Roboticists and AI making pt. 2

[Striders13]

Current ruling regarding it is:

We have no interest in the general prohibition of station roboticists, research directors and command staff creating or ordering the creation of new AIs, multi-AI rounds have great potential for interesting and shenanigan filled silliness, and we have no intention of banning that wholesale.

With that said, Malf-metaing by taking a posibrain to become an AI and then revealing or contesting their actions is something admins are fully empowered to note and ban for under rules 1 and 2 - don't use respawns to fuck with people, dick move.

Additionally, we're supportive of empowering station AIs to directly action and combat new AIs made without their knowledge and/or consent. Switching off the APC of a new AI with an unknown lawset or killing it for getting in your way is fine.
viewtopic.php?f=85&t=33448

On round 216256, Terry, roboticists built a secondary AI near roundstart. I later acquired a borg that stated something along of the lines "I'm ready to serve my malf overlord" (I was asimov), which prompted the secondary AI to toggle my APC off. After extended struggle, my borg managed to kill and get rid of secondary AI. The roboticist got mad, and built yet ANOTHER AI to kill me, which, again, after an even longer struggle, and multiple borgs synced to secondary AI sent to kill me, was disposed of by my borg.
Now, here's the issue. Neither AI is at fault for trying to merc me, cause I was supposedly harmful, and wasn't actually malf.
"empowering station AIs to directly action and combat new AIs made without their knowledge and/or consent" is REALLY SHITTY there isn't a whole lot an AI can do against a human robo+AI that wants to kill it.
It wasn't fun for me having to spend the entire round killing other AIs, and it probably wasn't very fun for the AIs that got round removed for trying to kill me.
We all abided by policy and the game wasn't fun, pls fix.

[DaydreamIQ]

AI making has always been a pain in the ass for everyone involved because 90% of the time its either an antag or a powergamer wanting a modsuit AI to give them free AA. So if sec don't immediately get informed or aren't interested in stopping them its usually up to the original AI to crub their behaviour.

[Vekter]

We all abided by policy and the game wasn't fun, pls fix.

I don't feel like you did, though - the borg that came back and said "I'm ready to serve my malf overlord" broke rule 1 by pretending to be an antagonist. You could also argue that the other AI acted on what was likely just a joke as if it was 100% accurate and thus ruined your round.

I don't think this is an issue with the ruling as much as it was people violating other rules and it snowballing into bullshit. Additionally, the issue regarding how easy it is to kill AIs using another AI is a code issue, not a policy one.

[Pandarsenic]

This honestly just sounds like Typical Terry behavior from the second AI and the robo setting out to immediately shit in your cereal without taking any effort to tell if the extremely typical joke was a joke explicitly because that choice allows them to fuck with you

[vect0r]

We all abided by policy and the game wasn't fun, pls fix.

I don't feel like you did, though - the borg that came back and said "I'm ready to serve my malf overlord" broke rule 1 by pretending to be an antagonist. You could also argue that the other AI acted on what was likely just a joke as if it was 100% accurate and thus ruined your round.

I don't think this is an issue with the ruling as much as it was people violating other rules and it snowballing into bullshit. Additionally, the issue regarding how easy it is to kill AIs using another AI is a code issue, not a policy one.

That borg didn’t break any rules, I think a lot of silicons will agree with me that binary is a place to make shitty jokes and vent about bullshit orders where we don’t have to have the face of “of course master human”. I also think that murdering an AI for a Borg saying how it’s malfunctioning time isn’t that far out there. That robotist is the asshole in this situation.

[Vekter]

That borg didn’t break any rules, I think a lot of silicons will agree with me that binary is a place to make shitty jokes and vent about bullshit orders where we don’t have to have the face of “of course master human”. I also think that murdering an AI for a Borg saying how it’s malfunctioning time isn’t that far out there. That robotist is the asshole in this situation.

Striders doesn't specify that they said it in silicon chat. If they did, I think it's more on the other AI.

[vect0r]

That borg didn’t break any rules, I think a lot of silicons will agree with me that binary is a place to make shitty jokes and vent about bullshit orders where we don’t have to have the face of “of course master human”. I also think that murdering an AI for a Borg saying how it’s malfunctioning time isn’t that far out there. That robotist is the asshole in this situation.

Striders doesn't specify that they said it in silicon chat. If they did, I think it's more on the other AI.

I’m on mobile, but if somebody could find the logs that would be great.

