From: utzoo!decvax!yale-com!leichter Newsgroups: net.ai Title: Re: The Mind"s I Article-I.D.: yale-com.711 Posted: Sat Jan 22 11:11:48 1983 Received: Mon Jan 24 03:59:44 1983 References: yale-com.707 How will you make an intelligent machine do the dirty work for you? Human societies had slavery until very recently; getting HUMANS to do your dirty work for you is a (depressingly easily) solved problem. The basic trick is to make the slaves believe that the situation they are living in is right - and make sure they have no hope of any other life - like just across the border. Since you control their environments, especially their upbringing, you can arrange to do the former; the latter is a matter of what kind of society your neighbors have. Historically, slave revolts are a fairly rare, and even less often successful, occurence. What about intelligent computers? Here, you don't even have to worry about indirect methods of indoctrination; you can control the data base the systems start with, what kind of likes and dislikes they have, and so on. Any program that was really like a human would have a (PERSONAL) concept of pain. It would be easy to include some simple command that triggers intense pain. Further, it's unlikely that anywhere in the world would a society of free computers exist; there would be no "underground railroad" to run to. While I agree that there are real MORAL questions to deal with here, I think the PRACTICAL issues would be pretty easily solvable. Of course, you could argue that a real ability to revolt is a necessary part of a "really intelligent" program. In the abstract, you would be right; it \\\if what you want is an accurate model of HUMAN intelligence, that would probably be a necessary part. However, we have no trouble recognizing as human the "faithful manservant" who really believes that "his place" is to serve. Talking intelligently to such a person is not particularly hard. -- Jerry decvax!yale-comix!leichter