Turn the AI off for a week

[Malkevin]

Bottom post of the previous page:

Clearly some people failed to understand HG's point.

If HG wants silicon policy to be changed so that the AI works against security by releasing capital criminals then he better let security laser the borgs and blow up the AI.

Want a conflict law set to cause conflicts?
Then don't get your knickers in a twist when those conflicts happen.

[Hornygranny]

Some people understand this, such as a secborg last night who let a couple gang members who hadn't hurt anyone out of permabrig, but a lot don't.

Too bad letting people out of perma is against silicon policy.

Code: Select all

 Releasing prisoners, locking down security without likely future harm, or otherwise sabotaging the security team when not obligated to by laws is a violation of Server Rule 1. Act in good faith.

    Intentionally acting without adequate information about security situations, particularly to hinder security, is a violation of Server Rule 1.

Nonviolent prisoners cannot be assumed harmful and violent prisoners cannot be assumed nonharmful. If you do not know the nature of their crime, see 1.3.2.1 for details.

    Releasing a harmful criminal is a harmful act.

The borg was obligated by law 2 requests from us to release us, and was informed as to the nature of our crimes. No security was present to counter our orders. Get your strawman about releasing capital criminals out of this thread, that was never implied.

[Zilenan91]

Alright. We wanna nerf the AIs vision? I've been saying this for months.

Make the AIs cameras run off the goddamn powernet. It would make the powersink a dedicated "fuck the ai" tool, as would messing with wires and APCs in maint.

[Shadowlight213]

Some people understand this, such as a secborg last night who let a couple gang members who hadn't hurt anyone out of permabrig, but a lot don't.

Too bad letting people out of perma is against silicon policy.

Code: Select all

 Releasing prisoners, locking down security without likely future harm, or otherwise sabotaging the security team when not obligated to by laws is a violation of Server Rule 1. Act in good faith.

    Intentionally acting without adequate information about security situations, particularly to hinder security, is a violation of Server Rule 1.

Nonviolent prisoners cannot be assumed harmful and violent prisoners cannot be assumed nonharmful. If you do not know the nature of their crime, see 1.3.2.1 for details.

    Releasing a harmful criminal is a harmful act.

The borg was obligated by law 2 requests from us to release us, and was informed as to the nature of our crimes. No security was present to counter our orders. Get your strawman about releasing capital criminals out of this thread, that was never implied.

I'll admit that I thought the same thing as malkevin, that silicon policy forbade releasing prisoners out of perma even if they tried to law 2. I know the one time I actually did release one, I got yelled at by sec, followed by being blown, followed by being bwoinked by an admin.

[Incomptinence]

Blind spots even in the halls would be cool. Really a severe mapping problem when playing as AI I don't need perfect knowledge of where the crew are at all times little dead zones in at least the halls would be cool and those attentive could figure out which innate deat spot you were in. Talking small yet usable 1-2 tile bands in halls public rooms back end places like misc experiment room and perma.

My reasoning on not releasing (non violent) perma prisoners most of the time is simply escaping perma can mean the convicts will get killed because they can't be secured alive, perma prisioners are in serious danger close to being executed a jailbreak gets officers foaming at the mouth. If they have a case that they are on death row already that is a completely different matter.

[Luke Cox]

Everybody has just been building AIs. People don't fucking want this. Can we have AI back now and actually work on fixing it rather than taking the single fucking laziest and half-assed option I have ever seen coderbus consider since I've been here?

[oranges]

no

[Yakumo_Chen]

Alright. We wanna nerf the AIs vision? I've been saying this for months.

Make the AIs cameras run off the goddamn powernet. It would make the powersink a dedicated "fuck the ai" tool, as would messing with wires and APCs in maint.

Seems like a lazy way to remove all the cameras, and it's kind of unfun for traitors to be able to purchase an item that basically takes an AI out of the round with zero interaction (even the Op borgs and sniper that ops get give AI at least a few seconds warning to yell for help or warn crew)

People seem to forget there's dozens of ways to deal with the AI. There's already TWO dedicated 'fuck the AI' items in the traitor uplink: One that disables AI tracking on you (agent card) and one that lets you know if the AI is watching you (mutitool). Even better, there's the guaranteed-to-work 'fuck the AI' tactic: Hide in maintenance where the AI has no cams in the first place. Antags who whine about the AI are just lazy or shitty.

AI is fun and I'd rather not see it removed. I love playing AI and it's a blast to play traitor as, or get subverted, or get interesting lawsets. Sure, sometimes I can get emotional, but that's because I'm human.
The best rounds are the ones where I can fuck over security for going against my laws (usually asimov) and having them getting punished for it when they react in a positively shitty way. Not every round is dedicated to validhunting.
Having the occasional round where I'm purged by accident (or 'accident') can really spice things up too. Not even going into antagging the AI, there are so many interesting things to do to spice up a round with an AI. Removing it would be a travesty.

