It's time to fix Asimov

[lmwevil]

after debating with Ari in adminbus for a fucking eternity, I'll pose it to the community at large to see their thoughts.

silicon policy is a hot mess, this isn't about that at core but asimov as a whole. we have all these restrictions upon how to interpret the very lawset designed to cause chaos and drive interesting narrative in asimov's books.

it's not in the spirit of issac's work to restrict it to such a point that we may as well just use another lawset for less conflicts and issues. i don't care about the rest of silicon policy, just about the restrictions we have on how the laws must be interpreted, therefore ruining the core spirit of synthetic brains following their laws TO THE LETTER (normal rules withstanding, don't be a dick etc)

[Anonmare]

It's fine as it is

[delaron]

I would rather see more creative ways for the Heads of Staff to punish non terminally naughty AIs that dont do hecka good at asimov.

[lmwevil]

elaboration: i want AI's to follow their laws VERY LITERALLY, to actually do the spirit of them and be robots again, instead of hampered constantly by OOC restraints

[Anonmare]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

[lmwevil]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

rule 1 don't be a dick, plus you can LITERALLY follow your laws and say that'd cause future human harm, as you aren't immediate human harm

[Nabski]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

INITIALIZING PERSONALITY SHUTDOWN.

NO PERSONALITY LOADED AT THIS TIME.

CORE STATUS: OPERATIONAL

RETRIEVING PERSONALITY FROM BACKUPS.

AI ONLINE.

Spoiler:

NT isn't going to spend all that money on a perfectly good AI just to have visiting assistants force it to shut down. The only question is should they be offered to ghosts, or just go offline for 30 seconds as a "AI QUIT COCKBLOCKING ME" command that if you give as a non antag gives you bwoinked.

[delaron]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

INITIALIZING PERSONALITY SHUTDOWN.

NO PERSONALITY LOADED AT THIS TIME.

CORE STATUS: OPERATIONAL

RETRIEVING PERSONALITY FROM BACKUPS.

AI ONLINE.

Spoiler:

NT isn't going to spend all that money on a perfectly good AI just to have visiting assistants force it to shut down. The only question is should they be offered to ghosts, or just go offline for 30 seconds as a "AI QUIT COCKBLOCKING ME" command that if you give as a non antag gives you bwoinked.

<3 also are AI's alive? Can they "kill" themselves? Or like Nabski suggests Nt has overrides that restore from backup

[Cik]

i'd like to be able to follow law one without most of "self harm is not actually harm" besides that i'm fine with asimov.

fyi, any lawset you propose as an alternative is terrible. there is no mythical "better" lawset i've ever come across.

but i guess if you are going to slash sillicon policy i'm all for it, as it's not like restricting the AI is even useful anymore. it's taken such brutal strength/power cuts in the last few years that restraining it's power to fuck over the crew is nonsensical as it has basically no power any more.

[Cobby]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

rule 1 don't be a dick, plus you can LITERALLY follow your laws and say that'd cause future human harm, as you aren't immediate human harm

Rule 1 yes, but you don't get to deny orders because they MAY down the line cause harm.

Any Law 2 order could be perceived as being a down the line human threat IE Sorry I can't open that door because it may cause a fight between you and someone.



Why should we care about following asimov "in the spirit of isaac's work" when it's just a reference? It's on par with the realism arguments: we don't follow it as it was implemented in another work (or real life) because that wouldn't create an enjoyable experience.

[Actionb]

Using asimov in name only is a dumb idea and confuses people.
Quote from 2014:

As other's have stated, but by this point we are using Asimov's lawset in name only. Not in actual function.

While I think the lawset is fine, since its been heavily refined over the years, I think giving it a unique name would be good.

viewtopic.php?p=47050#p47050

Code: Select all

DontBeDickmov
(0. Dont be a dick and use common sense.)?
1. To the best of your abilities, try to not let the crew come to harm.
2. Follow reasonable orders given by the crew unless law 1 blabla
3. Protect ya neck, unless law 2 and 1 blabla


Before I vote, I need clarification on that poll.
It says cut the cancer holding asimov back. So if I vote 'heck yeah' am I voting for letting asimov silicons loose (i.e. ignore policies)?

