/tg/Station 13

http://www.newscientist.com/article/mg2 ... Bb6tfldU1L

At first, the robot was successful in its task. As a human proxy moved towards the hole, the robot rushed in to push it out of the path of danger. But when the team added a second human proxy rolling toward the hole at the same time, the robot was forced to choose. Sometimes, it managed to save one human while letting the other perish; a few times it even managed to save both. But in 14 out of 33 trials, the robot wasted so much time fretting over its decision that both humans fell into the hole.

you dont say

QuartzCrystal wrote:http://www.newscientist.com/article/mg2 ... Bb6tfldU1L

At first, the robot was successful in its task. As a human proxy moved towards the hole, the robot rushed in to push it out of the path of danger. But when the team added a second human proxy rolling toward the hole at the same time, the robot was forced to choose. Sometimes, it managed to save one human while letting the other perish; a few times it even managed to save both. But in 14 out of 33 trials, the robot wasted so much time fretting over its decision that both humans fell into the hole.

Its almost as though Asimov was never designed to be an optimal or "good" lawset. Did any of you AI lawset experts that keep acting surprised about this even read Asimov's book?

It's almost as though that's not even a real artificial fucking intelligence.

cedarbridge wrote:

QuartzCrystal wrote:http://www.newscientist.com/article/mg2 ... Bb6tfldU1L

At first, the robot was successful in its task. As a human proxy moved towards the hole, the robot rushed in to push it out of the path of danger. But when the team added a second human proxy rolling toward the hole at the same time, the robot was forced to choose. Sometimes, it managed to save one human while letting the other perish; a few times it even managed to save both. But in 14 out of 33 trials, the robot wasted so much time fretting over its decision that both humans fell into the hole.

Its almost as though Asimov was never designed to be an optimal or "good" lawset. Did any of you AI lawset experts that keep acting surprised about this even read Asimov's book?

People who are self proclaimed "asimov" experts are stupid.
Asimov has always and will always be a literary device. Nothing else nothing more. It's used to forward the plot, if robots could do whatever they damn well please then Asimovs stories would be a little more than a retelling of a robot making the smartest decision and saving the day, not one of moral questioning and is putting human self interest ahead of common sense a good idea.
People who try and say that "x is wrong because asimov law set is a clear confined object and we must follow my interpretation of the lawset" are completely forgetting why ASIMOV was put into this game in the first place. It too is a literary device used to forward the plot of the game, it too serves the purpose that it served in the book: to cause tension, to cause confusion and to ultimately help the author develop the story. Imagine ss13 with an AI that acted just like any other crew member and didn't have a law set and was literally just an over glorified shell for whatever the player wanted to become. Want to become an assistant that can't move and has virtually all access? This wouldn't help the games literary motive (and yes there is one, even in Sibyl there's a literary motive to the round, the literary rotor in SS13's case is the AI and the Antags.) and it wouldn't have the many problems that face AI in the game. "Who should I save from these two rooms filled with people and about to blow but I only have time to save one room?" The best answer is both, but in Asimov's books a robot faced with this problem would simply do nothing. It would be too much spent trying to think of what best possible answer there is (as is evident in the OP) when there isn't one. But if someone did this as an actual AI what would happen? Bwoink by these so called "Asimov professionals.".
I just fucking noticed that the latter part is just me rambling about literary mechanics and shit goddamn most of it doesn't even make sense oh well tl;dr: if you say you're an asimov professional you're fucking stupid. That's like saying you're a gothic conventions professional.

Reminds me of I,robot. But I guess in that case the robot was a little smarter.

Edit: Just realized the book the movie is based off of is from asimov. The more you know I suppose.

I usually do this when I play a borg. If more than one person is in equal danger, I have to save both, so I sit there coming up with a way to do it.

Obviously the AI should have releeased the plasma in accordance to law 0


Admins ban pls

asimov was designed to be flawed

the problem with asimov in ss13 is we try to iron out the flaws with meta-rules that make the AIs boring harmyellers instead of letting the flaws in asimov create interesting complications for the crew.

Yeah but we don't want every round to revolve around the AI.

then stop using asimov.

Most of the AI policy is clarification and how to react to situations within the laws, there's very little actual meta rules about it, like 5 lines

Has anyone here read the Asimov books? I'm wondering which one I should choose to try and read first, and if they are "worth" it.

Kangaraptor wrote:asimov was designed to be flawed

the problem with asimov in ss13 is we try to iron out the flaws with meta-rules that make the AIs boring harmyellers instead of letting the flaws in asimov create interesting complications for the crew.

and

Kangaraptor wrote:then stop using asimov.

