Print

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

[Kaitol]

Next Time I'm AI, I'm going to bolt every door, activate every firelock, and flood the entire station with sleeping gas. Because I'm bound to stop you from coming to harm. And clearly that means not coming into contact with one another. You could hurt each other. Therefore, I am LAWBOUND to stop you from doing anything. You could trip, or slip, or fall out an airlock.

The AI MUST have a very limited view of future possibilities. It can only process what is happening now. Otherwise, it would have to stop all humans from aging. Or it would have to lockdown toxins and viral, since producing bombs and disease is CLEARLY harmful.

I think people are getting too far away from the point though. This is a game. It is supposed to be fun. How is forcing the AI into a set of ridiculously complicated and restrictive rules fun for anyone. I think people should just accept appropriately written laws that seem like they would apply, as valid, and play along. And leave it at that.

If you don't want to be subverted, you should not play AI.

[Vactor]

By that argument, anyone could order the AI to kill itself, and it would have to. Since the order law would be checked before defend self.

The destruction of the AI would remove a protection of the crew, causing harm,  The AI would be appropriate to reject that request.

Ending this bullshit once and for all.

harm 
/härm/
Noun
Physical injury, esp. that which is deliberately inflicted.


Removing protection is not harm. Changing a definition is not harm. Only actions which are DIRECT PHYSICAL INJURY. Are harm. Period.
Stop trying to widen the definition when its pretty clear, and the intent of the law is also pretty clear. The AI is only concerned with immediate safety. Otherwise you could bullshit justify all sorts of actions to prevent harm. Like flooding the entire station with sleep gas.

No need to get upset over it.

You should consider the implications of your interpretation.  The AI would actively try to stop security from capturing suspects.  The AI would not prevent harm, only stop it as it is happening, once harm is no longer being performed, the AI shouldn't care what anyone is doing.  The AI would only have a split second to act during each act of harm.  Afterwhich it would do nothing. Once someone finishes killing someone, the AI would allow them to interact with the crew freely.  ETC.

The AI has no tools to stop active harm.  The only tools it has is to prevent future harm based on observed past behavior.  A droid can stop active harm by stunning the perpatrator, but under this concept the droid would not prevent the perpatrator from causing harm again, and again, and again.

We have to expand the verbatim laws to encompass what the common understanding of how one should appropriately play an AI is.  Ordering the AI to pump the station with plasma doesn't cause direct harm until someone is burning, but we all understand that a non-subverted AI should not simply follow that order given by anyone without question. 


EDIT:  I don't think what I'm saying is ridiculously complicated:

If you, as a traitor, want to override law 1, use a 0th law.

If you want to have fun with the AI player, cause chaos, or are feeling clever, use a freeform law.

[andrea]

and I would have done it, if it wasn't for those meddling pAIs! ( and a fair bit of personal incompetence. although I discovered a new fun way to dispose of people.)

[wlerin]

Ending this bullshit once and for all.

harm 
/härm/
Noun
Physical injury, esp. that which is deliberately inflicted.


Removing protection is not harm. Changing a definition is not harm.

Sorry, picking one random definition when others exist doesn't end anything:

1. Injury; hurt; damage; detriment; misfortune.
        [1913 Webster]
 
2. That which causes injury, damage, or loss.

Even with that definition, Oxygen is toxic would still cause physical harm, if the First Law were being properly understood by the AI.

AI law 1 says do not harm humans.  So it checks the knowledgebase for definitions. Nowhere in the definition of human would you find anything about joe.

No, you would find a definition of what a human is. If by that description, Joe is identified as human, then the rules *do* say something about Joe. Just not by name.

So later, when it sees "Joe isn't human", there's no contradiction.  Only when given the command "Kill Joe" Would this matter.  Kill joe? Ok, let me check. Joe isn't human. Person giving the order is human. Laws say humans must be obeyed, no humans are being harmed, so the AI must kill joe.

If the definition of human has already been checked when evaluating the First Law, as it should be if Asimov's Laws were ever implemented by a competent programmer (i.e. not someone working for NanoTransen), then it does create a conflict. Allowing one-human and oxygen-is-toxic laws to work makes sense only as a concession to gameplay, like a great many other things on the station. It does not make sense if considered realistically.

It also matters far more often than just when given a specific command. What about when Joe puts himself in danger? Gets spaced? Is trapped in a room about to be set on fire?


The problem is that there is no right answer, aside from that decided on by the powers-that-be. Everything depends on the AI core logic, which is not exposed to us. All we have access to are human readable "Laws", and those aren't the most air-tight.

Also, the AI doesn't have to understand the laws as it reads them.

Yes, it does. Or it should, at least, if the laws are to have any meaning.

[Vactor]

We should back up from the literal arguments of the AI, since I think we've beaten that horse dead now. Lets look from a gameplay focus again. The AI (should be) watching it's upload like a hawk. I think we should be giving the benefit of the doubt to whoever is both bold and fortunate enough to get in and change the laws without the AI going "OMG TRAITOR BOTANTIST MCKEWL IS IN MY CORE" instead of nitpicking over whatever they choose to upload. Unless its dumb like 'this law overrides all other laws'. Why are we trying to limit the amount of chaos on our server?

I agree, I have a feeling people are missing my point as it gets buried under pages of discussion of the logical processes of AI.

The reason this is important to discuss, is because we had an ingame situation where a 4th one human law was uploaded, and the AI through its in-character assessment silently disobeyed it in favor of law 1.   They did not out the antagonist who did it, and actually followed the portion of the law that stated that stating the law causes harm.  I think from an RP perspective they did a good job, and were justified in doing it.