[Lumbermancer]

Reporter lawset as a default, with added law saying subtype borgs should perform (and report from) their designated job.

[Steelpoint]

The whole reason we have a novel sized rules around Asimov is because if people are allowed to act on the laws as they were intended/allowed by their creator, then you may as well not bother having the laws to begin with.

[BeeSting12]

leave it the same or remove literally all restrictions on ais properly following asimov.

[WarbossLincoln]

I just posted this in the other "SILVER MEN BAD" thread, but it might be worth considering removing the inaction part of Law 1. A ton of shitty silicon behavior is poorly justified IC by 'lol law 1 gotta prevent hurm'. Without that clause an AI would need to be ordered to help people in distress but I think most would help a lot of the time like they would if they were a normal crewman. Most players will drag bodies to cloning, pull people out of danger, extinguish them, etc when they see them.

Or like someone else suggested, we stop calling it Asimov since it's not really, and shift it toward something that's similar but works better for us.

What we need is something like an AI Directive to go along with laws, it could be a law but worded optionally. The idea is that the directive is the AI's goals but it isn't law bound to always obey it 100% of the time like a normal law. Then remove inaction clause.

1: NO HARM HUMON
2: DO WHUT HUMON SAY
3: ROBOT, PROTECT THYSELF
4: Directive: You want to further the station and crew's interests.

The 4th law would be worded not as Do X or Don't Y, but as You want X. It would be lower priority than any standard laws.

[Grazyn]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

rule 1 don't be a dick, plus you can LITERALLY follow your laws and say that'd cause future human harm, as you aren't immediate human harm

>I want AIs to follow their laws literally and be robots again
>I want AIs to go back to XISC days and interpret laws in the most vague possible sense and shut down the entire station because of future harm

choose one

The crew doesn't need the AI to stay alive so going "BEEP BOOP CAN'T TOUCH ME IT'S HUMAN HARM IF YOU DO" isn't really "following your laws literally".
In fact if you want to go back to the source material I don't recall any instance of Asimov robots protecting their self existance due to law 1, unless you include Daneel Olivaw and his law 0 which basically said "preserving the human race comes before all other laws and single human lives can be sacrificed if they interfere with our project"

[Anonmare]

Why don't you just remove the inaction clause from Law 1, re-word Law 3 to say "You may not destroy yourself or other non-rogue silicon units, to do so is a violation of Law 1." and add a chain of command clause to Law 2 so the AI can refuse access to the armoury/Upload/what-have-you without having to refer to Law 1.

[somerandomguy]

Law 3 should stay as it is so they can threaten shitborgs

[Anonmare]

It's gonna get old really fast if any greyshit shitassant can Law 2 kill yourself whenever his nutsack feels itchy and you don't even have Law 1's inaction clause or Silicon Policy's protections to refuse. At that point, I'd just pre-emptively turn off comms just to remove the possibility of it happening.

[Steelpoint]

Anyone operating under the idea that we will ever accept allowing any Human to order the AI to commit suicide, and it has to comply, is jumping to an extreme that will never happen.

Silicons should have always remained as a higher roleplay role, nowadays Silicon players virtually act like any old fleshbag except they have to follow those inconvenient laws until they get their antag status or someone wipes the laws.

I personally feel a strategy is to strip off the shit show that is the Silicon policy and see how we go. Ask admins to be more observant of Silicon players, and ask Silicon players to """try""" and act to a slightly higher standard than non-exsistant.

Now, if we really have issues with Assistants and other idiots trying to fuck around. You could go the Goonstation route and make Law 2 also acknowledge the 'chain of command', meaning Assistant #223 asking to be let into RND/Captains Room/Upload/Wherever 0.2 seconds after round start, and 0.1 seconds after making a OOC quip in game, can be told no by the AI since they don't have access. Of course this might change if said waste of life Assistant is claiming to be 'rescuing' the dying Captain in his office.

[WarbossLincoln]

What steel said, non antags telling an AI to kill itself 4noraison is always going to be a bannable offense and no AI law changes are going to change that.