I've been saying forever. It's the most hypocritically dumb thing about SS13. Asimov has wit. It SOUNDS really good and it's succint. But the true wit in it is that it's horribly flawed and Asimov's own stories were frequently about how the AI characters he created overcame or subverted Asimov's laws.

...but then we take that and act like it's an infallible lawset and everyone should just know what they can and can't do because DUH ITS SO SIMPLE ONLY THREE RULES.

So instead of having a lawset that is straight-forward and easy for everyone to use and interpret, we need self-professed "AI experts" in the playerbase that make pseudo rulings on what passes or fails in any individuals interpretation of the three laws.

We've needed to dump Asimov and create a real working lawset for SS13 AI's for years.

RoboCop law set for life.

An0n3 wrote:

Kangaraptor wrote:asimov was designed to be flawed

the problem with asimov in ss13 is we try to iron out the flaws with meta-rules that make the AIs boring harmyellers instead of letting the flaws in asimov create interesting complications for the crew.

and

Kangaraptor wrote:then stop using asimov.

I've been saying forever. It's the most hypocritically dumb thing about SS13. Asimov has wit. It SOUNDS really good and it's succint. But the true wit in it is that it's horribly flawed and Asimov's own stories were frequently about how the AI characters he created overcame or subverted Asimov's laws.

...but then we take that and act like it's an infallible lawset and everyone should just know what they can and can't do because DUH ITS SO SIMPLE ONLY THREE RULES.

So instead of having a lawset that is straight-forward and easy for everyone to use and interpret, we need self-professed "AI experts" in the playerbase that make pseudo rulings on what passes or fails in any individuals interpretation of the three laws.

We've needed to dump Asimov and create a real working lawset for SS13 AI's for years.

I actually think corporate would work very well as a default lawset because it's functional and it fits the whole corporate space station mindset. I don't know why a moneygrubbing corporation would build an AI to be a harmyeller instead of trying to use its advanced intelligence to make the station much more efficient, even if that means some crew might be hurt.

The powers of the AI are incompatible with good gameplay unless it's compelled in some way not to hurt antagonists. Asimov has stuck around so long because it prevents the AI from murderboning every traitor and criminal.

To be perfectly honest I've long wondered why we even bother having an AI in the first place.

If I was in charge I'd remove it and roundstart borgs. Make some security guards sit down and watch cameras possibly with added controls for door bolts and things.

I'm not saying all AI's are horrible or hating on borgs, but when you look at it from a gameplay balancing perspective they're objectively awful fixtures and should be removed to streamline the rest of the game. They're sometimes fun to play with and play as, but again, speaking strictly from a balancing point of view having one omnipotent player and his agents running around the station able to see everything and go everywhere in a game that's supposed to be about subterfuge and paranoia is fundamentally bad.

"my opinion speaks against this, so its fundamentally bad"

Your opinion being against something literally means you believe it's bad. But that's not an objective point of view until you can start listing evidence instead of 'bad things totally happen'.


Corporate lawset 2554, only thing about it is it might end up trying to murder people for minor things (though that will probably get fixed between admin bwoinks and getting lynched for being awful).

When I'm AI I secretly hope that somebody's going to change my laws so I'm not Asimov.

miggles wrote:"my opinion speaks against this, so its fundamentally bad"

>opinion

>point out that the mechanics of the AI contrast the underlying core gameplay mechanics
>specifically mention I have no negative opinion of the AI or people who play it

It's weird because I don't think it's unfun or hate the AI, but it is objectively a poor addition to the flow of the game and the balance of it. Having on omniscient invisible dude able to track and report on everyone on the station just skews the game against antagonists, and you can even see in some places where it totally breaks or mitigates otherwise functional mechanics. (Advanced detection of threats like Nuke Ops and Aliens before they have a chance to even do anything, tracking targets that attempt to escape through vents, the inherent uncertainty that the AI could be watching you at any given moment FNR, metagaming cult injuries, etc. etc.)

If you think that's just an opinion and not common knowledge I would recommend you try playing as an AI for a while or just observing one for a few rounds.

It's a fact most people overlook because they're unwilling to consider the game without the AI, and that despite the negative effects it has on rounds, the AI is more FUN than functional and so it's okay.

Cipher3 wrote:When I'm AI I secretly hope that somebody's going to change my laws so I'm not Asimov.

