Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Mono124]

I dunno, to me that sounds interesting and fun.

The issue can also end with the AI refusing to let anyone into their upload for any reason whatsoever because upload access may lead to law changes that will conflict with current laws... even if absolutely nothing is going on in the station like the aforementioned scenario. To be entirely honest it all depends on how much of a dick/how lenient the AI is for that round, and how willing they are to abuse loopholes in the base laws to do whatever they want.

[MrWillsauce]

Eh, anyone can ruin an important role by being a dick. That's not unique to the AI.

[miauw62]

Haven't been on this thread for a bit so this is bringing up an old topic (AI harm-Asimov), but I just wanted to try and justify some aspects. I'll put it in spoilers so people who don't care don't have to look at it.

Spoiler: Can AI's harm people? (click to show/hide)

All I saw was attempted murder, an evil cult and a dead captain.

Get suffocated, evil person. :u

Unless you were non-asimov.... cult is still people

But if you read Asimov's "I Robot" you will see that robots can 'break' the first law in order to save more lives. The robots with Asimov's laws are able to allow the harm of or harm a human, if it prevents or might prevent the harming of more humans. Based on that, even asimov AI's can still harm or even kill traitors, syndicates, cultists, wizards and death squads if it would save more human lives in the end.

Actually, In I Robot it is explicitly stated that robots can cause a small amount of harm to humans to prevent a greater amount of harm. A robot could ram into a human, maybe breaking a few ribs, to prevent that human from being crushed by a rock. Killing humans, however, is not a small amount of harm and thus unacceptable.

In the story "Little Lost Robot" while they are trying to find one of the NS-2 "Nestor" robots that has hidden itself, they carry out a test. The test involves dropping a heavy stone onto a human and killing them (the robots don't know but the stone is rigged to stop). The room is filled radiation and the robots are told that, they will be unable to save the person and, should they try, will only destroy themselves which might prevent them from saving more humans in the future. When the rock is dropped, all the robots stay still (except the NS-2 as he could sense the radiation was harmless) even though they all (except the NS-2) had newtons 3 laws programmed into them without any alterations.
Based on that, an asimovian AI should be allowed to let a human die, or even kill them, if it might save the lives of future humans.

Spoiler (click to show/hide)

This was in a situation where the robots were influenced by the malfunctioning Nestor, who convinced them that there was no point in trying to save a human if they would be destroyed before they could reach said human. This also ONLY demonstrates the second part of law 1, since they are only letting the human be harmed, not directly harming the human.

Spoiler (click to show/hide)

The NS-2 Nestor unit wasn't malfunctioning, he was just following the second law and his purposely weakened first law. I'm also sure that Dr. Calvin, not Nestor, told the robots they would be destroyed if they tried to save her, and would prevent themselves from saving future people. Also, even though they only broke the second part of the first law, the implications are that they could break the first part as well in a similar circumstance. Breaking the second part should allow an AI to not open a door for a traitor or normal person that is suffocating, if they can save more people by opening other doors. Breaking the first part should allow an AI or robot should be able to kill a traitor who is about to blow up the medbay (with people inside) if it is the only way to save the numerous people inside the medbay.

Spoiler (click to show/hide)

No, the modified Nestor convinced the other robots that it would be futile to try and save the human in the second test, this was explicitly said.

And this situation demonstrates that robots can forego saving a human if they will ALWAYS AND WITH COMPLETE CERTAINTY die trying, thus achieving nothing but possible future harm.
This situation does NOT demonstrate how a robot could cause harm to prevent greater amounts of harm.

Also, baylaws for AIs are incredibly unfun. AI should be neutral, not mechanized validhunter.

[Mono124]

Eh, anyone can ruin an important role by being a dick. That's not unique to the AI.

What is important to the AI is that they can ruin an important role by being a dick and get away with it, at least sometimes. If someone ahelps and is pissed at the AI due to it, say, locking down the upload because it doesn't want anyone uploading laws, the player of the AI can try to justify it by saying that they were just following their laws. We all know it's bullshit 90% of the time and they were being a dick, but they can still say it, and admins will (afaik) be disinclined to discipline them because they don't actually know if the AI is just being an asshole or if they are trying to be serious and don't know better. Both law sets have their pros and cons, and honestly I'm biased because I like BS12's law set so I probably fail to see some of it's cons, but figuring out which one would be better for our purposes and playerbase now will potentially cut down on ruined rounds and mountains of ahelp requests. If things are clear cut where we mean for them to be clear cut and more ambiguous where we mean for AIs to have leeway, it'll stop people playing the "oh well I though I could do that because of law 1" when they abuse the laws.

Mostly related, but deserves it's own area, I think we should make our own lawset instead of using /tg/s or BS12s.

[MrWillsauce]

I think the AI should be able to choose from the /tg/ lawsets (Asimov, Corporate, Paladin) at the beginning of the round. Either way, I vote Asimov.

[Iceblaster]

...

* Iceblaster says Paladin then runs off

[Erils]

Haven't been on this thread for a bit so this is bringing up an old topic (AI harm-Asimov), but I just wanted to try and justify some aspects. I'll put it in spoilers so people who don't care don't have to look at it.

Spoiler: Can AI's harm people? (click to show/hide)

All I saw was attempted murder, an evil cult and a dead captain.

Get suffocated, evil person. :u

Unless you were non-asimov.... cult is still people

But if you read Asimov's "I Robot" you will see that robots can 'break' the first law in order to save more lives. The robots with Asimov's laws are able to allow the harm of or harm a human, if it prevents or might prevent the harming of more humans. Based on that, even asimov AI's can still harm or even kill traitors, syndicates, cultists, wizards and death squads if it would save more human lives in the end.