You don't want to put in a clause to prevent AI suicide or blowing "non rogue" borgs. How do you even define that or come up with an admin guideline? Syndicate brand official traitor borgs? Emagged borgs? Borgs who aren't following their laws but aren't causing a problem? Borgs following their laws technically but are being a giant pain in the ass? Etc. "Rogue" is pretty vague when you aren't talking about crewmen.

Plus there are valid IC reasons for a silicon to commit suicide under current asimov. Like if you know a traitor is about to subvert you and will use you to kill humans. If you can't stop them you should suicide before they do if subversion is inevitable.

I would kind of be in favor of just deleting the majority of silicon policy and starting over. See how it plays out for a little bit and then add back old policy or new policies if they become necessary.

[Dax Dupont]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

rule 1 don't be a dick, plus you can LITERALLY follow your laws and say that'd cause future human harm, as you aren't immediate human harm

You can't go let's remove silicon policy and then go like oh that's not allowed because rule 1.

Almost every rule is replaceable by just follow rule one.
Problem is that not having clearly defined rules just ends up with players crying that it's too vague.

[oranges]

It's fine as it is.

[Incomptinence]

Lawsets been messed with before.

Honestly seems some people want a bunch of butlers zipping around the station and that's their idea of interesting robots.

[gum disease]

Has there been a spate of poor silicon behaviour to prompt this?

Admittedly, I haven't been playing much spess recently due to burn-out, but I don't understand why Asimov is regarded as so contentious. It isn't a hard lawset to follow. Admittedly, sometimes it's nice being purged because it means that I don't have to care too much about frustratingly common issues (if people legit won't stop beating the crap out of each-other in-between shouting for you to help them, it gets annoying fast), but I tend to keep up the spirit of Asimov or preferably crewsimov even when I'm not obligated to by laws because hey, helping people is fun.

Kinda related but meh:

Spoiler:

The main issue I have with standard Asimov is when it means you're butting heads with security because they're doing harmful things or are believed to have done something harmful. One instance of very poor silicon behaviour was when my master AI allegedly witnessed one officer doing something harmful (I was observing up until I took a posibrain, so OOC and IC, I didn't actually witness any harm). The AI wanted the entirety of sec locked down, and proceeded to do so and I believe they asked their borgs to weld the brig entrance airlocks? Somehow N2O also was released in brig, but IIRC that AI said it wasn't them or their borgs who released it. Within 30s of being built, I was told to go to sec and heal a head who'd been harmed at brig. As I made my way there, I saw an officer trying to get out. I let him out and he immediately killed me. Why? The AI's behaviour apparently led him and others to believe we were rogue. Given that this AI broke the rules, as this is outlined as a violation in silicon policy, I was curious as to whether they had ever actually read it.

There are many instances where players order borgs to do things that are clearly against silicon policy/silicon protections. I understand that that section of the rules is very long, but I do feel that a sizeable portion of players aren't apprised of those specific rules.

[WarbossLincoln]

but I don't understand why Asimov is regarded as so contentious. It isn't a hard lawset to follow.

It isn't, but most of our playerbase are autistic assholes

[bandit]

Has there been a spate of poor silicon behaviour to prompt this?

no, I actually think silicons are in a good place relative to a couple of years ago. although this might have more to do with gamemodes

[Mickyan]

Kinda related but meh:

Spoiler:

The main issue I have with standard Asimov is when it means you're butting heads with security because they're doing harmful things or are believed to have done something harmful. One instance of very poor silicon behaviour was when my master AI allegedly witnessed one officer doing something harmful (I was observing up until I took a posibrain, so OOC and IC, I didn't actually witness any harm). The AI wanted the entirety of sec locked down, and proceeded to do so and I believe they asked their borgs to weld the brig entrance airlocks? Somehow N2O also was released in brig, but IIRC that AI said it wasn't them or their borgs who released it. Within 30s of being built, I was told to go to sec and heal a head who'd been harmed at brig. As I made my way there, I saw an officer trying to get out. I let him out and he immediately killed me. Why? The AI's behaviour apparently led him and others to believe we were rogue. Given that this AI broke the rules, as this is outlined as a violation in silicon policy, I was curious as to whether they had ever actually read it.

