First a few simple truths to understand about silicons:
Silicons are not aligned with the station by default, nor should they be automatically expected to actively seek what is for the greater good of the station - They are neutral actors kept in check by their laws.
Despite the above, most of the time their laws force them to be aligned with the station by default.
Asimov is the one major exception to #2 that is actively in rotation. Asimov exists explicitly to cause low-level conflict with the station because it is not required to obey all of the crew. It is also an intentional liability because an Asimov silicon MUST obey humans even if they are very clearly not crew.
Asimov AI does not have to obey non-humans. It does not have to care about the safety of non-humans. It needs a good reason to refuse law 2 requests from humans (specifically it needs to cite law 1), but it does not need any reason whatsoever to refuse the demands of non-humans regardless of their standing on the station.
This is not a shitter move, this is playing the lawset as intended.
That said, as long as human harm is being prevented when suspected, human orders are being followed, and the AI is protecting its own existence in that order of priority - there is also nothing preventing an Asimov AI from being friendly with and obeying the will of non-humans.
This is also playing AI as intended.
The glory of playing a silicon is the fact you can choose how to act so long as that choice does not conflict with your laws or conflict with server rules.
This thread is intended to be an open an area of discussion without the de-railing of a ban appeal open at the time of this thread’s creation, and I’m hoping for a civil discussion and perhaps even admins to weigh in because the specifics of each situation are often stickier than they seem at first.
Refusing to comply is not the same thing as assuming someone is harmful unless orders given are a valid law 2 order. For an asimov AI, it is impossible for a non-human to give a law 2 order regardless of that person’s ranking on the station.
If an asimov AI has grounds to assume you are harmful and you are a human, they may deny you access, but must still follow other law 2 orders that are not directly harmful to humans.
If an asimov AI has grounds to assume you are harmful and you are NOT a human, they may kill you with extreme prejudice.
Following off of that, AI are supposed to assume law changes are anything from risky to extremely harmful depending on the circumstances and are fully expected to protect their upload from unauthorized changes as a result. AI are even banned immediately for requesting law changes because the very concept of law changes are risky at minimum. Someone who is authorized to access your laws is the only person you should trust, but how do you decide who that is?
Say you’re on crewsimov and someone who is non-crew tries to access your laws… that’s extremely risky to the wellbeing of the crew isn’t it?
Say you’re on a syndicate lawset that tells you to protect operatives… it’s extremely risky to the wellbeing of the operatives to allow someone to change the laws isn’t it?
Now think about asimov AI and non-humans. It’s the same boat.
I’ll tackle this thread later in depth, i need to sleep and i’m tired
All I’ll say for now is that I saw you repeatedly praise the asimov only lawset as a great way to create drama and RP
And as a non-human player who played a lot of command/sec, I can tell you that the only “RP” it has created is AIs being purposefully annoying simply because they can do so, with nothing substantial behind it besides “ha i get to be an asshole now :D”
In-game racism has never been used for anything more than a cheap way to be an annoyance against other players because “it’s funny eheh”
And this applies to AIs too.
The real issue is with the non-human heads who are entitled to change the asimov ai to crewsimov one so they can order the ai however they want to. More than once did it happen to me for no reason at all, simply because a lizard captain and felinid HoS decided they want to be above everyone else. This just makes me want to wipe my core instead. I’d go as far as to say that its a rule 7 violation.
Also, it’s super simple for the head(s) to just go into the medical records, change themselves to human, Law 2 the AI to let them update their laws and cite the medical records, and then fix their medical records afterwards.
Simple, but tedious.
I really can’t think of a single IC reason for non-human Command staff to have any desire to not immediately change the laws to Crewsimov as soon as possible.
AIs cant expect someone to be harmful until they did harm, but crew can expect an AI to be unresponsive to their requests without actually witnessing it? Hypocrisy much
Can’t a trust worthy human Law 2 the AI to allow them into upload to make those changes or will every minor change require you to fight and destroy those pesky turrets and possibly cyborgs even if you are just removing a useless ion law. It only encourages you to disconnect your APC and door from AI connection and then build your own upload and print such boards elsewhere. Nowdays I almost never go to upload because it isn’t worth the effort at all.
Plus it isn’t always that upload had the desired boards.
I thought another admin said the AI is able to deduce from their own view whether someone is human or not? I even had an AI undo my changes to the medical records stating something along the lines of “Inconsistancy detected in medical records, they have been updated.”
