Thoughts about methods to make the AI immune to law changes

A place to record your ideas for the game.
Post Reply
User avatar
zxaber
In-Game Admin
Joined: Mon Sep 10, 2018 12:00 am
Byond Username: Zxaber

Thoughts about methods to make the AI immune to law changes

Post by zxaber » #472937

A traitor with a hidden upload can be absolute cancer for the crew, and especially the AI. There is little the crew can do except have someone ready with their own upload at all times, and the AI will get yanked back and forth constantly between protecting the humans and killing them all. I'd like to explore some ideas that I could PR if any are popular enough.

I want to preface this by saying that I don't think restrictions on uploading AI laws would heavily affect borg balance. While law changeing is a valid tactic against rogue borgs, it's not guaranteed (against malf), and borgs have other harder counters anyway. I am willing to hear counter-arguments.

Anyway

• Idea 1: Uploads can only affect an AI in a core. This would mean that, to stop an AI from being constantly (un)subverted, you would card it while its laws were set as you want them. This would have a fairly big penalty for both the AI, and whoever is on the AI's side, as said AI could no longer use cameras at all. This means no tracking people, or opening doors, or general borg coordination. It's worth noting that this would also affect AIs in mechs, but mechs have their own general weaknesses, and any traitor roboticist massing AI-controlled mechs could also just brainwash people into being his death-mech pilots. In fact, he could even delegate the construction of said mechs to the pilots, freeing him up to do other traitor things.

• Idea 2: Core Upgrade. This would be a research-node item (one node after AI tech?) that would result in a printable disk, which when applied to a core would prevent further uploads. Ideally, it would apply to the core itself, so that the AI could be carded and moved if uploads are suddenly required once more. Such a solution has almost no downsides to the AI or crew, except that the upgrade could cost expensive materials. That, in turn, could see it unavailable when needed most, for better or for worse. My main issue with this solution is that there's really no reason for the Captain not to apply this basically every round (probably after uploading some meme law).

• Idea 3: Lavaland Evac. A PR was made a while back to prevent this, ironically after a game where I had told the AI to have its borg send it to Lavaland when I watched a suspected traitor print an AI upload board (in the round in question, the traitor just went to Lavaland too so it didn't help). In any case, the PR didn't actually have the desired effect, but I would consider using this tactic as a form of bug abuse, since there was an attempted fix. The biggest two penalties for an AI being moved to Lavaland is that it cannot use the crew monitor (which is probably a bug in itself) and it most likely doesn't have access to a nearby intercomm to use if comms become ded. In turn, it's also the weakest preventative measure, as the traitor can simply teleport to the Lavaland base, or use the mining shuttle, or use the aux base to get to lavaland and make the upload. This would be less a PR change and more of a policy allow/disallow thing.

Further ideas are welcome. I personally favor the first idea, since it keeps things clear-cut and has the most reasons why the crew would want to keep the AI in their core except when absolutely necessary.
Douglas Bickerson / Adaptive Manipulator / Digital Clockwork
Image
OrdoM/(Viktor Bergmannsen) (ghost) "Also Douglas, you're becoming the Lexia Black of Robotics"
User avatar
Anonmare
Joined: Sun Mar 15, 2015 8:59 pm
Byond Username: Anonmare

Re: Thoughts about methods to make the AI immune to law changes

Post by Anonmare » #472992

Technically carded AIs are already immune to Law changes if their wireless is disabled.


An AI that is not powered should not be eligible for law changes (a risky method of avoiding law changes, since you're stunned while the power is off and slowly dying).
Upload consoles could announce the fact a law change occurred over the command channel but not give specifics (hacked modules make no such announcement).

A drastic method might be to make Upload consoles unique and not mass-producable. Circuit boards would need to be hardened against acid/cremation in return but ti would make AI subversion a much more risky endeavour
Image
Image
Image
somerandomguy
Joined: Sun Nov 05, 2017 7:41 pm
Byond Username: Astatineguy12

Re: Thoughts about methods to make the AI immune to law changes

Post by somerandomguy » #472998

Lavaland evac shouldnt be considered bug abuse if the fix wasnt merged, and is irrelevant if it was
User avatar
zxaber
In-Game Admin
Joined: Mon Sep 10, 2018 12:00 am
Byond Username: Zxaber

Re: Thoughts about methods to make the AI immune to law changes

Post by zxaber » #473004

Anonmare wrote:Technically carded AIs are already immune to Law changes if their wireless is disabled.


