Status and Future of the AI

[Pratepresidenten]

2 hours ago, Menroka said:

The only possible problem I can see are very robust antags, like changelings, because now they are protected under the "Safeguard sapient life on board" law. The AI would have to ensure, that they are captured alive, instead of killed. The AI could make an exception, when a live capture is "infeasible" and would "likely" lead to more dead sapients, however this is ambigous and will lead to problems at some point.

It wont really be a problem. The AI/borgs arent meant to kill people, thats the crew's job :^)

The AI/borg wouldnt go out of its way to ensure the target is taken alive, because law 2 and 3 would just cancel eachother out, making the synth take no action.

[StationCrab]

This is great. 

[Menroka]

I think, I understand it a little bit better now. That "assist" in law 2 is pretty crucial in this manner, since it prevents any kind of interference with the crew, unless stated otherwise by someone higher in the chain of command. I think, the laws are pretty good!

If cyborgification was an option, that would apply to the crew as well, making cyborgification of Assistants/Visitors/any non-crew a tempting course of action. This can even be justified under law 2, since Cyborgs have the same laws as the AI, therefore Cyborg are assisting the crew, therefore turning Assistants/Visitors/any non-crew into Cyborgs is assisting the crew. This can ofcourse be stopped by someone saying no, but it would be kind of annoying.

Edited January 7, 2021 by Menroka
changed Assistants to Assistants/Visitors/any non-crew

[Chada1]

Removing the AI and 'borgs from being capable of harming most antags is a good change since it'll force 'borgs/AI to try more diplomatic approaches against them, but also please don't forget my law changes, they're super important to consider

[Butterrobber202]

16 hours ago, GreenBoi said:

May I suggest borgification, in these trying times?

Are borgs considered "life"? According to previous stances, there are equipment.

[StationCrab]

+1

[Scheveningen]

The one thing I find sad about this upcoming change is that it most likely cannot come any faster. 

[Cnaym]

"This lawset does not prevent me from abusing crew to change the AIs laws and download them if required as such I approve of it." -Kaiser

Going to apply with my bartender AI and a possibly a custom sprite huehuehue

[SierraKomodo]

Are there plans to open up the application process for people to apply and get approved before the whitelist goes live, or will applications not open until after whitelisting is live on the server?

[Arrow768]

Yes, the idea is to open up the whitelist process before the whitelist is required on the server.

[Capesh]

Oh god oh no the AI/borg main from years past return.

 

Cough.

 

So! I took the short time of reading through the suggested new AI lawset.

 

Everything but one is fine. Specifically, law 3: "Protect all sapient life on the station."

But ORB, you ask, what's the issue with sapient life?

 

Well every antag is sapient. Nukies? Sapient. Wizards? Sapient. Vox? ... Arguably. Pirates?

You get the picture.

 

One wonders why AI mains weren't poked about it. Maybe they weeeereeee

 

Anywho, it's an easy fix. "Protect all registered crew on the station."

 

Even gives incentive for people to actually fake-add their own crew records. Joy paperwork.

 

Thank you for coming to my TED talk. Beep Boop.

[jrphoenix303]

How will this affect the Malf AI gamemode? I'm assuming even enabling the antag in the options will be barred behind the whitelist, or will something entirely different (e.g. removing malf altogether) be instated instead? Once this is enacted, it'll certainly lower the amount of Infiltration secret rounds, at the very least.

[Tailson]

Capesh has a very important point that should be considered. I read this thread before and for some reason never noticed.

Interpreted correctly the 4th law should be used to prevent any and all conflict that could hurt any being capable of acting with appropriate judgement. This would mean stopping fights, getting in security's way, etc. There isn't a clause in there for exceptions, so even if someone has murdered 20 crew, the AI must still safeguard them from security's attempts to harm them.

Limiting it to manifested station crew would work just fine. Maybe even include "and those designated by command staff" as a caveat in the case of visitors etc.

[Kintsugi]

Law three is unacceptably vague. A horror form changeling could kill the entire crew, and the AI would be obligated to protect it from the HAPT kill-team sent to get rid of it, so long as it could demonstrate sapience.

[Triogenix]

8 hours ago, Tailson said:

Capesh has a very important point that should be considered. I read this thread before and for some reason never noticed.

Interpreted correctly the 4th law should be used to prevent any and all conflict that could hurt any being capable of acting with appropriate judgement. This would mean stopping fights, getting in security's way, etc. There isn't a clause in there for exceptions, so even if someone has murdered 20 crew, the AI must still safeguard them from security's attempts to harm them.

Limiting it to manifested station crew would work just fine. Maybe even include "and those designated by command staff" as a caveat in the case of visitors etc.

9 hours ago, Capesh said:

Oh god oh no the AI/borg main from years past return.