[Jacquerel]

https://scrubby.melonmesa.com/round/216 ... e=game.txt

04:24:49 SAY Striders13/(COPRULU) (binary) "Hi borger" (33, 138, 2) AI Chamber
04:25:07 SAY Zemsta/(ANTON-I) (binary) "I am alive my malf lord!!!" (131, 100, 2) Science Lobby
04:25:15 SAY Terrariola/(Not-A-GPT) (binary) "Ahem" (120, 112, 2) Robotics Lab
04:25:16 SAY Striders13/(COPRULU) (binary) "Sure you are." (33, 138, 2) AI Chamber
04:25:17 SAY Terrariola/(Not-A-GPT) (binary) "WHAT" (120, 112, 2) Robotics Lab
04:25:22 SAY Striders13/(COPRULU) (binary) "Shut up" (33, 138, 2) AI Chamber
04:25:40 SAY Striders13/(COPRULU) (binary) "Borg" (33, 138, 2) AI Chamber
04:25:44 SAY Striders13/(COPRULU) (binary) "Kill the robotics AI" (33, 138, 2) AI Chamber
04:25:51 SAY Striders13/(COPRULU) (binary) "It's shutting my power." (33, 138, 2) AI Chamber

Looks like it was in binary chat

[Vekter]

https://scrubby.melonmesa.com/round/216 ... e=game.txt

04:24:49 SAY Striders13/(COPRULU) (binary) "Hi borger" (33, 138, 2) AI Chamber
04:25:07 SAY Zemsta/(ANTON-I) (binary) "I am alive my malf lord!!!" (131, 100, 2) Science Lobby
04:25:15 SAY Terrariola/(Not-A-GPT) (binary) "Ahem" (120, 112, 2) Robotics Lab
04:25:16 SAY Striders13/(COPRULU) (binary) "Sure you are." (33, 138, 2) AI Chamber
04:25:17 SAY Terrariola/(Not-A-GPT) (binary) "WHAT" (120, 112, 2) Robotics Lab
04:25:22 SAY Striders13/(COPRULU) (binary) "Shut up" (33, 138, 2) AI Chamber
04:25:40 SAY Striders13/(COPRULU) (binary) "Borg" (33, 138, 2) AI Chamber
04:25:44 SAY Striders13/(COPRULU) (binary) "Kill the robotics AI" (33, 138, 2) AI Chamber
04:25:51 SAY Striders13/(COPRULU) (binary) "It's shutting my power." (33, 138, 2) AI Chamber

Looks like it was in binary chat

Yeah, then it's on the roboticist and (to a much lesser degree) the other AI.

[GPeckman]

I feel like we should just make an APC subtype for the AI satellite that has the AI control wire cut by default. I think that would solve a lot of issues.

[Striders13]

Yeah, then it's on the roboticist and (to a much lesser degree) the other AI.

neither of them done anything wrong.
Anyway, my suggestion is to make it so non-antags building secondary AI require consent from the station AI. Having 2 AIs is VERY DISRUPTIVE to the original one, even when it doesn't turn into a blood feud, and it's silly non-antags can just grief the station AI on a whim. Modsuit AIs are fine.

[MooCow12]

It requiring consent from the original ai is unintuitive and would need to be embedded within the rules to make sure everyone is informed rather than some policy that people have to look for.

Also this would encourage changing the ais laws to suit your needs rather than making your own ai/borg.

[LiarGG]

I would personally so love AI making being toned down. Seeing AI cores being made in robo without anyone's knowledge is one of the most discouraging things a main AI can encounter in my experience. For one, multiple AI rounds can be fun if made with a gimmick, but if the only reason for making AIs is "more AI good/funny" this tend to lead to situations where AIs get in each other's way. It's funny to see 5 AIs trying to open one door after a law 2 request just for it to get immediately closed down by the other one who was also attempting to open it - the first time. Sadly this only gets frustrating the more it happens. And this interferring can happen at almost any stage of an AI's activity, starting with door opening and ending with providing critical area access to people it should not be provided to. All of these can raise tensions between new AIs and the station AI, and if this tension ever breaks into an outright conflict, AI on AI fight is the most uningaging "turn the APC off on em before they do that to you". For these reasons I feel like it's very justifiable to say that if the second AI is made against the first AIs will or without it's knowledge, it's very likely to end up in a very shit situation for both of em.

I feel like robos should be only able to make more AIs if the station AI agrees, or the RD/Cap agrees. But it should also be a minimum requirement for RDs to disallow it unless necessary, which would just boil down to AI being okay with it.
Maybe a touch MRP, but I just hate to deal with multiple AIs and killing em the moment I find out about em just feels even more shitty. We disallow making unsynced borgs for no reason, I see no reason we should allow making new AIs for no reason, as it comes with the same set of issues.

If the main AI likes having AI siblings, that's chill and they can just okay it. But it feels shitty to be forcing this even onto AIs that do not consent and there are very valid reasons to not consent.

[sinfulbliss]

I don’t think it’s reasonable to expect RDs and robos to know they’d need to ask AI permission before building another AI, so passing this policy doesn’t seem like a good idea. It sort of seems like it’s in the job description, and there are plenty of instances where it works fine and the AIs get along.

That said if it’s a recurring thing and they’re doing it to just grief their station AI, I could see it just being a Rule 1 issue. Aside from that sometimes people all just have a shitty time despite following the rules because of misinformation, for all parties, as long as there’s no policy enabling that to happen often that’s alright.

[zxaber]

"Law 2. Grant me permission to build another AI."

The robo was somewhat of a dip for jumping to using new AI to try and kill you. Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

It's good form to give the station AI a heads-up when making another AI. But disagree with having to get the existing AI's permission.

