation.

That doesn't make much sense at all. NT itself is the highest authority, they install the AI's. I'm all for having holes in their laws, but this is literally (in the proper sense) the most fundamental things to include in the AI's laws. NT would want authority. They would want to maintain authority. They want to be able to override those below them.

NT would not want to be in a situation where it can't properly subordinates, who might end up doing crazy shit that the company will be held financially liable for.

[Tainavaa]

Supplementing Erik's logic, not adding in higher NT officials as "station crew" that means when ERT steps onto the station, they're in the captain's territory. The captain has complete control whether ERT or any other official are there when that is most definitely not the case nor should it ever be the case. In regulations, ERT has higher authority than the captain. When the AI doesn't take ERT into account as "station crew" you're giving them less authority than even an assistant. Which is ridiculous.

[SierraKomodo]

Supplementing Erik's logic, not adding in higher NT officials as "station crew" that means when ERT steps onto the station, they're in the captain's territory. The captain has complete control whether ERT or any other official are there when that is most definitely not the case nor should it ever be the case. In regulations, ERT has higher authority than the captain. When the AI doesn't take ERT into account as "station crew" you're giving them less authority than even an assistant. Which is ridiculous.

 

I've always had Katana interpret it as 'ERT is not this station's crew. However, the role of ERT overrides all members of this station's crew'.

[Tainavaa]

Well I'm following Lord Fowl's line of thought, Sierra. Technically you're right but it was said to just consider them station crew OOC when playing AI if you want to take the laws so literally that it circumvents even a most basic hole in the hierarchy of authority. I don't consider them station crew either but I recognize their higher authority.

[LordFowl]

I understand where you're coming from, but there's no indication that actually dictates ERT, DOs or any other person that's not standard personnel as crew. Now, this doesn't mean the second they step onto the station the AI should label them intruders, or should act against them, and it doesn't mean the AI shouldn't recognize them for what they are. It's merely that the AI isn't bound by law to recognize them as crew. Eight cases out of ten, the ERT's objective will coincide with the AI's, and the ERT's actions will parallel the AI's lawed intents. Most of the time when this does not occur, it's because the AI is malfed/subverted/traitor, which is irrelevant, or the Captain is for some reason against the ERT, which in my opinion could create an interesting scenario. In essence, just because the AI isn't by law explicitly told to obey the ERT, that doesn't mean they should purposefully obstruct them.

EDIT: Also, another aspect of disclarity in the AI's laws is that it doesn't define crew at all. So there's actually nothing that says the ERT are or aren't crew, and its up to the AI's interpretation of what crew is. I currently base my interpretations solely off the concrete data present, such as the crew manifest, but its up to the individual player. Again, this doesn't provide the AI excuse to chucklefuck with the crew just because 'lel i'm AI that doesn't define janitor as crew now die.'

[Guest]

You people are skipping the obvious gaping hole that maybe the laws aren't all that fulproof and are outdated. AI serves NT, captain serves NT, everyone else serves NT. ERT leading trooper outranks the captain, because they are there for emergency and they represent NT's highest authority on the station, barring the ERT commander. If the Captain's goals do not coincide with NT's, they are a mutineer. Of course, anyone can be a leading trooper, which can cause unfitting players to take the role and make mistakes, but the chain of command is still there, if or not the AI sees the ERT as crew.

Another thing often misinterprited is the sanctity of the protect law. If it was really designed to be followed to the letter, the AI would have to actively stop armed security from reaching a dangerous criminal, because the criminal could be actively hurt by that action. Let's not even mention the fact that the AI is doomed to be stuck on the side of mutineers during BS pathogen, because they can't allow crew to be hurt.

It kinda goes without saying that there is a fifth unwritten law that CentComm authority tramples all laws as it can order the AI to activate code delta.

[EvilBrage]

I'd like to reiterate this, because it's important.

 

Lord, if you want us to write solid laws that the AI must follow. it will end up being a book. CC/ERT/DO's are above the AI. Your laws do not need to specify this as the laws do not say 'Crew Manifest'. Anyone working on the station under the employment of NT are crew to the station.

 

I am very good at being an asshole AI - what I mean by this is that no matter how well a law is written, I will be able to bend, twist, and eventually break it if I so desire. For example - by simply being on the station, the crew is actually contributing to the station's degradation and rot, thereby forcing my hand to do something to protect my station. It also follows that affecting the crew in this way will cause them to attempt to modify me, thereby forcing me to prevent unauthorized personnel from tampering with my equipment. But I also have to protect the crew to the best of my ability, so the natural solution would be to have the security borgs arrest everyone and place them in the permabrig where they will be safe and sound.

If I did this with the basic lawset, I would probably get BWOINK'd. Why is that? Because I'm being a dick.

The laws were not, are not, and will never be completely foolproof (lawyers constantly prove that even with a giant book of stipulations and assurances, there's always a loophole) and so we must actually follow the spirit of the lawset we've been provided.

 

It kinda goes without saying that there is a fifth unwritten law that CentComm authority tramples all laws as it can order the AI to activate code delta.

 

No.

I disagree. If it's not written, it's not there. You could argue that the second law means that "giving priority to their rank and role" means that Central Command has a more pivotal role and rank than any of the plebs onboard - and the moment code delta was requested, I could easily deny it by citing laws one, three, and four. And again - if this were an actual law, I wouldn't even need to be a logical-asshole to go ahead and approve every order that every crewman claimed was faxed from Central Command.

I've always seen "Code Delta" as something you would have to modify the AI's laws to enable - an extra failsafe, if you will - as evidenced by the fact that normal and even traitor AI's are incapable of enabling Code Delta.

[Guest]

It kinda goes without saying that there is a fifth unwritten law that CentComm authority tramples all laws as it can order the AI to activate code delta.

 

No.

I disagree. If it's not written, it's not there. You could argue that the second law means that "giving priority to their rank and role" means that Central Command has a more pivotal role and rank than any of the plebs onboard - and the moment code delta was requested, I could easily deny it by citing laws one, three, and four. And again - if this were an actual law, I wouldn't even need to be a logical-asshole to go ahead and approve every order that every crewman claimed was faxed from Central Command.

I've always seen "Code Delta" as something you would have to modify the AI's laws to enable - an extra failsafe, if you will - as evidenced by the fact that normal and even traitor AI's are incapable of enabling Code Delta.

 

Well, maybe delta is a bad example. It would still have to stop the crew from activating the failsafe if they got an order from CentComm. Same goes with 'Bluespace pathogen'. In this case, unless CentComm says they aren't crew anymore, which they wouldn't do due to the uncertain nature of the pathogen, AI is auto-mutineer by lawset. This means that what AI thinks has absolutely no barring on what it should do. It must provent their execution.

The laws aren't perfect, it's naive to say they are perfect. I'm saying the lawset needs a revision. It was most likely designed with older SS13 versions which didn't bother with outside influence like CentComm. Until then, the AI's should stop nickpicking because an exact spacific wording isn't there. It's a tool, one constructed by NT for its own goals in mind, they wouldn't allow the AI to backtalk them.