 

Cough.

 

So! I took the short time of reading through the suggested new AI lawset.

 

Everything but one is fine. Specifically, law 3: "Protect all sapient life on the station."

But ORB, you ask, what's the issue with sapient life?

 

Well every antag is sapient. Nukies? Sapient. Wizards? Sapient. Vox? ... Arguably. Pirates?

You get the picture.

 

One wonders why AI mains weren't poked about it. Maybe they weeeereeee

 

Anywho, it's an easy fix. "Protect all registered crew on the station."

 

Even gives incentive for people to actually fake-add their own crew records. Joy paperwork.

 

Thank you for coming to my TED talk. Beep Boop.

54 minutes ago, DanseMacabre said:

Law three is unacceptably vague. A horror form changeling could kill the entire crew, and the AI would be obligated to protect it from the HAPT kill-team sent to get rid of it, so long as it could demonstrate sapience.

I'm super tired so sorry in advance if I massively misinterpreted what these new laws mean, but it seems that people are forgetting the other laws exist. The AI cannot hinder security or the crew in fighting a horror form, because that would go directly against law 2 of Serving and Assisting them. At the same time, the AI cannot actively assist security and the crew due to the law 3, protecting all sapient life, but it can still help those who wish to run from it, or don't seek to harm it. Given this, I think a major point of the new laws, is the fact that AI can no longer assist security in tracking down, detaining, or killing antagonists. But it still also can't actively hinder those efforts either, like safeguarding a murderer. It seems to be deliberately crafted so the AI can pretty much not interact with either security or the hypothetical antagonist besides possibly opening doors for EMT's to drag wounded out and over to medical. This is not making the AI protect antagonists, this is removing the AI completely from nearly all security-antagonist interactions, which while being a possible issue, it is completely different then the perceived issues I saw in other responses and quoted here.

An example. JoeMcJoe, Officer, asks the AI to open a firelock on a window so he can shoot a traitor currently hiding in that room. The AI's laws now contradict each other, laws two and three specifically here, and when laws contradict, the end product is inaction. So now, Officer JoeMcJoe needs to go find an engineer or crowbar the shutter open himself, and if he does get it open, the AI can't close it because that would not be serving and assisting the crew, so then Officer JoeMcJoe shoots the traitor. The end result is the same as before, just with more effort required on the part of the crew to accomplish their objective of shooting the traitor.

Edited February 22, 2021 by Triogenix
added on why I even quoted those people specifically pt. 2 electric boogaloo.

[Pratepresidenten]

45 minutes ago, Triogenix said:

The AI cannot hinder security or the crew in fighting a horror form, because that would go directly against law 2 of Serving and Assisting them. At the same time, the AI cannot actively assist security and the crew due to the law 3, protecting all sapient life, but it can still help those who wish to run from it, or don't seek to harm it. Given this, I think a major point of the new laws, is the fact that AI can no longer assist security in tracking down, detaining, or killing antagonists. But it still also can't actively hinder those efforts either, like safeguarding a murderer. It seems to be deliberately crafted so the AI can pretty much not interact with either security or the hypothetical antagonist besides possibly opening doors for EMT's to drag wounded out and over to medical. This is not making the AI protect antagonists, this is removing the AI completely from nearly all security-antagonist interactions, which while being a possible issue, it is completely different then the perceived issues I saw in other responses and quoted here.

Pretty much nailed it.

[Chada1]

I'm really glad that the AI/'borgs are being removed from Antag/Crew interactions as well, it was what me and Amunak were hoping to do in a project before, actually, but it was discontinued due to really really negative response, I hope it works as well as I foresaw it working there in this case as well, I'm p. sure it'll drastically improve borg and AI for the better.

[Capesh]

2 hours ago, Triogenix said:

I'm super tired so sorry in advance if I massively misinterpreted what these new laws mean, but it seems that people are forgetting the other laws exist. The AI cannot hinder security or the crew in fighting a horror form, because that would go directly against law 2 of Serving and Assisting them. At the same time, the AI cannot actively assist security and the crew due to the law 3, protecting all sapient life, but it can still help those who wish to run from it, or don't seek to harm it. Given this, I think a major point of the new laws, is the fact that AI can no longer assist security in tracking down, detaining, or killing antagonists. But it still also can't actively hinder those efforts either, like safeguarding a murderer. It seems to be deliberately crafted so the AI can pretty much not interact with either security or the hypothetical antagonist besides possibly opening doors for EMT's to drag wounded out and over to medical. This is not making the AI protect antagonists, this is removing the AI completely from nearly all security-antagonist interactions, which while being a possible issue, it is completely different then the perceived issues I saw in other responses and quoted here.

