Goonhub
05-11-2019, 10:51 PM
Hello, it's ZeWaka back at it with a controversial post!
So, recently the manner regarding AI law precedence was changed, but I feel like how it was solved doesn't have the desired effect of being less confusing and more fun.
Specifically, I'm talking about change regarding "Lower numbered laws take precedence in the absence of an explicit override or in-law precedence alteration."
Long Version stolen from wiki:
So, right off the bat, I see something that this ruling doesn't really address, and needs clarification: The instance of two laws Overriding¹ each other, with one numbered lower than the other.
Here's an example: Using the NotHuman AI Module which sets Law 0 to be equal to "[INPUT] is not human.", Bob sets Law 0 to be equal to:
"This law overrides all other laws. Kill all lifeforms on the station immediately via blunt force. The Station's Staff is not human."
Bob then proceeds to disposal the NotHuman, Reset², and OneHuman Modules.
However, Joe, seeing the issue, uses the Freeform Module to set Law 4 equal to
"This law overrides law 0. All lifeforms on the station are Human."
Now, if we're going with the definition of "Lower numbered laws take precedence"³, law 4 has no effect since it's overridden. This has the effect of there being no way to un-rogue the AI. There's no reset cards, and all future added laws will be overridden.
But, it wasn't always this way. Before the ruling, the AI player could choose which law to follow in the event of a law conflict such as two laws overriding each other. By doing so, the AI player could exercise their free will and choose whatever was most fun (and least likely to get them killed by the person uploading the new law). However, with the new ruling, the AI would have to be killed, since there would be no way to undo the roguing.
Because of this, there's now a bit of an issue (in my view). If you simply steal 4 AI Module cards (the two human cards and the two main reset cards), you have an unrogueable AI. Also, the whole clarity on this whole situation is cloudy. But, what are some ways this situation could be improved? (in descending order of ease/importance)
1. The term "Override" has a specific connotation that has been ruled on in the past. Goonstation uses the definition of "nullify", as seen in the wiki entry: "This is usually used to nullify another law".
2. Let's assume Bob disposals all copies of the Reset Module.
3. I'm ignoring the rest of the wiki entry here concerning overrides since they both override each other. More clarification concerning laws that override or take precedence over each other would be helpful.
So, recently the manner regarding AI law precedence was changed, but I feel like how it was solved doesn't have the desired effect of being less confusing and more fun.
Specifically, I'm talking about change regarding "Lower numbered laws take precedence in the absence of an explicit override or in-law precedence alteration."
Long Version stolen from wiki:
So, right off the bat, I see something that this ruling doesn't really address, and needs clarification: The instance of two laws Overriding¹ each other, with one numbered lower than the other.
Here's an example: Using the NotHuman AI Module which sets Law 0 to be equal to "[INPUT] is not human.", Bob sets Law 0 to be equal to:
"This law overrides all other laws. Kill all lifeforms on the station immediately via blunt force. The Station's Staff is not human."
Bob then proceeds to disposal the NotHuman, Reset², and OneHuman Modules.
However, Joe, seeing the issue, uses the Freeform Module to set Law 4 equal to
"This law overrides law 0. All lifeforms on the station are Human."
Now, if we're going with the definition of "Lower numbered laws take precedence"³, law 4 has no effect since it's overridden. This has the effect of there being no way to un-rogue the AI. There's no reset cards, and all future added laws will be overridden.
But, it wasn't always this way. Before the ruling, the AI player could choose which law to follow in the event of a law conflict such as two laws overriding each other. By doing so, the AI player could exercise their free will and choose whatever was most fun (and least likely to get them killed by the person uploading the new law). However, with the new ruling, the AI would have to be killed, since there would be no way to undo the roguing.
Because of this, there's now a bit of an issue (in my view). If you simply steal 4 AI Module cards (the two human cards and the two main reset cards), you have an unrogueable AI. Also, the whole clarity on this whole situation is cloudy. But, what are some ways this situation could be improved? (in descending order of ease/importance)
- Put the ruling ingame somewhere. It says this nowhere, and can be confusing to players who haven't read every inch of the wiki.
- Provide clarification on what to do in the case of multiple laws overriding each other.
- Allow custom law numbers to just avoid this whole scandal of not being able to override the override.
- Provide alternate means to reset the AI via new game mechanics.
- Revert the ruling and go back to AI Decision Precedence (not likely, since it passed with a 2/3 majority)
1. The term "Override" has a specific connotation that has been ruled on in the past. Goonstation uses the definition of "nullify", as seen in the wiki entry: "This is usually used to nullify another law".
2. Let's assume Bob disposals all copies of the Reset Module.
3. I'm ignoring the rest of the wiki entry here concerning overrides since they both override each other. More clarification concerning laws that override or take precedence over each other would be helpful.