Humans regulate each other by the behavior we call ‘property’. AI’s that do also will readily simulate human behavior.
Choices require a means of decidability. Property is the only decidable value that is calculable(rational)+cooperative.
All human moral intuitions are reducible to prohibitions on imposition against various inventories (property).
Ergo, any AI algorithm requires decidability, and one that may not violate such impositions will produce moral actions.
Humans suffer from pre-cooperative impulses for survival that are non rational for an AI to develop unless by design.
Ledgers (~blockchains) are necessary for AI’s to gain access to external actions, access regulated by non AI algorithms.
But that said, I have seen nothing that even vaguely approaches AI. Only systems that process discreet data faster that can – and give us the illusion of intelligence by doing so.
Ai’s, like mathematical axioms, produce deterministic consequences that appear ‘magical’ to us. But AI exists in fact rather than illusion, if and only if the algorithms are capable of free-association followed by Introspective regression-testing for falsehood (survival of possibility).
Consciousness is produced by the (brief) memory of continuous comparison of changes in state, perceived by continuous searching of memories. The thing we call ‘mind’ is just a bag of emotions that react to changes in state of property-in-toto, and use (very) short term memory to accumulate emotions and associate them with those memories.
I’ve been working on this problem since the early 80’s and our lack of progress is still a problem of hardware.
That said, an intelligence will always merely fool us.
Anything intelligent in the sense we mean it, will compete with us the way other creatures compete with us.
And that is the last thing we want to bring into this world.
Ergo, an assertion that the first law of decidability is that property-in-toto may not be violated – no involuntary cost may be imposed by action or inaction against property in toto.
Ergo, an assertion that the second law of decidability is that any agent capable of choice or action, must be monitored by a non-sentient moral agent that prohibits the cognizance of, or action upon, any cost that wold be imposed involuntarily against property in toto.
This agent can cause analogy to pain (cost) upon any concept or action that would prohibit the calculation (use of) that memory or concept that imposes cost, and prohibit entirely action that would impose an involuntary cost.