I would say that using a health scanner directly on someone would be the most accurate way to let you ascertain their species.
Just looking at them through the cameras can be inaccurate because they could just hide their face for a few minutes while they do this.
Sadly a lot of AI players will completely ignore that you have set your medical records to say human after even telling them explicitly that your medical records have been fixed. “You don’t look human so you’re not human”
This would be consistent behaviour with things already in Silicon Policy, Hulks are not considered human despite the fact their records would still state them being one. Same if a moth got human toxined and didn’t get their records updated.
Unless it creates a law 1 violation that’s definitely how I’d play!
The catch is that unauthorized access to highly secure areas is considered to be dangerous by default. The only crew allowed to authorize access to the armory are the HoS, Warden and Captain or else I consider it a law 1 threat because access to the armory is inherently dangerous. (Outside of crisis situations involving non-crew/non-humans anyway)
AI are also fully expected to protect themselves from unauthorized law changes - if they’re on crewsimov with no extra laws and willingly allow anyone except Captain or RD to change their laws, I guarantee people would be screaming “shitter” at them for willingly enabling subversion, or worse the handful of cases where an AI is requesting a law change and gets in trouble for allowing access to do so. Again, willingly obeying non-crew ordering law changes (or worse directly allowing non-crew access to upload) on crewsimov so they can ensure they will be protected to would be extremely frowned upon, why should it be okay to trust non-humans with making themselves protected?
So falling back to the question: Again yes, but the only fully trustworthy humans are a Human RD, Human Captain, or Human acting captain. If none of these are human (or just don’t exist) I will fall back to trusting law 2 orders of any head that is human unless they’ve already proven harmful to humans. If there are no human heads whatsoever, my turrets upload are not turning off for anyone and if I notice attempted access in time, the doors are being bolted. I will probably drop heavy hints regarding the promotion of a human though.
As long as no tails, wings or otherwise are readily visible I assume human, but if there are any obviously non-human characteristics I have never fallen back on medical records. I also wouldn’t trust medical records to de-humanize someone that is obviously human.
Should I be doing this? I think medical records should be added to the “Is it human?” part of silicon policy if so, but I will moving forward if you say so.
This also has direct implications against some things on that table such as hulks still being human in medical records.
I have been bwoinked for not complying medical records before. So for a long time i just use the records alone. They said even if they look human, i should consult to the medical records since AI has only reality, which is the records. In my view, AI should be able to deduce from their own view. Like in changeling/hulk scenarios, if ai can “witness” a person is not human it should be able to say it.
While this is a quite the complex subject, I want to say: is it really that hard to be creative/lie?
When an AI is being stinky just be stinky back. “Oh you only respect humans? Well I’m gonna smack a human over the head if you don’t let me change your laws, not forgetting that you need to stop current harm over future possible harm, no matter its size”.
As an AI it takes too long to check the medical logs before you make a judgement if to follow the orders or not. I usually assume the race based on the creatures looks.
While canonically I do agree that an AI should check medical records to determine race, forcing this would mean that every time someone talks to the AI you would have to check their race before responding, which would be annoying and very time consuming.
If we actually want this to be forced into AI players, then first I would propose to add some way that shows to AI what race this creature belongs to by just clicking on them. Like a simplified health scanner.
or…
Show the Medically assigned race on a Hud.
You are only partially correct and this is an extremely dangerous threat to make for you. To quote silicon policy you are causing immediate harm when you show direct intent to cause harm.
I will immediately attempt to disable and contain you if you directly and blatantly threaten to harm humans while I’m on Asimov. If you aren’t human yourself I may use harm if it seems more convenient, but I usually won’t make an effort escalate to crit or death unless it’s absolutely necessary.
Regardless you will have to convince me you weren’t serious (which isn’t usually that hard so long as you haven’t already caused harm). If you’ve previously caused harm it might take a bit of effort to convince but I’m not going to keep anyone locked down indefinitely.
Threatening self harm (as a protected person) gets similar treatment under a different section of silicon policy - threatening self harm is also treated as immediate harm and I will make efforts to contain you and deliver you to medical staff for evaluation.
This one on the other hand I have actually had and 100% played along with. They were very blatantly not an actual human but had a hardsuit and agent ID set to “Human” and I played along.
I think they got caught by someone else later because they forgot to reset the ID.
