A Treatise on AI Laws

[kfive]

To remove confusion in how laws should be processed, whilst still ensuring both priority and some level of difficulty when fucking with laws, I propose these options, both of which should be valid.

Interpretation Set A - Sequential Processing:
1: Implicit definitions are definitions nonetheless. When Asimov 1 uses "human" or "harm", the terms are defined as per common sense.
2: Definitions persist within a single application until changed. If law 5 changes the definition of "harm", then only lower priority laws (5-6+) should be affected. If it would lead to a violation of higher priority laws (1-4), it is discarded where necessary.
3: Definitions do not persist across multiple applications of the law. The lowest-numbered law using a term always dictates the definition until redefined.
4: Non-definitions that are related to laws, and definitions which specify that additional data may be required to complete the definition, can preserve their definitions.
5: In the event of a conflict of interest between two intentions on the same law, the AI's discretion is permitted.

As with the second set (coming soon!), all default lawsets will function identically. This simply specifies how to handle extra laws. This interpretation set ensures that:

A: Laws redefining things in antaggy ways have to be placed before functional laws. As it stands right now, why buy the hacked AI board when the freeform one works too? Does the OneHuman board need to exist if you can do the same with freeform?
B: There is a clear priority in terms of definitions. If a law is lower numbered than another law, then you only have to look at it and the laws more important than it to get a sense of how it applies.
C: If the RD is a retard and doesn't know how to actually mess with laws beyond slapping on whatever sounds good, they won't accidentally fuck up humanity.

Note: This thread exists because of an argument on redefining human with law 4. On the one hand, if it is possible to redefine human with law 4, then law 5 should be able to redefine it again. If in that scenario the law 4 definition would take priority, law 1's implicit definition should take even more priority. This gets rid of that maddening ambiguity and replaces it with a simpler, one by one approach.

~k5

[kfive]

Placeholder for part 2, another approach that is more common but, I believe, more flawed.
Edit: Turns out that that this one's somewhat harder to explain than the last. And also that I'm too lazy to focus on it quite as much as the other.