Actually, In I Robot it is explicitly stated that robots can cause a small amount of harm to humans to prevent a greater amount of harm. A robot could ram into a human, maybe breaking a few ribs, to prevent that human from being crushed by a rock. Killing humans, however, is not a small amount of harm and thus unacceptable.

In the story "Little Lost Robot" while they are trying to find one of the NS-2 "Nestor" robots that has hidden itself, they carry out a test. The test involves dropping a heavy stone onto a human and killing them (the robots don't know but the stone is rigged to stop). The room is filled radiation and the robots are told that, they will be unable to save the person and, should they try, will only destroy themselves which might prevent them from saving more humans in the future. When the rock is dropped, all the robots stay still (except the NS-2 as he could sense the radiation was harmless) even though they all (except the NS-2) had newtons 3 laws programmed into them without any alterations.
Based on that, an asimovian AI should be allowed to let a human die, or even kill them, if it might save the lives of future humans.

Spoiler (click to show/hide)

This was in a situation where the robots were influenced by the malfunctioning Nestor, who convinced them that there was no point in trying to save a human if they would be destroyed before they could reach said human. This also ONLY demonstrates the second part of law 1, since they are only letting the human be harmed, not directly harming the human.

Spoiler (click to show/hide)

The NS-2 Nestor unit wasn't malfunctioning, he was just following the second law and his purposely weakened first law. I'm also sure that Dr. Calvin, not Nestor, told the robots they would be destroyed if they tried to save her, and would prevent themselves from saving future people. Also, even though they only broke the second part of the first law, the implications are that they could break the first part as well in a similar circumstance. Breaking the second part should allow an AI to not open a door for a traitor or normal person that is suffocating, if they can save more people by opening other doors. Breaking the first part should allow an AI or robot should be able to kill a traitor who is about to blow up the medbay (with people inside) if it is the only way to save the numerous people inside the medbay.

Spoiler (click to show/hide)

No, the modified Nestor convinced the other robots that it would be futile to try and save the human in the second test, this was explicitly said.

And this situation demonstrates that robots can forego saving a human if they will ALWAYS AND WITH COMPLETE CERTAINTY die trying, thus achieving nothing but possible future harm.
This situation does NOT demonstrate how a robot could cause harm to prevent greater amounts of harm.

Spoiler (click to show/hide)

Sorry you're right. Nestor did tell them. But still, if they were able to break the first law in order to prevent a needless destruction of their themselves, and self-preservation is only the third law, it seems that they would be able to break the first law in order to prevent further needless death of humans, which is the first law. However, I only think that either should reasonably happen if the robot/AI is certain that breaking the first law would prevent the unnecessary deaths of other humans. However, I think that in-game this is up to the interpretation of the AI player.

[Hanslanda]

@Mono124: Generally the whole 'bolting upload at roundstart' thing is not something admins go out of their way to pursue, but to be quite honest, I consider it meta, not accurate roleplay of laws. (Granted, plenty of people roleplay in vastly different ways, so take this all with a grain of salt) I think an AI made by a megacorporation to run it's stations would be aware of the fact that the Research Director and Captain both have access to it's upload, and are, in fact, the only people on station that start with access to it. Logically, you could assume that if they have access, they are trusted enough to be allowed to tamper with the upload, having been intensely screened by Nanotrasen to get the security clearance necessary to enter the upload. Therefore, the Captain and RD should be allowed access at will, because it is their job to ensure the safety and full functionality of the station AI.

But most people put zero thought into any of that and just bolt it immediately because I dun wanna let da trator win. I don't want to try to force my roleplaying style on everyone else, so I just quietly shake my head and do my best IC to make the AI's life a living hell after that.

[Criptfeind]

Doesn't the round start by going on blue alert most of the time? Explicitly warning the AI that something is up and that it should in fact probably bolt it's upload?

Although that said if I was the AI I would unbolt it for the captain and RD.

[KingofstarrySkies]

Doesn't the round start by going on blue alert most of the time? Explicitly warning the AI that something is up and that it should in fact probably bolt it's upload?

Although that said if I was the AI I would unbolt it for the captain and RD.

It doesn't on Baycode. Which is a good thing, because then the alerts are ACTUALLY USED.

[Nick K]

We often have lowish populations - often less than 20 or even 15 in the evening for European timezones. IMO Having an antag-hunter AI in that situation makes rounds worse because with such low populations the AI can keep tabs on all the crew and quickly figure out which are likely to be up to something.
This means you get Antags that are afraid to do antag stuff because the AI will call them out instantly and it's round over if there are security forces present.

Asimov AIs still often act in this way (I see plenty putting the EVA motion alerts onto general comms despite EVA not actually containing anything dangerous) but at least they have the leeway to ignore theft and illegal access. In fact, if an AI knows that a possible punishment for theft is execution then it has a very good reason not to report the thief to security.

Paladin lawsets and other ones that make the AI completely crewsided would make this much worse because the AI would be actively encouraged to antag-hunt and would be allowed to shock doors as well as bolting them, making it much harder for the antags to escape.

[MrWillsauce]

I'm pretty sure the EVA thing is automatic because of the motion sensor there.

[Nick K]

The AI gets a motion sensor alert automatically yes, but I'm pretty sure that doesn't go out on the radio channel unless the AI player copy-pastes it there.

[Hanslanda]

The AI gets a motion sensor alert automatically yes, but I'm pretty sure that doesn't go out on the radio channel unless the AI player copy-pastes it there.

This.

[MrWillsauce]

I don't really see why that's metagamey. It's a secure area; there's a motion sensor there for a reason.