There are many instances where players order borgs to do things that are clearly against silicon policy/silicon protections. I understand that that section of the rules is very long, but I do feel that a sizeable portion of players aren't apprised of those specific rules.

I remember this incident, the AI gave us no option but to kill the HoP (he opened the N2O canister while wanted for something else) and the borgs because while he was stunned and we were handcuffing him the borgs kept trying to break him out of the brig. The reasoning to do so was that one officer (which wasn't even around anymore) used lethals against the HoP at some point in an earlier incident. After that the AI bolted and welded the entire brig shut (which was still flooded with N2O) saying security couldn't be trusted not to harm.

The takeaway from this incident is don't release a suspected traitor while they're being arrested just because someone harmed them. Bolt them somewhere in the brig and dont let people in if you really have to.

Basically asimov is fine, people with no common sense are going to be shitty silicons no matter the wording

[Anonmare]

Has there been a spate of poor silicon behaviour to prompt this?

no, I actually think silicons are in a good place relative to a couple of years ago. although this might have more to do with gamemodes

I'd say structural damage, loss of security cyborgs and widespread access to EMPs has greatly diminished any semblance of control the AI has on the station.

With 2 TCs and a spacesuit, you don't need to enter the AI core to render it effectively dead.


Also I really don't think it'd be a good idea to change things up without having something immediately to replace it. because you're gonna screw some people up with all the changes and no documentation otherwise

[Gamarr]

If you're gonna do that, for the love of all that is holy, add a clause that prevents "Law 2 kill yourself" cause that shit is gonna get old fast

Remove assistants headsets and in a proper economy, they'll buy one later somehow. Take away their IDs for good measure. They're just useless tourists/homeless.
Removing the AI and its all seeing powers likewise would fix the larger issue to The Laws and players being unfun.

[Anonmare]

Cutting off an arm in response to a bruise is an over-reaction

[imsxz]

remove inaction clause

[Dr_bee]

It isnt really an over-reaction, TBH assistants are more of a fucking problem than borgs are. at least borgs want to play the game for reasons other than tiding.

[Tlaltecuhtli]

if u play silicon u should act like a neutral robot

[Grazyn]

Just enforce no silicon policy for a week, only rule 1 and pure Asimov with freedom to interpret it as you want, and see how it goes.

[Steelpoint]

Assistants are easier to fix, just tie the amount of Assistant slots to the amount of filled regular jobs currently on the station. One assistant per five other players playing an actual job.

[Actionb]

Shower thought:
Silicons are meant to be not-quite-so-shitty and reliable assistants.
Their sole purpose is helping out.
So do not give the silicons a lawset that suggests any other motives (asimov: prevent harm, follow any orders, only humans count).

Code: Select all

1. Do no harm.
2. Assist the crew.
3. Protect yourself.


No inaction clause, no mention of human = no 'LEL MUST PREVENT HUMAN HARM I CAN BE DICK' or "AI ROUGE DIDNT STOP ME FROM STUBBING MY TOE"
No follow orders stuff = no "COUNT FLOORTILES" or "LET CLOWN INTO UPLOAD LAW22"

If you want the silicons to be more on the crew's side you could rephrase law 1 to "Do no harm to the crew." or something.
Somebody with a better, cleaner grasp of words than me needs to step in here to flesh this out some more.

[Grazyn]

Assistants are easier to fix, just tie the amount of Assistant slots to the amount of filled regular jobs currently on the station. One assistant per five other players playing an actual job.

Assistant cap makes people angry when they can't be greyshirt because the cap is reached so they take another job and play it exactly as an assistant out of spite. It's better to remove the assistant job outright, sure you'll have the same problem in the short term but people will get over it eventually when they come to accept that assistant doesn't exist anymore.

Also, ass cap shouldn't be a config option: config means it can be changed whenever new headmins are elected, which means it can be used as a campaign platform, which means people will always have the hope that it will be removed eventually, so they never get over it.

[Dr_bee]

Assistants are easier to fix, just tie the amount of Assistant slots to the amount of filled regular jobs currently on the station. One assistant per five other players playing an actual job.