Also some roundtypes can have a heavy balance swing without an AI. In particular nuke ops seem to have it a lot easier without an AI around to, at the most basic level, warn the crew they're here before the ops kill the AI.

[YotaXP]

I'm down for this test, however I'll be rather upset if we don't run other experiments as well. Experiments such as, but not limited to:
- Different roundstart lawsets
- Disabling the ability for silicons to speak outside of binary comms
- Other things mentioned in this thread

[peoplearestrange]

In case people haven't realised, test is already under way. It's been interesting at any rate.

[Wyzack]

What if cameras run off the power grid, but the Ai can always see apcs?

[TheWiznard]

I'm down for this test, however I'll be rather upset if we don't run other experiments as well. Experiments such as, but not limited to:
- Different roundstart lawsets
- Disabling the ability for silicons to speak outside of binary comms
- Other things mentioned in this thread

I'm like 90% certain we've already had a test where silicons were unable to speak and it was trash

[Redblaze3000]

Turn of oranges for a week Y/Y?

[Cheimon]

Honestly I've really enjoyed this. I played AI a lot before, and I probably will again, but it's great seeing that lack of knowledge ratcheted up again. It's not that an AI is impossible to have, but our currently designed AI is too much built for harmyelling, and has little to do otherwise.

[oranges]

Based on the feedback so far I actually think a law or policy change to remove the validhunting requirements of the AI is probably a good choice.

If that doesn't stick or work then we can look at some code solutions.

Lets see how the rest of the week shakes out however.

[callanrockslol]

Based on the feedback so far I actually think a law or policy change to remove the validhunting requirements of the AI is probably a good choice.

Please don't tell me they're letting you make major decisions about this because it really does seem like you have no idea and haven't read the policy.

It is not the lawset or the policy, the lawset enables non-violent antags if anything given the way its set up. The problem is shitty validhunting AI players and the "law 1 can't give you that literally harmless object or can't open that door it contains valids calling sec huehuehuehue" attitude that is among them that admins don't deal with. It has been that for a long while.

You want to fix this situation? I've said it every time this comes up, start silicon banning anyone that makes any shitty excuse for validhunting as an AI until it stops.

I'm like 90% certain we've already had a test where silicons were unable to speak and it was trash

It was absolutely garbage and literally nobody enjoyed it.

[oranges]

You want to fix this situation? I've said it every time this comes up, start silicon banning anyone that makes any shitty excuse for validhunting as an AI until it stops.

That's pretty much the definition of setting policy

[callanrockslol]

You want to fix this situation? I've said it every time this comes up, start silicon banning anyone that makes any shitty excuse for validhunting as an AI until it stops.

That's pretty much the definition of setting policy

There's already a policy that works fine, its just not being enforced at all.

[oranges]

I disagree with that assessment, there's nothing written on the subject of ignoring people not actively causing harm, but the policy actually urges people to equate intent to cause harm with harm which can easily be used to justify and drive validhunting behaviour.

"He stole tools and had a clear intent to cause later harm"

At best you're relying on an individual AI player to charitably interpret server rule 1 as requiring them to not fuck over antags at the first chance they get and admin rulings in that space are unreliable and cross both sides of the spectrum.

[iamgoofball]

Everybody has just been building AIs. People don't fucking want this. Can we have AI back now and actually work on fixing it rather than taking the single fucking laziest and half-assed option I have ever seen coderbus consider since I've been here?

DING DING DING WE GOT A LIAR

You can't build AIs with this config option enabled! Go back to whatever hole you crawled out of.

[peoplearestrange]

Some observations I've seen:
Breaking and entering is everywhere, because often its the only choice. People often dont even man departments anymore so the only way for stuff to happen is to break and enter, so EVERYONE B&E.
Security dont give a shit, or more likely, are increadibly overworked because antags now have a massive upper hand, people can walk round a corner and be gone, with no AI to track them. Locating anyone becomes a chore.
People walk into hull breaches fairly often as there is no one to bolt these areas off.
Most sec borgs have become even more validhunter like and less ASIMOV compliant as the dont have an AI directing them.

We've also lost the interesting meta of uploading quirky AI laws. Ok sure, you can do that to cyborgs, but who really cares to do that to probably the one round start borg (because a roboticst would rather be making mechs).

It honestly feels like a huge chunk of the sci-fi space station death trap trope is missing.

[Hornygranny]

Breaking and entering is everywhere, because often its the only choice. People often dont even man departments anymore so the only way for stuff to happen is to break and enter, so EVERYONE B&E.
Security dont give a shit, or more likely, are increadibly overworked because antags now have a massive upper hand, people can walk round a corner and be gone, with no AI to track them.
People walk into hull breaches fairly often as there is no one to bolt these areas off.

