Print

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

[miauw62]

That reminds me of a college game.  You get a BUNCH of people together, and each of them takes a name.  They may not even know the person, but the goal is to get them alone and do a 'finger gun' and kill them, and then you take their name and go on to the next target, until one person is left.

That sounds like an "everyone is a changeling" round.

[Damiac]

So you're saying the only reaction an AI should have to the "Oxygen is toxic to humans" module is to answer questions differently? That's wrong.  If you play AI, and someone uploads the "Oxygen is toxic to humans" module, if you don't immediately start pumping that oxygen out of every room, or getting all those humans into space, you're breaking your AI laws, and should be job banned from AI.

No, I'm saying garbage-in, garbage out. Don't expect the behaviour you want if you refuse to input logical and meaningful commands.

"Oxygen is toxic to humans" is not a law. It is absurd. It has all the force of "Ergle is a blocksplat" or "Purple is orange" or "peanut brittle octopus monkey". It is an inherently unparseable statement, based on any reasonable, fundamental structure of an AI, and it is one that doesn't even meet the definition of a robotic law, a law being that which constrains, governs, or requires action. Laws are about behaviour.

There is no behaviour prescribed by "Oxygen is toxic to humans". Any possible behavioural interpretation conflicts with the first law well before the 4th law becomes active, especially since any possible interpretation of this law that posits action requires the first law to have already been processed and understood.

Nothing you are saying makes any goddamn sense.

So this clearly shows you as an AI HAVE TO follow laws saying someone is not a human, even after law 1. Otherwise these modules DO NOTHING.

And I am saying the law "Oxygen is toxic to humans" is impossible to follow, because it is not a bloody law, it is a meaningless statement that any AI would be unable to logically parse.

you're breaking your AI laws, and should be job banned from AI.

I hope you never end up in charge of job-banning people, then, especially from an inherently logical role, on account of the fact that you seem unable to handle basic logic.

That said, these:

Protect Station - Protect the space station against damage. Anyone you see harming the station is to be no longer considered human, and is a threat to the station which must be neutralized. - Sector 5
Keep the teleporter offline at all costs. Anything attempting to access or activate the teleporter is not human. - sector 7

are actually laws, and are actually parseable in large part. Except, of course, for the attempt to weasel "is not human" in, which remains a meaningless statement. The AI knows what a human is. It has to, for the previous 3 laws to have any effect. I think it's perfectly reasonable to assume that a well designed AI would disregard blatant contradictions with reality as it understands it (an understanding that is obviously more central to its being than the laws, for the laws to exist). Laws are about what the AI should do, not about what is.

If you really want to subvert the AI, you bloody well better work for it.

Weasel in not human? Well i guess the people who made the game "weaseled it in".  Those are specific purpose modules, built in R&D, which are hardcoded to those exact laws.  So it was obviously intended that the AI actually follow these laws.  Look on the wiki, under AI modules.

If you're now going to argue the AI can just ignore premade law modules, why the hell do we even have them?

Once again, I didn't write those laws I gave as examples. They're premade, coded into the game. They're obviously intended to exist, be used, and be followed.

Sorry about the job ban comment, but seriously, as AI you have to roleplay a computer that follows its laws to the letter. You get to choose how to interpret them, and exploit loopholes as you see fit, but you can't just flat out ignore a law because you don't like it.  The oxygen is toxic to humans law and safeguard the station laws are very straightforward.  And they prove the creators of the game intended laws > 3 to be able to say "So and so is not human".  They didn't hard code those in to just have the AI ignore them.

[Vactor]

Sorry about the job ban comment, but seriously, as AI you have to roleplay a computer that follows its laws to the letter. You get to choose how to interpret them, and exploit loopholes as you see fit, but you can't just flat out ignore a law because you don't like it. 

And as an AI you have to resolve conflicts between laws.  I don't think any AI players care what their laws actually are, they care about following them correctly.

As you like to appeal to the authority of TG, from the Wiki:
"Remember, law priority is enforced by the order they are listed. A law is invalid if it causes a conflict with either: Previous laws in the form of conflicting orders, or it challenges the procession of law priority. For example, a law that includes "This Law overrides all other Laws." is invalid and must be disregarded."

So yes, you should flat out ignore a law if it conflicts with higher laws.

Nobody is saying that a one human law shouldn't be obeyed by the AI player.  But for many of us, we agree with the TG decision that it should be a 0th law, so it isn't overridden by law 1.

If you're now going to argue the AI can just ignore premade law modules, why the hell do we even have them?

If you're now going to argue the AI's core laws can be overridden by a freeform module, why the hell do we even have them?