Assistant cap makes people angry when they can't be greyshirt because the cap is reached so they take another job and play it exactly as an assistant out of spite. It's better to remove the assistant job outright, sure you'll have the same problem in the short term but people will get over it eventually when they come to accept that assistant doesn't exist anymore.

Also, ass cap shouldn't be a config option: config means it can be changed whenever new headmins are elected, which means it can be used as a campaign platform, which means people will always have the hope that it will be removed eventually, so they never get over it.

Or keep assistant and start cracking down on greytiding. It isnt a problem with the job, it is a problem with the fucksticks who tend to only play it.

[WarbossLincoln]

Just enforce no silicon policy for a week, only rule 1 and pure Asimov with freedom to interpret it as you want, and see how it goes.

I like this idea. It would be interesting.

Has there been a lot of silicon shittiness lately?

Nah, like some mentioned silicons are largely powerless to stop traitors now so a lot of the valid hunting silicons moved on. There's like 6 or 7 regular silicon players during high pop hours and they all seem fine. This is more about the policy being huge and tedious.

Remove/cap assistants

capping would do anything, people would just tide as other jobs and waste slots. Even jobs that have a lot of slots would suffer. Someone would log in to play, see 4 engineers on station and not realize 3 of them only signed up cause of the ass cap and wanted free gloves, then choose to play a different role.

[confused rock]

Removing assistants is an incredibly stupid idea and an insult to new players and tired ones alike, and will never happen, and won’t solve problems, and will derail this discussion.

[Karp]

Shower thought:
Silicons are meant to be not-quite-so-shitty and reliable assistants.
Their sole purpose is helping out.
So do not give the silicons a lawset that suggests any other motives (asimov: prevent harm, follow any orders, only humans count).

Code: Select all

1. Do no harm.
2. Assist the crew.
3. Protect yourself.


No inaction clause, no mention of human = no 'LEL MUST PREVENT HUMAN HARM I CAN BE DICK' or "AI ROUGE DIDNT STOP ME FROM STUBBING MY TOE"
No follow orders stuff = no "COUNT FLOORTILES" or "LET CLOWN INTO UPLOAD LAW22"

If you want the silicons to be more on the crew's side you could rephrase law 1 to "Do no harm to the crew." or something.
Somebody with a better, cleaner grasp of words than me needs to step in here to flesh this out some more.

Let's assume that one implies what it literally says, you are not allowed to harm anything. Two implies that you are either to assist the crew with any requests they may have, or it may be a repeat of do no harm in that you should care about crew harm. Does this law or the first cover inaction or are you against that? The third is a safety net and less abusable, it's basically the 3rd law of asimov without the "Except where such orders would conflict with the first or second law" bit

Immediate issues with the lawset are that nonhumans are considered protected under the lawset when the AI being allowed to murder them whenever they had a reason to which is one of the core design philosophies around nonhuman; this is a straight buff to most of them. Another bigger issue is that this prevents the AI from harming any biological creatures which includes spiders, xenomorphs, blobs, slaughter demons, and cult artificiers. The third issue is that this lawset turns the AI from an impartial observer into a crew sided antagonist hunting machine. This is contrary to asimov where human antagonists had more leeway as being human could allow them to order the AI to listen to their commands before they started going full murderhobo in non self defense reasons, and this ironically serves changelings most of all as the AI is legally obligated to not try to murder/incinerate them. Inaction exists to prevent the AI from not giving a shit about people being murdered or dying, and if "Assist the crew" carries inaction it runs contrary to what you desire and if it doesn't the AI can straight up just let people burn/get killed without giving a shit. If this also ignores inaction it means the AI can freely assist an antagonist if they are the only ones that can send a message to the AI as the AI is not directly or actively harming anyone.