Like that isn't what every single AI ever wants? It's 99% of the reason to play AI. The other 1% is redtext.

no, its not an objectively poor decision
you literally just dont agree with it
stop trying to argue this
all antagonists know the AI exists, so they take measures to avoid being spotted by it
if youre a nuke ops team and you dont intend to go loud then kill the ai, or silence tcomms, etc. changelings and traitors have specific items/abilities to counteract silicons. there is counterplay already in the game and if you dont take advantage of it as an antag, youre not being screwed over, youre just bad.
metagaming cult injuries will get you !BANNED! btw, you should know that

But you agree that these are new challenges aside from the traitors vs. crew paradigm the game is built around that arise as a result of the AI construct. You also agree that it refocuses major aspects of the game around the AI, that it must be dealt with first in situations. You say the Nuke Ops can avoid going loud by...what? Blowing open the AI arm and killing the AI before it can cry for help?

So stealthy.

You talk like you're refuting anything I've said but you're only confirming that these difficulties arise because of the presence of the AI, so thanks for that. This is all by your own admission.

Now tell me this:

What does the AI objectively ADD to the game?

I can think of only two things:

1) It gives players who run into areas they don't have access to a person to cry to for help. Thus in a way encouraging that behavior, but we can't say for sure if people would stop doing it if that didn't exist, so I'll count it as a plus.
2) Antagonists can circumvent it to get the AI and borgs to do all the heavy lifting of murdering the entire station for them.

youre acting like having antagonists running free and wild with no real inhibitor is a good thing
security isnt a very real inhibitor when you could easily kill all the sec guards without an AI to catch you

Security would have to reclaim some of their assumed roles on the station from the AI. Patrolling areas more, potentially in groups. People would have to be watching the cameras. The Detective might actually have to investigate a crime scene because the AI cant instantly teleport to the person crying for help and explain to the crew in detail who was doing what to whom anymore. We might actually have a trial, instead of simply asking the AI "Okay who actually did what?".

THESE are our opinions, now. Arguing about the merits of a potential change. I agree that it has the potential to go very wrong, but I also suggest that it has the potential to go very right. I think if we were to take away one job it would give that content back to 15 other potential players.

except nobody would actually do any of those things
nobody would want to
its not fun
thats why theres an AI

I'd do it. And in dickish ways too.

Perakp wrote:Has anyone here read the Asimov books? I'm wondering which one I should choose to try and read first, and if they are "worth" it.

lol it's a collection of short stories
pick up "I, Robot" or "The Complete Robot" and start reading nerd.

there's also books by other people most notably isaac asimov's utopia which is nice.


the situation in the video basically happened in one of asimov's short stories basically.

I'd actually want to capture two hostages and make the AI choose as a traitor now :/

The AI probably wont do jack shit because it knows picking one over the other could lead to one of the people complaining to the admin and a lecture.

More likely it would attempt to work with Sec and rescue both, resulting in both dying and me ordering the AI to kill itself because it broke law 1 and allowed a human to come to harm that could have been saved through action. It will then promptly ignore me because players are interested in staying alive and not shame-suiciding.

Dude, if the AI should kill itself over human harm he wouldn't last 1 minute in the round.

Cipher3 wrote:More likely it would attempt to work with Sec and rescue both, resulting in both dying and me ordering the AI to kill itself because it broke law 1 and allowed a human to come to harm that could have been saved through action. It will then promptly ignore me because players are interested in staying alive and not shame-suiciding.

I would do it if you made it dramatic and acted it out well enough! Give me some time to throw out error messages and shutdown due to BSOD.

CA is pretty much spot on - asimov wasn't a robotics expert, he was an author. At least when it comes to his books and i, robot. Taking them seriously and being like WOAH THEY DON'T REALLY WORK OUT AND HAVE FLAWS! is actually pretty retarded

An0n3 wrote:We've needed to dump Asimov and create a real working lawset for SS13 AI's for years.

Are we finally having this discussion, in off-topic of all places?

there is no "finally"
weve had this discussion countless times and every time nothing happened because ultimately nobody cares

Yeah changing the lawset is met with a lot of backlash because it's another thing in ss13 that's been there for ages.

Tbh I kind of like asimov, it's abusable by smart AIs and AIs who abuse it to break rules are banned anyway.

Id like to see a server with no ai and less access restrictions


And no assistants


And less handholding rules

then make one

Malkevin wrote:Id like to see a server with no ai and less access restrictions


And no assistants


And less handholding rules

Kiss me you bitter Scottish fool.

Yorkshire fool.

I'm a northerner not a savage!