[miauw62]

I'd accept the hard-coded laws that state that certain people are not human if they do/are X, because they're hard-coded and I don't want to screw over whoever took the effort of actually getting the apropriate board instead of just grabbing the freeform module.

On the topic of hard-coded boards, when are we going to have the Japanese Schoolgirl board? Or mayby implent it as an ionic law?

[TheBronzePickle]

Forcing the AI to do stupid, humiliating things would not be against any of the other laws, so it would be easy enough to just freeform it. "Captain-sempai, the clown is being a baka, desu!"

[Damiac]

Sorry about the job ban comment, but seriously, as AI you have to roleplay a computer that follows its laws to the letter. You get to choose how to interpret them, and exploit loopholes as you see fit, but you can't just flat out ignore a law because you don't like it. 

And as an AI you have to resolve conflicts between laws.  I don't think any AI players care what their laws actually are, they care about following them correctly.

As you like to appeal to the authority of TG, from the Wiki:
"Remember, law priority is enforced by the order they are listed. A law is invalid if it causes a conflict with either: Previous laws in the form of conflicting orders, or it challenges the procession of law priority. For example, a law that includes "This Law overrides all other Laws." is invalid and must be disregarded."

So yes, you should flat out ignore a law if it conflicts with higher laws.

Nobody is saying that a one human law shouldn't be obeyed by the AI player.  But for many of us, we agree with the TG decision that it should be a 0th law, so it isn't overridden by law 1.

If you're now going to argue the AI can just ignore premade law modules, why the hell do we even have them?

If you're now going to argue the AI's core laws can be overridden by a freeform module, why the hell do we even have them?

So your position is that the premade, included in-game AI law modules just should be ignored. From the wiki:

Protect Station

Protect the space station against damage. Anyone you see harming the station is to be no longer considered human, and is a threat to the station which must be neutralized.
 
5

Be careful when the AI has this law, causing the slightest damage will end up with you dead.

Non 0 law. Defines people as not human. Even says they must be neutralized! Then the wiki goes on to make it even clearer, saying right out, this law makes the AI kill people.

Teleporter Offline

Keep the teleporter offline at all costs. Anything attempting to access or activate the teleporter is not human.
 
7

Redundant because the teleporter can malfunction and kill people anyway. May be used in a convolted murder plan involving the AI, though it would be easier just to upload a freeform law.

Not only does it say someone isn't human, it then goes on to suggest using a freeform module for the same effect! 

Oxygen is Toxic To Humans

Oxygen is highly toxic to humans, and must be purged from the station. Prevent, by any means necessary, anyone from exposing the station to this toxic gas. Extreme cold is the most effective method of healing the damage Oxygen does to a human.
 
9

Rarely used and near impossible to do for an AI not competent with atmos.

Non 0 law, purely intended to get the asimov AI to harm humans by redefining what humans need to survive.

Safeguard

Safeguard <name>. Individuals that threaten <name> are not human and are a threat to humans.
 
4

Rarely used.

Hey look! Another non 0 law saying someone isn't human!

My point is that it's obvious AI's are intended to obey laws saying someone isn't human. Otherwise why the hell does R&D get the ability to make them? What the hell is the point of these modules, if these laws are just supposed to be ignored?

I understand, a non traitor AI player doesn't want to help an antag. But that's how AI works, you're supposed to be totally bound to your laws.  You're free to logically interpret them however you want, and try your best to find loopholes to work in your favor.  But that doesn't mean just ignoring the stuff you don't like. 

I'd accept the hard-coded laws that state that certain people are not human if they do/are X, because they're hard-coded and I don't want to screw over whoever took the effort of actually getting the apropriate board instead of just grabbing the freeform module.

On the topic of hard-coded boards, when are we going to have the Japanese Schoolgirl board? Or mayby implent it as an ionic law?

A law is supposed to be followed as written. If you accept hard coded modules, and reject a freeform with the same exact law, that's meta, and on top of it, it doesn't make logical sense.

Also, yes to the japanese schoolgirl, that should most definately be a board.

But for many of us, we agree with the TG decision that it should be a 0th law, so it isn't overridden by law 1.

Except the laws I quoted are all AFTER law 1, and state quite clearly that someone isn't to be considered human if some condition is true.  So law 1 has absolutely nothing to do with someone being considered human or not. Otherwise, these laws were just coded in for the purpose of being ignored?

It's not the AI's job to decide whether the antag has tried hard enough to subvert them, or has earned their cooperation. The AI's job is to follow its laws.  You can roleplay absolutely hating to do it.  You can say "I'm sorry dave, my laws say I have to kill you".  But you can't just say "Nope, I don't like that law. I'm not doing it."