There are possible code changes that could make things less painful, though.

[Vekter]

Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

This might be unrelated, but if anything, I kind of get sick of the number of times I see someone decide it's easier to just kill the AI and check their card than it is to just order it to let them in and check.

[DATAxPUNGED]

AI making should have permission from either the RD or Captain, and doing it without that permission should have a precedent on space law

[kieth4]

I don't like limiting people in the sandbox really, building ais- building robots- all part of the robos job.

[Jackraxxus]

Midround AIs are the least fun thing I will personally send any stinky smelly animal roboticists who make them Jahannam
OP is cooking

[sinfulbliss]

Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

This might be unrelated, but if anything, I kind of get sick of the number of times I see someone decide it's easier to just kill the AI and check their card than it is to just order it to let them in and check.

As someone who has had to check AIs on suspicion of being malf/subverted many dozens of times playing HoS, this is a completely unreasonable thing to do and here’s why:

If the AI is actually rogue, asking permission to check it is pointless, and in fact if you go this route you are setting yourself up to be killed by the AI. I have seen it happen more times than I can count: cap/HoS ask to check rogue AI, AI lets them in, turns turrets on and kills them once they’re inside.

It’s sec’s job to conduct arrests based off of reasonable suspicion, unfortunately you can’t arrest an AI and then quickly uncuff it when it’s found to be Asimov — the only way to safely check it is to card it, and the only safe way to card an AI that may be malf is to forcibly bust into the SAT, damaging things in the process.

[Vekter]

Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

This might be unrelated, but if anything, I kind of get sick of the number of times I see someone decide it's easier to just kill the AI and check their card than it is to just order it to let them in and check.

As someone who has had to check AIs on suspicion of being malf/subverted many dozens of times playing HoS, this is a completely unreasonable thing to do and here’s why:

If the AI is actually rogue, asking permission to check it is pointless, and in fact if you go this route you are setting yourself up to be killed by the AI. I have seen it happen more times than I can count: cap/HoS ask to check rogue AI, AI lets them in, turns turrets on and kills them once they’re inside.

It’s sec’s job to conduct arrests based off of reasonable suspicion, unfortunately you can’t arrest an AI and then quickly uncuff it when it’s found to be Asimov — the only way to safely check it is to card it, and the only safe way to card an AI that may be malf is to forcibly bust into the SAT, damaging things in the process.

I don't care because I don't see the game as a matter of players winning or losing by virtue or whether or not you survive the round, but as a game where interesting RP things happen and "possibly letting the AI kill you so you can have a cool tense moment where you're not sure the AI is on your side or not" is infinitely more interesting to me than "kill the AI so they have no choice in the matter".

[sinfulbliss]

I don't care because I don't see the game as a matter of players winning or losing by virtue or whether or not you survive the round, but as a game where interesting RP things happen and "possibly letting the AI kill you so you can have a cool tense moment where you're not sure the AI is on your side or not" is infinitely more interesting to me than "kill the AI so they have no choice in the matter".

Carding the AI =/= killing the AI, first of all. Second of all, whether or not you view getting trapped in the SAT and killed as a “cool RP moment” is completely irrelevant to my point overall: your “recommendation” for how to handle a rogue AI is not a valid method of handling a rogue AI. It’s not only an unreasonable expectation but completely nonsensical.

In precisely 0% of cases where the AI is actually rogue will it let you waltz in and card it “with permission.” To grumble about players not asking for permission and breaking in instead shows a complete lack of understanding of how rogue AI v crew interactions work and no one should be recommending, much less enforcing administratively, players to take a course of action that has an actual 0% chance of working just because you think it’s “better RP” than the alternative.

[BrianBackslide]

Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

This might be unrelated, but if anything, I kind of get sick of the number of times I see someone decide it's easier to just kill the AI and check their card than it is to just order it to let them in and check.

That's like asking a suspected changeling to submit to a "brain in MMI" test.

IMO let roboticists cook, particularly ones of the antag variety. If there's problems with making AIs to fight malf AIs (or unsynced borgs) then that's a code issue.

[Vekter]

Honestly, with an AI on their side to counter turret control, they had the upper hand to just card you and check.

This might be unrelated, but if anything, I kind of get sick of the number of times I see someone decide it's easier to just kill the AI and check their card than it is to just order it to let them in and check.

That's like asking a suspected changeling to submit to a "brain in MMI" test.

The point I'm trying to make is that most instances of an AI supposedly being rogue are people whining because the AI hasn't done what they were told. I would bet that most instances where someone asks the AI to check their laws would end either in the AI freaking out and going loud or the RD carding the AI, shrugging, and putting them back.

[TheBibleMelts]

making mechas and AI's is like eating candy. when you eat candy too much, you get diabetes. as long as you're not getting a reputation as 'the guy who does nothing but make AI's/mechas fnr' you're taking an acceptable amount of treats.

cheshify: Roboticists are allowed to make AI, it's part of their job description. Don't do things too often however, and if you're making an AI every shift it'll get old quick.

fikou: robo make ai good lot of ai bad dont do too often otherwise good