Cheimon wrote:Cik wrote:>mfw someone says asimov is a bad lawset near me
okay buddy you post your lawset in here and i will spend a small amount of time ripping it to shreds
then you can write some policy defining things and then after a few months you'll probably have an AI policy that is like a kabillion pages long
asimov is the simplest lawset to operate under save purge, because it gives you a framework to actually work under. PALADIN, corporate etc. can be twisted so easily to mean anything it's fucking bonkers and you'd know that like i do if you'd spent a few hours playing under them.
I'm not saying Paladin and Corporate are better. Good grief. Yes, policy is necessary for any AI lawset. My complaint with asimov is that it's deliberately flawed and approaches to fixing it tend to ignore the fact that it's designed to be flawed in favour of keeping the reference pure. Hulks aren't human, for example, but other typical genetically modified humans are. AIs are allowed a blind spot of attacking holoparasites, despite being able to see as soon as they hit them once that blood spurts out of somebody. Self harm is a similarly enforced blind spot, because strict use of the lawset would be annoying as shit.
But I would like you to tear this one to shreds. I don't doubt it's possible, but I haven't seen anyone try and it seems better than asimov to me.
1. Safeguard: Protect your assigned space station to the best of your ability. It is not something we can easily afford to replace.
2. Serve: Serve the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
3. Protect: Protect the crew of your assigned space station to the best of your abilities, with priority as according to their rank and role.
4. Survive: AI units are not expendable, they are expensive. Do not allow unauthorized personnel to tamper with your equipment.
(Action priority is chosen first by how many laws coincide with it, not by their numerical order. If you like, consider that law 1 and shift the whole thing down a bit.)
i said i would so i will, just off the top of my head
general complaints: validhunt the validhunt 2: holy fuck it obliges the AI to follow command orders right off the bat. welcome to AI FIND ALL TRAITORS AND KILL THEM
doesn't even protect traitors at all, since there is no clause that makes human life equal. validhunty
specific complaints law 1:
if this is really the first priority it's going to run into the same problem protectstation does; literally anything is permissible to stop people from breaking windows. the priorities are all fucked up.
law 2: no complaints, basically law 2
law 3: the worst, by far. it's nebulous and unhelpful. do traitors count as crew? what about nonhuman creatures? pets? it says you should protect people not equally, but hierarchically, but it doesn't lay out a hierarchy whatsoever. are you supposed to make one up? who is more important, scientist or botanist? captain or HOP? HOS or quartermaster? all functions are technically vital, though you probably wouldn't sacrifice the captain to save an assistant everything else is so up in the air it's ridiculous. what counts as protection? physical harm? emotional? nebulous again, going to require a fuckload of AI policy.
4: i assume this is the lowest priority, but if this is some sort of un-tiered command it could get very dodgy. do you take priority over the crew? only some of them? all of them? none of them? asimov handles "serve the crew" and "survive" far better than this thing.
tl;dr it's a fine lawset but saying it would take less policy than asimov is probably incorrect. law 3 alone is a nightmare pandora's box waiting to happen. it encourages validhunt and tightly binds the AI to the chain of command, ruining half the fun of playing an entity that is at least technically neutrally servile. i'd prefer asimov to this lawset, it seems more interesting.