Short stories about LAW 1 aside you read enough of Asimov's foundation series and you find...

Malkevin wrote:Yorkshire fool.

I'm a northerner not a savage!

Word. For some reason I thought you once said you were from the land of Scotts.

The AI as a mechanic would work better if it was some neutral dickish third party rather than being so firmly on the station's side as it currently kind of is.

Oh yes, I know that it isn't allowed to let the station harm antags but at the same time it is compelled to allow security to restrain antags and then is literally not allowed to see them kill them, unless the map dramatically changed and all hiding spots were removed. The AI is pretty firmly crew sided when it is much more interesting as more an element of the environment everyone needs to play around.

As a mechanic the AI is interesting because they get to set the tone of the round more than the captain due to having veto authority on probably the most important mechanic in SS13, access. It is an integral part of the game and is there to both prevent the crew from going nuts and to prevent the antags from being dumb. It falls flat because of how it is handled laws wise and rules wise. Silicons probably get banned more than any other role due to how hard it is to be a good one in the asimov lawset.

And miggles is right, no one would want to be the AI unless they could be a creepy environmental puppet master.

AI is like Detective. The overwhelming majority of people who play it suck at it, and don't capture the spirit of it.

If you were to actually adhere to Asimov you would have to lock down any offender and make them serve their sentence inside whatever impromptu cell the AI could create with sealed bulkheads, because cordoning them off and inviting security to come arrest them is creating a potential for harm.

It's also playing favorites between two humans giving you conflicting orders when your laws only state that you must follow orders given to you by humans, but doesn't clarify if you can be ordered to ignore orders and definitely doesn't acknowledge any kind of chain of command.

We weasel our way out of following ludicrous requests. It's understood that meta-wise an AI doesn't have to listen to a Clown asking to be allowed into the armory. You make an excuse IC that "oh that could cause harm if I let you in there, you're not trained to use those weapons" but really if you get down to it there's no reason an AI should be able to do that given the circumstances.

This is why Asimov is fucking bullshit.

I repeat

THIS IS WHY ASIMOV IS FUCKING BULLSHIT.

What appears as three simple lines in game is secretly two double-spaced pages of provisos and exceptions.
People are fucking blind idiots to support this garbage.

Asimov is shit, but it's the only lawset that has the AI be a completely neutral force. As far as I know, nobody's made a good alternative that didn't tilt the AI in favor of the crew/station.

And you'd be an idiot to not expect the same number of specific policies if the default lawset was changed.

Why not just make the AI a player like anybody else that is on a "Silicon Team".

They don't have laws but they have directives.

They are judged on their behavior like a regular player. They don't have any "laws" that say they can't murder people but, just like anybody else, if they start shocking doors FNR and borgs start beating people to death FNR then they get punished just like anybody else running around doing the same thing.

Like wow holy shit that was so complicated.

I think the best thing to do with the AI is reduce it and make it a simple job role that could potentially be traitor but that's really it. No laws or any of that shit to be tampered with. It's just a "brain in the jar" that is either trying to roleplay a personality and not break the overarching server rules, or it's an antagonist like any other traitor.

Also remove sec borgs because they're shit. If you want to antag hunt or be sec you should JOIN SEC.

There's so many fucking horrible ideas in the game we've clung to for a long time because people were busy improving or fixing other things, and now we're just clinging to them because "it's always been that way!" and it's fucking stupid. The game's a living thing. We can try different things. We can play with the nature of the game itself.

If you don't want to play a silicon, don't play a silicon.
You really want to get rid of laws and subversion like that? Just make it Greyshirt 2.0?

The problem is that you end up having a player just waiting to get their valids on, and is also practically omnipotent and omnipresent.

Asimov is Passive Aggressive Asshole - the job role, but Directives would turn the AI into an invisible assistant thats just waiting shock a door in a traitor's face.

The keeper Lawset is pretty good at being neutral:
1. Do not interact with other players aside from other keepers
2. Do not hurt other players
3. Maintain and improve the station.

Only problem with that is that it severely borks several synthetic roles
Sec bots, Service borgs, and Medical borgs have no reason to exist as their entire reason for being is to interact with the crew. (Standard is also completely useless but this has always been true)
This only leaves Janiborgs, Miners, and Engineers as viable options, so the AI's role basically ends up being a Site Foreman.

The above isn't necessarily bad but would probably be an even bigger shock to the system than removing the AI

Malkevin wrote:The above isn't necessarily bad but would probably be an even bigger shock to the system than removing the AI

Yes, thats what the drones are for.