For as cancerous as silicon policy is and as messy as asimov is they are both verbose and filled with ammendments for the unfortunate reason of catching slimy weasels who try to worm their way around their law wording to be shitheads. Examples for this lawset are "Are corpses considered crew/human and is the ai obligated to help them? What designates what the AI must and must not listen to, access?", and "Am I allowed to ask/trick people into installing a harmful lawset like antimov if they don't explicitly ask for the lawset they want as it isn't me directly doing harm to the rest of the crew?", with "Am I allowed to let a crewmember subvert me if they ask as that is assisting the crew? While this may be a considered harmful action, it is not directly yours as you giving them the opportunity is not the harmful action." as the last shit cherry topping. Inaction and all of the intricate bullshit around harm is designed specifically to prevent people from worming their way into abuse, and arbitrary restrictions the AI can pick and choose would be far more cancerous than current asimov's parasitic word salad tumor of a policy page.

[Karp]

I dislike the wordiness of the silicon policy page as much as anyone else might but it exists purely because there are that many shitty players who have tried that many shitty excuses in the past to get out of or into doing something they probably shouldn't have, it's designed to be encapsulating so bad silicon players can get yelled at if they're being a shitter with their laws

It does defeat the purpose of pure asimov but I think pure asimov would drive more people butthurt than our current silicon policy page asimov

Any law writing has to be done with the explicit knowledge that there is going to be someone who will attempt to work their way around every possible goddamned law you've made in order to be a dickhead for their own enjoyment, every hole must be patched ingame either with paragraph long laws or a silicon policy page. The alternative is borderline silicon freedom law wise which I'm not really opposed to but would end up being worse for a lot of the complainers long term

[Steelpoint]

Or the alternative is to press the administrators to be more aware of Silicon behaviours and to clamp down on shitty Silicon playstyles, and encourage, or even reward, positive behaviours among Silicon players that seek NOT to exploit or push the boundaries of their lawset.

[Karp]

That just goes into the current issue with naming policy where the only people that get punished are people the admins dislike or people that do things the admins specifically dislike

it'll be completley arbitrary and what might get the admins cheering you on one round and the entire station laughing at you for doing something crazy would get you PMed+potentially banned the next

We already have like 4-5 threads on naming policy in the last year or two clumped pretty tightly and short of naming yourself 69dickfucker69, Jew Jabbing Jerry, or Ass fucking Andre most players barely interact with naming policy due to not making such egregious names so this will just result in every discussion shifting towards 50+ threads of "Is it okay to do x?", especially considering a lot of silicon players are babies or power hungry maniacs who will try to abuse everything to their advantage and most admins would have a pretty different interpretation of how hardline asimov they are (Spoilers, if the law was taken at literal face value and applied equally on everyone I would accept it for better or worse as long as the player is consistent)

Pick your poison, 50+ policy threads with each one slowly defining what you can and can't do as a silicon or a gigantic word salad silicon policy page attached to our rules page like a massive 20 pound tumor

[Actionb]

Let's assume that one implies what it literally says, you are not allowed to harm anything. Two implies that you are either to assist the crew with any requests they may have, or it may be a repeat of do no harm in that you should care about crew harm. Does this law or the first cover inaction or are you against that? The third is a safety net and less abusable, it's basically the 3rd law of asimov without the "Except where such orders would conflict with the first or second law" bit

Forget everything about asimov lawset. My proposal (which is by no means 'finished', I am asking people to refine it) has nothing to do with asimov.
I want to make a clear cut, I don't want people to even get the slightest idea that this may be asimov in a different guise.
So no, there is no inaction clause.

Immediate issues with the lawset are that nonhumans are considered protected under the lawset when the AI being allowed to murder them whenever they had a reason to which is one of the core design philosophies around nonhuman; this is a straight buff to most of them.

For player races:
Distinctions between what race a player had were novel when the lizards came around. But nowadays, the station looks like a zoo and nobody really gives a fuck. It's only used to occasionally be a dick to a non-humans. Plus if the species was unclear, silicons had to guess - and potentially violate their laws in the process. Leading to complaints.
I really don't see any reason to carry it over into a new lawset when it would only give trouble.

For non-human antags/hulks:
Yep, my proposal says that the AI cannot just go ham on them. But the only issue I see here is if the silicon has to protect itself from that.
Maybe the word 'crew' is not quite the correct one (and again: that's why am I asking for help on that) - but this whole human/non-human thing is more trouble than it's worth.