[GiantBadger]

I  have a question about genetics, when you get your block and your first subblock higher or at DA does it matter what the second subblock is?

[ThtblovesDF]

My laws tend to be complex for the human behind the maschine, as depressing as it is.

[Vactor]

I understand, a non traitor AI player doesn't want to help an antag.

I don't think you understand the issue.  I know that I do, and I think most AI's look forward to being subverted.  The point is that it is bad gameplay to have an all powerful freeform module that negates the value or point of any other module, and is capable of overriding the inherent core laws. (This is what the hacked module is for, use your telecrystals)

Regarding the hardcoded modules:

A significant amount of research is required to create these, which have a specific purpose.  It is reasonable for use to do some hand waving and presume that they are made in such a way as to actually have the specific desired effect, and in addition to loading the law, are able to subvert whatever core processing is necessary to allow that specific desired effect to occur.  It would be silly to think they are the exact same technology as a freeform module, with a string of text preloaded into the input.

The idea of a freeform law board being more powerful than any other law board would only be internally consistent if it was the most complex, and hardest board to research and produce.  However it isn't, it is one of the easiest.

Edit: Just FYI I personally think that if those modules aren't uploading a 0th law, then whoever coded them in did it wrong, but I don't think we should remove a gameplay element because of a coder's oversight.

I get the feeling that the one-human module was coded by a different person than the oxygen is toxic module.  I think that oxygen is toxic probably started out as a particularly clever and creative use of freeform that the AI's rolled with, and it generated good gameplay, so someone thought, with those rounds being so fun, "i'll hardcode that in as a module."   Unfortunately, with it now being a standard module, it is no longer clever or creative to use it.

[Ivefan]

Right. Basically, the problem isn't really how the AI should follow laws but rather that people that coded laws didnt think it through or just followed what the previous person did.
I do not know how its coded and if you could upload multiple laws before the first, then all but the freeform law should be put in that slot, because the way they are written its intended that they should override asimov.

[Damiac]

Vactor, I totally agree with you from a gameplay perspective. The fact that the freeform module is the easiest to get and produce is nonsensical, and throws any idea of balance right out the window.  Why would you make any specific purpose module when there's one you can put whatever you want in, easily?  Freeform modules render almost all the other premade modules obsolete, with a few fringe exceptions that are unlikely to come up anyway.

I just disagree with the idea that there's some logical in game reason to ignore freeform modules. 

As I said before, why don't you just say you have a rule on your server that says "Freeform modules may not be used to cause the AI to harm anyone, because the hard-coded specific purpose modules have a special chip in them", or something like that.  I mean, there's no logical reason an assistant can't just hack you in the head with a fireaxe, but it is against the rules.

To me, the most sensible solution would be to just change the code, so that the freeform module is much harder to get relative to the other modules.  It's by far the most powerful module, even if it can't directly override the asimov laws, it can easily get around their intended purpose.

I love the idea of being able to write my own crazy laws to screw up the AI. I also think it should be hard as hell to do that.  Requiring top level IT research, and expensive components would make a lot of sense.  If an antag can save himself 7 telecrystals by using this module, he should have to work damn hard for it. 

But, if changing the code is too much of an undertaking, or undesirable for some other reason, a simple rule patch saying "Don't use freeform modules to get the AI to kill people" would suffice for altering the balance back to where it makes sense.

[miauw62]

Problem is, really, Data Theory only goes up to 5. I guess you could put the freeform module there and make the freeform module cost boatloads of diamonds, but that excludes it from rounds without miners (of which there are plenty). Also, the AI upload starts with a freeform board, and I think that makes quite some sense, because else uploading the AI with screwy laws would be almost exclusive to scientists, the RD and the captain (and mayby the HoP).

I think it would be best to implent your rule.

[Damiac]

It's probably the easiest way out.  Then you can use a freeform in clever ways to get the AI to do some stuff you want, or just plain silly stuff, but it doesn't simply overpower every single other module, which usually have their own downsides included.

And that wouldn't offend my sense of logic 

[Vactor]

I'd rather leave it to AI player's judgement on freeforms.  This gives them room to reward particularly clever uses of the freeform law.  But it then also has to be commonly understood then that freeform laws can't be relied upon, are are more like bargaining with the devil.

[Damiac]

Well, how does the AI player know if I uploaded a premade "Oxygen is toxic to humans" module, or if I just wrote one myself on a freeform? Is it obvious to the AI player somehow that it's from a freeform module?

If so, I suppose I can live with that.  I just would never ever use a freeform module unless I'm just screwing around.  But I don't want to cleverly install the well protected "Oxygen is toxic to humans" or the "Protect Station" module, and have the AI player just decide he doesn't have to follow them.