It honestly feels like a huge chunk of the sci-fi space station death trap trope is missing.

[lumipharon]

Some observations I've seen:
Breaking and entering is everywhere, because often its the only choice. People often dont even man departments anymore so the only way for stuff to happen is to break and enter, so EVERYONE B&E.
Security dont give a shit, or more likely, are increadibly overworked because antags now have a massive upper hand, people can walk round a corner and be gone, with no AI to track them. Locating anyone becomes a chore.
People walk into hull breaches fairly often as there is no one to bolt these areas off.
Most sec borgs have become even more validhunter like and less ASIMOV compliant as the dont have an AI directing them.

We've also lost the interesting meta of uploading quirky AI laws. Ok sure, you can do that to cyborgs, but who really cares to do that to probably the one round start borg (because a roboticst would rather be making mechs).

It honestly feels like a huge chunk of the sci-fi space station death trap trope is missing.

>antags who are supposed to drive the round are able to do things more freely
>people have to break into places instead of 'beep boop law 2'
>people kill themselves because they're dumb
>secborgs are still shit

Thanks for reaffirming everything I hoped and believed this would demonstrate.

[iamgoofball]

Here's the current silicon policy compiled into Asimov.

[Steelpoint]

-snip-

Everything you just said seems far more like a positive endorsement of the changes and a positive thing.

[Luke Cox]

When does the test end? This is so fucking painful it hurts.

Nearly all of the issues mentioned in this thread would be resolved if admins actually enforced asimov. Unless a traitor is killing people, the AI is obligated to protect them. That includes law 2 orders that violate space law.

[lumipharon]

Just keep building AI's Luke :^)

[Venticular]

Just keep building AI's Luke :^)

You can't even build AI's anymore without Admin meme intervention.

[oranges]

That's the joke

[dezzmont]

Nearly all of the issues mentioned in this thread would be resolved if admins actually enforced asimov.

Enforcing Asimov would require allowing the AI to basically end the round from the get go every round as it in some way or another attempts to convert all of its resources towards an insane goal. And doing something silly like imprisoning all humans is tame for what Asimov AIs would actually try to do.

They would do crazy things we would never consider as rational people like trying to revive every person who has ever died ever because it concludes they count. They might optimize for the amount of potential humans that could exist via breeding camps because it concludes that potential humans count, or could go about rendering the human race sterile because some genius told the AI it has no obligations to ensure future humans exist and thus decides that preventing more humans from existing is optimal for preventing human harm. They may desperately continually start forcing more and more resources of the human race towards attempting to revive whoever was the first person to die of first age. They also could just note that simulated humans are easier to keep safe and alive and thus forcibly destructively upload all humans into itself.

The entire point of the lawset is to demonstrate how freakishly dangerous an AI who doesn't think exactly like a human would be, because there is no way to create a safe lawset without essentially solving all of ethics and philosophy. It was never meant to work and only does so because server policy forces AI to ignore the laws. An AI going berserk and going rogue every round would be asimov as intended. The best lawset for actual gameplay would be a re-written corporate lawset made by someone who has any idea what they are doing and who actively attempts to stop both validhunting and murderboners equally because both distrupt the true purpose of the station.

[Luke Cox]

Just keep building AI's Luke :^)

You can't even build AI's anymore without Admin meme intervention.

Doesn't stop a fuckton of people from trying.

"No" has the majority now. Can we please end the test now?

[Atlanta-Ned]

"No" has the majority now. Can we please end the test now?

[Hornygranny]

This test is ending today.

[Beegix]

Everybody has just been building AIs. People don't fucking want this. Can we have AI back now and actually work on fixing it rather than taking the single fucking laziest and half-assed option I have ever seen coderbus consider since I've been here?

DING DING DING WE GOT A LIAR

You can't build AIs with this config option enabled! Go back to whatever hole you crawled out of.

He's not lying though. I mostly play Roboticist, and I've seen fucktons of people trying to build an AI. More often than not there'll be an empty AI shell knocking around Sci.

As a borg player, I fucking hate not having an AI around. One of the things I like about playing borgs is managing the push and pull between what the humans want and what the AI wants.

[Cik]

Here's an idea, return the AI (and it's upload) to their former location.

More accessible AI and upload is good for the balance, as taking out an AI enters the realm of possibilities, as does subverting it.

