Re: Does a parental model work for having A.I. that can be trusted?

Eliezer S. Yudkowsky (sentience@pobox.com)
Wed, 04 Aug 1999 18:48:10 -0500

You're being far too anthropomorphic. It doesn't make a damn's worth of difference whether or not AIs are raised permissively or harshly, they can't possibly get a "taste for blood", and I'd be seriously surprised to ever find two AIs with the same architecture and conflicting goals.
"Parenting" isn't going to help at all. All of these things rely on a
base of human emotions that won't exist in an AI.

I suggest:
http://pobox.com/~sentience/AI_design.temp.html

The sections you want are "Interim Goal Systems" and "The Prime Directive".

This should give you some idea of what the inside of a first-stage AI's mind is likely to actually be like.

-- 
           sentience@pobox.com          Eliezer S. Yudkowsky
        http://pobox.com/~sentience/tmol-faq/meaningoflife.html
Running on BeOS           Typing in Dvorak          Programming with Patterns
Voting for Libertarians   Heading for Singularity   There Is A Better Way