Are ethics supposed to be Aumann-agreeable?
If they were, uFAI would be a non-issue. (They are not.)
Ethics ought to be Aumann-agreeable. That would only imply uFAI is a non-issue if AGI developers were ideal Bayesians (improbable) and aware of claims of uFAI risks.
How do you know thatthey are not?
If they were, uFAI would be a non-issue. (They are not.)
Ethics ought to be Aumann-agreeable. That would only imply uFAI is a non-issue if AGI developers were ideal Bayesians (improbable) and aware of claims of uFAI risks.
How do you know thatthey are not?