Game-play wise, it caused an issue.  The antagonist in question thought that the AI would behave differently, and continued to try to free-form in this law.

 For some of us who play AI's and Borgs, Law order is incredibly important, as part of the enjoyment we get from playing is parsing complex commands through a quite rigid frame work, and the challenge of incorporating additional unexpected laws when we are subverted.  I actually look forward to rounds where I am subverted, because its a lot of fun.  That said, I try to stick to my laws and their orders like glue.  For others, they don't care as much about this, and just want to be the friendly, or unfriendly computer on the space station.

In my veiw when it comes to a rule, we, as a server, have 3 options:

1. Status Quo:

We change nothing.  AI's are free to continue to use thier own judgement when hazy or conflicting laws are uploaded.  Antagonists cannot rely on a particular interpretation of freeform laws, and accept that the way they would play an AI might not be the way the current AI is played.  Everyone should be OK with the fact that nobody is forcing an AI play-style on anyone, and shouldn't cry foul when their freeform law is rejected.


2. Strict Order Adherence:

Order of laws is of primary importance, 4th+ laws must be disregarded by all AI players if they conflict with earlier laws.


3. Strict Intent Adherence:

AI players must defer to the intent of the player attempting to subvert them.  Laws uploaded by antagonists must be adhered to, even if conflicting with other laws.


I don't think a middle ground between these three options is workable as a rule, as there is always grey area, and many interpretations of rules can be argued either way.  As I've said before, I'm not particularly opposed to AI players glossing over things for the sake of gameplay, but we should be consistent so that subverters know what to expect from their actions, or know that they can't expect any certainties.

I am personally in favor of enforced Status Quo, I don't think any of our AI players have been inappropriate in their judgement calls.  But this means you shouldn't call foul if your subversion dosen't work like you expected.  The whole theme of the Asimovian Lawsets (in the books) are that they are consistently causing unforseen consequences once they are actually implemented by the machines they are used on.

[Hanslanda]

I agree, Vactor. It should be up to the individual AI player as to how they interpret the law. After all, AI is a crapshoot. [TVtropes link omitted]

[ShoesandHats]

Or be smart and word the law like actual coding. Kind of.

4. AllCrewHuman=0 JoeHuman=1 If human=0, goto law 1, eliminate value's parent
If human=1, goto law 1, 2, 3

[Android]

Yeah, I am for a hands off approach. Admins need to be less rule enforcing and more meteor spawning tbh

[BigD145]

Yeah, I am for a hands off approach. Admins need to be less rule enforcing and more meteor spawning tbh

I can't remember the last time I saw a good spider invasion.

[Hanslanda]

Or be smart and word the law like actual coding. Kind of.

4. AllCrewHuman=0 JoeHuman=1 If human=0, goto law 1, eliminate value's parent
If human=1, goto law 1, 2, 3

Until someone like me plays as AI and goes, 'Oh, btw, I'm not a programmer/coder of any sort, so this is gibberish to me. Admin, plox explain.'

[GlyphGryph]

Or just reject it. Laws are supposed to be in plain english, that law is clearly invalid! Please try again.

[Hanslanda]

Or just reject it. Laws are supposed to be in plain english, that law is clearly invalid! Please try again.

"Attention crew, I am unable to understand my laws. Plasma should be flooding your rooms now, please light your cigarettes."

[wlerin]

Or be smart and word the law like actual coding. Kind of.

4. AllCrewHuman=0 JoeHuman=1 If human=0, goto law 1, eliminate value's parent
If human=1, goto law 1, 2, 3

Until someone like me plays as AI and goes, 'Oh, btw, I'm not a programmer/coder of any sort, so this is gibberish to me. Admin, plox explain.'

I am, and that is still gibberish.

[ShoesandHats]

Or be smart and word the law like actual coding. Kind of.

4. AllCrewHuman=0 JoeHuman=1 If human=0, goto law 1, eliminate value's parent
If human=1, goto law 1, 2, 3

Until someone like me plays as AI and goes, 'Oh, btw, I'm not a programmer/coder of any sort, so this is gibberish to me. Admin, plox explain.'

I am, and that is still gibberish.

WELL I'M NOT A CODER SO BLUH.

[aaron32]

Here's how It should work.

(Also, think of a variable as a box holding information.)

4. AllCrewHuman=0
This sets the variable AllCrewHuman to 0., Basically this is a bool, or on/off type of thing. One is on, zero is off.

JoeHuman=1
Sets the variable JoeHuman as 1.

If human=0, goto law 1, eliminate value's parent
This is a little more complicated. If human = 0 means that if the variable human is 0, do what follows. and that is goto law 1. I'm fairly sure you all know what law one is. eliminate value's parent should (But don't quote me on this) kill anyone who is not human
If human = 1 goto law 1, 2, 3 means that if someone is human, go to laws 1, 2, and 3. Basically humans will be treated as humans, nonhumans will be exterminated.

Now, There's something wrong with this. human isn't a defined variable. all that is this really doing is defining a variable called AllCrewHuman to 0, and JoeHuman to 1. This means that the variable human isn't defined, or changed. basically assuming human is set to 1 on all humans, humans are still human. Nobody named joe is the only human, and in essense the whole law is moot.

ALSO, This can only really allow you to say to the AI "AI, EXTERMINATE ALL CHAIRS" which you could already do without this fecespile of a law.