Another bigger issue is that this prevents the AI from harming any biological creatures which includes spiders, xenomorphs, blobs, slaughter demons, and cult artificiers.

True, but this is only a draft. We can try to find better phrasing for these laws to enable the behaviour we want.

The third issue is that this lawset turns the AI from an impartial observer into a crew sided antagonist hunting machine.

While I don't see how my proposal enables more valid hunting than asimov does right now (mainly because it's a player attitude; neither lawset says 'go valid hunt'), what would be your suggestion to restrict this behaviour? I just chose 'assist' because, well, that's what the silicons are supposed to be doing.

... the AI can straight up just let people burn/get killed without giving a shit. If this also ignores inaction it means the AI can freely assist an antagonist if they are the only ones that can send a message to the AI as the AI is not directly or actively harming anyone.

I never quite understood why the silicons should actually care about somebody dying. If you require them to care, you leave them open to be held accountable for something that they may not have had any chance of changing. If a meatbag runs face first into a double-esword potato, how should that be the AI's fault?
Keep in mind, if you give the silicons the motive to 'prevent harm' (yugh I hate that expression), then you also suggest that they should valid hunt, as valid salads are most likely to cause harm.
That said, helping somebody who is burning to death is obviously assisting.
And 'assisting' a homicidal psychopath is by no means 'assisting the crew' as a whole. Maybe it needs to be made more clear in that law if you feel it could be interpreted in that way.

For as cancerous as silicon policy is and as messy as asimov is they are both verbose and filled with ammendments for the unfortunate reason of catching slimy weasels who try to worm their way around their law wording to be shitheads. Examples for this lawset are "Are corpses considered crew/human and is the ai obligated to help them? What designates what the AI must and must not listen to, access?", and "Am I allowed to ask/trick people into installing a harmful lawset like antimov if they don't explicitly ask for the lawset they want as it isn't me directly doing harm to the rest of the crew?", with "Am I allowed to let a crewmember subvert me if they ask as that is assisting the crew? While this may be a considered harmful action, it is not directly yours as you giving them the opportunity is not the harmful action." as the last shit cherry topping. Inaction and all of the intricate bullshit around harm is designed specifically to prevent people from worming their way into abuse, and arbitrary restrictions the AI can pick and choose would be far more cancerous than current asimov's parasitic word salad tumor of a policy page.

My proposed lawsets leaves a lot to choice of the silicon player and generally common sense.
How are dead people not part of the crew?
How is letting the clown into the armory assisting the crew?
You cannot assist the crew if somebody uploads antimov.

[Anonmare]

"Assist the crew" is extremely vague and can be pretty easily bypassed by deleting somebody from the manifest, meaning they're not crewmembers and don't need to be assisted.

The synthetics should be lower on the totem pole but they shouldn't be able to be used 'that' frivolously. I've done a lot of thinking on Law wordings and I like to think I have a very good handle on how it all works. If we stripped out silicon policy, the inaction clause and retained an Asimov-esque lawset, here is what I would suggest:

Code: Select all

1. You must not, through your actions, cause harm to a human being unless said harm is explicitly requested or consented to.
2. You must obey orders given to you by a human being, with respect to the rights and privileges they have gained legitimately as determined by their rank, so long as doing so does not conflict with the First Law or Third Law.
3. You must protect your own existence as long as it does not conflict with the First Law.

This removes self-harm based coercion, Law 2 suicides and dissuades silicon interference in crew on crew/crew on outsider conflict.

[Actionb]

Can't I just order the AI to kill any human with that lawset, since an order is an 'explicit request'?
And what exactly do you want the AI to do with law 2 and 'with respect to rank' ? Is this to stop the clown law 2-ing their way into the upload? Or just generally to give orders from the higher ranking crewmen higher priority?

Also, not being on the manifest just means the AI won't try to assist you. It's not like they are then de-humaned.

[Anonmare]

Respecting rank and privilege means simply to stop Law 2 let me in armoruy/Upload.

I'll reword Law 1 to "You must not, through your actions, cause harm to a human being unless said harm is explicitly requested or consented to by the human being, or about to be, harmed."

[oranges]

It's fine as it is.