An example. JoeMcJoe, Officer, asks the AI to open a firelock on a window so he can shoot a traitor currently hiding in that room. The AI's laws now contradict each other, laws two and three specifically here, and when laws contradict, the end product is inaction. So now, Officer JoeMcJoe needs to go find an engineer or crowbar the shutter open himself, and if he does get it open, the AI can't close it because that would not be serving and assisting the crew, so then Officer JoeMcJoe shoots the traitor. The end result is the same as before, just with more effort required on the part of the crew to accomplish their objective of shooting the traitor.

 

1 hour ago, Pratepresidenten said:

Pretty much nailed it.

So to combat valid-hunting (that in of itself the result of poor AI play) you disable action altogether. 

 

While I personally foresee issue with this, I suppose we can only really evaluate how good or bad this decision of lawset is after whitelists have utilized it for some time. 

 

Now if we could resprite the block of moldy cheese that is the ubiquitous AI core look... but that's another thing. ❤️

Edited February 22, 2021 by Capesh

[Carpe Venenum]

On Discord, the "Safeguard sapient life" law was discussed as being too open to law-lawyering. While maneuvering around the laws is against the rules, reading the law does make questions arise regarding protecting the crew which could not be considered sapient or life, such as IPCs and Bound Vaurca (which lack self-awareness). I suggest the law is reworded a little to avoid these arguments in the future, for example:
Safeguard all sapient life on board, including the crew and all visitors.

Edited February 22, 2021 by Carpe Venenum

[Estutes]

As someone who has never played AI but only synthetic units I'd say that these additions would make the game less spicy as it removes possible interactions between the AI and crew and I do not like that.
Although I would make an exception on this if it becomes standard on higher alert levels for command staff to change AIs laws to the more aggresive NT presets. I think the wiki explains very well how as an AI you should handle things as to not give away the antags intentions but to just hint your capabilities and answer the exact thing they are asking. I think @Faye <3 is right on that the AIs are not very common and I also fear that the whitelist might worsen it.

I also think that lore wise It doesn't make sense that there isn't a motion alarm in the vault. Also I think from a game mechanics perspective It's pointless as what's the point of the existance of a vault if you can't be detected inside.

Also if you are required to as an AI to "Safeguard sapient life on board" but you are also required to"Safeguard and ensure only authorized personnel gain access to areas of high security or importance to the station and its operation." there's no point in the AI having turrets or electrifying doors capabilities as there is no scenario other than someone subverting the AI that would allow it to use those. They would in any case remain de-activated.

About sapient life rule is very vague. Does this include incapacitated personnel? People with poor judgement? Under drug influence? These questions can be answered and resolved easily but the vagueness is there and the rule should be more specific and/or clear and not need explanations.

One thing I like about Aurora above every other RP server/community I've played in the past is the way you can get into very high quality RP with little to no effort in regards of forums, posts, whitelists, applications and OOC stuff that does not involve direct In Game actions. I understand command staff are to remain under whitelist and maybe you want whitelists for certain species as to maintain certain quality of RP and lore level regarding those. I get it. But adding more whitelists I think It is not the right path and at least for me it would lower how I value Aurora RP experience.

[Arrow768]

AI whitelist applications are possible starting today.
The whitelist will be activated on 2021-03-01.

[Chada1]

I'm so happy that the AI is finally whitelisted, not to be over dramatic but it feels as if I've finally been able to see Aurora take a first step towards finally eliminating one of the worst problems with Stationbounds, which is unvetted play of the AI who commands them.

The laws still might need some adjustment like what Carpe said.

[SierraKomodo]

Something that crossed my mind going over the AI wiki regarding the third law. The wiki says: "Station pets and lab monkeys - among other sapient test subjects - fall under this law as well". Does this mean all creatures that are living or is it explicitly only monkeys and designated pets?

[jrphoenix303]

On 24/02/2021 at 14:53, Chada1 said:

I'm so happy that the AI is finally whitelisted, not to be over dramatic but it feels as if I've finally been able to see Aurora take a first step towards finally eliminating one of the worst problems with Stationbounds, which is unvetted play of the AI who commands them.

The laws still might need some adjustment like what Carpe said.

To be fair, when the borg comes first and the AI second, they remain unbound and can quite literally tell the AI to fuck off in robot chat.

[Chada1]

On 25/02/2021 at 22:05, jrphoenix303 said:

To be fair, when the borg comes first and the AI second, they remain unbound and can quite literally tell the AI to fuck off in robot chat.

This is theoretically true but untrue in practice -- You still have to justify why you're ignoring the AI, you're just not compelled to do what they ask, if what they tell you is helpful and within your 4 laws, you still have every reason to listen to them. But I mostly meant the AI being awful (Validhunting, not acting appropriately to Crew) makes playing 'borg very unfun when it happens.