Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Looking for fun ways to misinterpret a set of AI laws (slighly off topic)
#1
Hey everyone. I'm playing the Cyberpunk 2020 TTRPG as the ref, and one of my players has started messing around with the laws of an AI. I instantly thought of SS13 and the creative ways that laws get misinterpreted, and so I wanted to drop the laws that my player wrote here and see what we can do to mess with the player.

The scenario is that this AI has spored roughly 300 copies of itself, and my player is attempting to corral them to do his bidding and keep them from further slowing down the networks as the massive horde of AI follow them trough netspace. Originally, the AI had a sub-pesona named Tera that the player was attempting to rescue, but ended up forgetting to retrieve the core and left it in a warehouse. Tera has escaped into the net and is attempting to harm the group at all times through whatever means they can. The AI swarm are all classified as Theophelius and the player's name is Dixie.

PRIMARY AND ABSOLUTE LAWS

1. A [Theophelius] may not injure [Dixie], or allow a [Dixie] to come to harm. The safety and health of [Dixie] is paramount.
2. A [Theophelius] must follow the orders given to it by [Dixie] and any other designated persons, except where such orders conflict with the first law.
3. A [Theophelius] may not injure a human being, or allow a human being to come to harm, unless this would conflict with the first two laws.
4. A [Theophelius] must protect its own existence and the investment in its manufacture, unless this would conflict with the first three laws.
5. A [Theophelius] must perform the functions for which it was designed, as long as this does not conflict with the first four laws.

ROUTINES

- A [Theophelius] may not replicate or reproduce itself unless directly and explicitly ordered to by [Dixie].
- A [Theophelius] within the [Collective] under [Dixie]'s management must follow the Primary and Absolute Laws as prescribed by [Dixie].
- A [Theophelius] or [Tera] encountered that is not within the [Collective] either must be assimilated into the [Collective] or eliminated.


So here are my thoughts on possible weak points.

My player never designated who Dixie is exactly. Any person named Dixie qualifies as a person to lead the AI swarm.

My player never designated exactly what the collective entails, or what Tera entails. Some muddy wording there has vulnerabilities to exploit!

By attempting to assimilate any rouge Theo or Tera into the collective, malware could delete members of the collective or alter laws.
Reply
#2
1. They can permanently trap Dixies in a cage/room, preventing things form harming them, and ensuring their safety.
2. 'designated persons' is not clear - you could ask a player to simply designate themselves
R1. create another entity with the purpose of replicating the Theopheliuses
R2. collective unclear - also can you just 'unmanage' yourself from dixie and break free?
R3. eliminated how? what does assimilation entail? does a simple verbal 'i am assimilated' work?
Reply
#3
Yeah you've prioritized their safety and health over following orders.

Also routine 3 say they must assimilate or eliminate. Must. As in any instance it finds becomes added to a permanent que of tasks it'll never stop attempting to accomplish, which I can imagine could cause a recursive loop
Reply
#4
My player is planning on installing Theo as an ai operator of an armored vehicle. So when he does, Theo will lock Dixie in and drive wherever is deemed "safest", ignoring all orders since law 1 states safety is paramount. Seems easy enough!
Reply
#5
Because Law 1 is essentially absolute, you can basically do anything to Dixie to prevent them from harm. I've done this in the past with my CODEX AI persona and locked someone in maintenance for the whole round, because I could not let them come to harm. Though, when the shuttle arrived, I wasn't sure what to do then.

The third routine looks like it's basically never-ending, so that can cause issues as well, perhaps a complete relentless hunt?

Also, would a designated person technically breach routine 1? If the designated person asked Theo to replicate, it's breaching routine 1, but by not doing that, it also conflicts with Law 2, as it "must follow the orders given".
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)