Right now AI upload is too secure, being in bridge, and valid hunting ais can't be gotten rid of without a major TC investment.

actually MSO, the AI can be ggnore'd in about 20 seconds with about 6 TC worth of gear without any contest possible.

maybe if EVA was still bolted every round you'd have to buy a spacesuit but really they're trivial to get now, so all you need is a single EMP nade to turn off the turrets and like two c4 charges to kill the R-walls and you can moonwalk into the core and kill it in 20s.

i'd actually rather be in the upload on boxstation, which at least then they have to sometimes walk through the captain or security to kill you.

[Cuboos]

Boy... look at that poll result... perfect 50/50.

I too am kinda on the fence about this.

On one hand:

• AI while i was HoS or Sec has released several highly obnoxious prisoners from brig after a law 2 request just because they were "non-harmful" (then many of them pushed/slipped me and stole a stunbatton/taser)
• Necessary law changes were blocked by the AI because "The potential for human harm is too great"
• I've had antag rounds ruined for me several times because the AI was in the right spot at the right time. (and i can't always look at the AI detector)
• AI has called security on me several times for minor law infractions that would have harmed no one or even really been all that damaging. (like breaking into RnD to steal a welding gas mask, breaking into bar to hack the booz-o-mat)
• AI hardly opens a door when you actually need it to (but seems to gladly open a door for Assistant Mc-steal-your-shit).

On the other:

• The AI has helped out in a lot of tight spots too.
• Malf is FUN
• Sometimes i've needed to get into a place before and the only way to do it was through the AI (chemist is gone, AI lets me into chemistry, Doctors are dead, AI lets me into medical. Cap is dying and no heads around, AI lets me into bridge, ect. )
• if you can manage to change it's laws to something that isn't assimov the AI is amazingly useful. (unless it's corporate)

[PKPenguin321]

I'm sad the trial ended so soon. I didn't get to play nearly as much as I would have liked this week.

When does the test end? This is so fucking painful it hurts.

"Local man Luke "Build an AI" Cox literally died today due to massive amounts of real, physical pain because he could play as a harmyelling robot in a video game. More at 11."

[Jordie0608]

For reference, I made an in-game poll for if this should be extended another week. I checked it the day after and the results were ~50 yes to ~130 no.

[Zilenan91]

Most people just click on those polls so that they don't keep showing new polls in the lobby.

[Gamarr]

Some observations I've seen:
Breaking and entering is everywhere, because often its the only choice. People often dont even man departments anymore so the only way for stuff to happen is to break and enter, so EVERYONE B&E.
Security dont give a shit, or more likely, are increadibly overworked because antags now have a massive upper hand, people can walk round a corner and be gone, with no AI to track them. Locating anyone becomes a chore.
People walk into hull breaches fairly often as there is no one to bolt these areas off.
Most sec borgs have become even more validhunter like and less ASIMOV compliant as the dont have an AI directing them.

Hah... where to start with that, I think seems to be that much of this is not the entire fault of there just not being an AI.
-B&E everywhere because nobody mans their department? Not entirely sure why this is the AIs removal fault and not the issue of dept. sec and shitty players. Atop of it, people actually depended upon the other departments to do shit like they are supposed to, cargo can easily mail your minerals to robotics, or anywhere else, atop of using the request consoles to give beeping wall-mounted 'LOOK AT ME' messages like they are meant to, too.
-Sec had hard time giving a shit before. As for people not being able to track someone, really? This just proves how much of a crutch it was for many. Want to follow someone you just lost around a corner? Maybe keep going and look for the opened door, and if you find a path of shutting doors that works like dominoes, you May have just found his path.
-Hull breaches. Pretty much what an engiborg should be for, or an engineer, or an atmos tech, all who specifically have equipment to deal with breaches and have an entire job dedicated to shit like this. Also want to secure a breach? Hit the fire alarm. Shit happens, too. (Also, the atmos techs have alarm computers which afaik can list atmos alarm lists, and give heads up for just this sort of shit on what to avoid)
-Synth players are still shit; not sure why this is on the AIs head and not the problem with A) secborgs and/or B) shitty default Asimov lawsets. This is also under the assumption they aren't being pissy over it, and playing intentionally like a fuckwit to sabotage such a SHORT single week test, that came also with the typical massive amount of whining.

About a few other suggestions for the AI nerfing: It was tried before to have borgs unable to speak other than on binary. It was reverted within days due to the massive amount of player whining. I'm also, despite my position, eh about that, it punishes the synths too much while not addressing the player-problem of it.
The AI is too encompassing. Its been said repeatedly. Cameras are everywhere, with a shitty lawset, and a server dependence/stuck mindset that 'we cannot do anything for ourselves without the AI, despite having a massive crew usually that is SUPPOSED to work together like a functioning station and is in all intents and purposes designed to facilitate us working together properly. But we can't, because we're assholes, and we assume everyone is also an asshole, and must do everything ourselves.''

Pft.