An AI that is not powered should not be eligible for law changes (a risky method of avoiding law changes, since you're stunned while the power is off and slowly dying).
Upload consoles could announce the fact a law change occurred over the command channel but not give specifics (hacked modules make no such announcement).

A drastic method might be to make Upload consoles unique and not mass-producable. Circuit boards would need to be hardened against acid/cremation in return but ti would make AI subversion a much more risky endeavour
If possible, I'd like to not make the actual subversion more difficult, just to give the crew some way to fight back if/when a subversion is found out. With that said;

•Upload consoles could announce law changes - I feel like this tips the balance away from any antag without an uplink (and also slightly towards uplink'd antags), to the point where subverting an AI is almost useless. As soon as that report goes out, the command team will likely be either banging on the upload to reset, or else just grabbing EMP genades or the Ion to kill the AI and figuring out the laws later. With this, we might as well remove the non asimov-clone boards because no one is gonna one-human an AI if it tells the command staff.

•Making upload consoles boards unprintable - This one puts all the power into he who holds both console boards. It would mean the AI gets jerked around less, and that subverting in the first place would be more difficult, but also that the moment the AI gets subverted, they likely will never be fixed; a good tator will hide the boards, and otherwise a borg will do it. It would, for the record, also encourage captains or RDs to just secure the spare upload board at round start. You could policy against that, but it would end up feeling like another nuke ship rule.
somerandomguy wrote:Lavaland evac shouldnt be considered bug abuse if the fix wasnt merged, and is irrelevant if it was
The fix was merged, but it didn't work. It was intended that AIs on Lavaland would be unable to affect machines on the station, but it was unsuccessful. Either way, an attempt at fixing it was made and merged, so I'd chalk it down as unintended.
Douglas Bickerson / Adaptive Manipulator / Digital Clockwork
Image
OrdoM/(Viktor Bergmannsen) (ghost) "Also Douglas, you're becoming the Lexia Black of Robotics"
Tlaltecuhtli
Joined: Fri Nov 10, 2017 12:16 am
Byond Username: Tlaltecuhtli

Re: Thoughts about methods to make the AI immune to law changes

Post by Tlaltecuhtli » #473006

give ai uploads a gps signal when they are on
User avatar
MisterPerson
Board Moderator
Joined: Tue Apr 15, 2014 4:26 pm
Byond Username: MisterPerson

Re: Thoughts about methods to make the AI immune to law changes

Post by MisterPerson » #473129

I like the core upgrade idea. It's risky for both the AI (surprise law upload, carding, EMP) and the person applying the upgrade (woah there AI why are your turrets set to kill?). Lots of fun.
I code for the code project and moderate the code sections of the forums.

Feedback is dumb and it doesn't matter
somerandomguy
Joined: Sun Nov 05, 2017 7:41 pm
Byond Username: Astatineguy12

Re: Thoughts about methods to make the AI immune to law changes

Post by somerandomguy » #473154

MisterPerson wrote:I like the core upgrade idea. It's risky for both the AI (surprise law upload, carding, EMP) and the person applying the upgrade (woah there AI why are your turrets set to kill?). Lots of fun.
allow changes sometimes or something so it's laws aren't fixed
i.e. traitor subverts ai and upgrades it
Dr_bee
Joined: Fri Dec 23, 2016 6:31 pm
Byond Username: DrBee

Re: Thoughts about methods to make the AI immune to law changes

Post by Dr_bee » #473159

A cooldown on law changes besides reset or purge would be a simple change to mitigate some of the horror for the AI.
User avatar
DemonFiren
Joined: Sat Dec 13, 2014 9:15 pm
Byond Username: DemonFiren

Re: Thoughts about methods to make the AI immune to law changes

Post by DemonFiren » #473173

the most fun I've had in any round was that one where my master AI had an amazing 164 law changes

that was "fun" as in "‼FUN‼"
Image
Image
Image
ImageImageImageImageImage

non-lizard things:
Spoiler:
Image
Stillplant
Joined: Thu Jan 17, 2019 7:58 pm
Byond Username: Stillplant

Re: Thoughts about methods to make the AI immune to law changes

Post by Stillplant » #477171

Introduce a law upload authentication code. Each AI core has one, which can be set by using a multitool on the board.

AI upload boards have a list of codes. A upload can only upload laws to an AI if it possesses the correct authentication code. All roundstart upload boards are authenticated to upload to the roundstart AI. The code can be found on a piece of paper in the Research Director's office.

This would solve the problem that subverting the AI can be done by printing 2 boards and buildig a console, but it isn't fully secure.
Post Reply

Who is online

Users browsing this forum: No registered users