X-Message-Number: 30176 From: Date: Thu, 20 Dec 2007 03:16:07 -0500 Subject: Re: "evil" AI & consciousness --_----------=_1198138568305909 Content-Disposition: inline Hello again Mr. Ettinger. You said, in quotes: "My point is that a system without feeling, without subjectivity, cannot have motives in the sense that we do. It doesn't want anything or fear anything. It can only have programmed goals, states to attempt to reach or to avoid, which is very different." You just described my ferocious grizzly bear, in better detail than I did. Except that it does "want" something, just as a programmed machine would. Some other species of bear fear humans and will run, unless they are very hungry or trapped. If that grizzly bear is coming at you, though, believe it - he has a programmed goal, and fears nothing. You can shoot him and he will still come at you, unless you break bone in the skull. "These goals must be very explicit and unambiguous. Any attempt by the programmer to paint with a broad brush will inevitably result in freeze-up, and trying to foresee all future possibilities in detail is hopeless." Many animals, when frustrated by ambiguous programming, will just run around aimlessly, until they find something that seems to match their programming, such as that grizzly bear, and then he comes after you to take a big bite out of wherever he can. Things happen already in computers, akin to this - programs running haphazard until they end up causing damage to your hard drive or causing the operating system to fail. "In any case, to repeat myself, when some programmer thinks he is near a super-intelligent program, he will build in safeguards, e.g. in certain situations requiring a pause for external input. That there is little present effort to do this simply reflects the fact that such programs are nowhere on the horizon." He will?? Maybe you would, because you are a nice fellow. But some folks, like those who run the SIAI, don't seem to believe in any safeguards at all. If they do, why don't they ever talk about them? But let's assume for a brief second that any programmer would, as you say, build in a pause for external input when he thinks he is nearing a super-intelligent program. You certainly cannot conclude that there is no such proximity to achieving a super-AI, because the programmer may not know when the conditions are present to satisfy that. We certainly do not know now ahead of time what algorithms are necessary to spawn such intelligence. The other factor is publicity. You cannot say you know anything about what "present effort to do this" indeed exists, because the general public is not privy to all ongoing research projects in any field, much less this one. -- Got No Time? Shop Online for Great Gift Ideas! http://mail.shopping.com/?linkin_id=8033174 --_----------=_1198138568305909 Content-Disposition: inline Content-Type: text/html; charset="iso-8859-1" [ AUTOMATICALLY SKIPPING HTML ENCODING! ] Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=30176