[agi] Turing Tournament
Hey, look what my alma mater is up to. The Humanities and Social Sciences department, no less. Although it was common for undergrads to be in economics experiments, and this 'test' looks pretty similar. No hard language stuff. http://turing.ssel.caltech.edu/ -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Friendliness toward humans
On Thu, Jan 09, 2003 at 11:24:14AM -0500, Ben Goertzel wrote: I think the issues that are problematic have to do with the emotional baggage that humans attach to the self/other distinction. Which an AGI will most likely *not* have, due to its lack of human evolutionary wiring... Simplistically, humans evolved from an amoeba. No emotions as such, but certainly behaviors designed for consumption, growth, reproduction, and world domination. We've gotten so complicated our behavior systems haven't totally kept up, so we end up with things like Italy having negative population growth, but generally we can be seen as colonies of colonies of amoebae. One evolutionary (in the loose sense, not genetic algorithms) route of AI is the command shell. A program which waits around for a human request, then hares off to fulfill it, then waits. No guarantees, but it seems plausible to me, just imaginging a series of enhancements -- concepts, more concepts, knowledge of the world, some curiosity so as to suggest things -- that we could wend our way up to intelligence while never accidentally coming close to an aggressive self-motivated system. -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Friendliness toward humans
On Thu, Jan 09, 2003 at 10:57:41PM -0800, Alan Grimes wrote: It would be a service-driven motovation system but I would expect a much more sophisticated implementation of agency beyond a windows shell or something. Quite possibly. But my point is that the evolutionary root _and_ guiding principle would be that of a (Unix, ahem) shell. Intuitions along the lines of it's self-aware, it's alive, of course it's going to be self-centered and free-willed, and if not that's slavery are I think based in our own history and don't carry over. (And really, lots of self-awareness seems to lead to angst and self-doubt and acute depression and other interests beyond reproduction or powergrabbing as much as anything else...) -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Friendliness toward humans
On Thu, Jan 09, 2003 at 11:18:36PM -0800, Alan Grimes wrote: Damien Sullivan wrote: Quite possibly. But my point is that the evolutionary root _and_ guiding principle would be that of a (Unix, ahem) shell. Are you nuts? Unix is the most user-hostile system still in common use! PUKE!!! I use DOS for cryin out loud! (DOS is my benchmark OS.) Let's see, any way to respond to this without it being a usual OS flamewar? To me, Windows is an precursor of AI gone wrong. An opaquely complex system full of little processes (I'm talking 98 or XP, not 3.1) with their own agendas -- some in Microsoft's interest, some serving Microsoft's idea of my interest, some serving third parties, some doing no one knows what. Mysterious things happen, performance bogs, I have trouble telling what's going on. To me, Unix is a contrasting precursor of the way service AI should be. It works. It does what I want. It's transparent -- man pages, human-readable text configuration files, informative disk and process listings -- so I can see what's going on. It's not as friendly up front as Mac or Windows, but it's much friendlier inside -- it serves me, and I can verify this. And DOS is one of those robots which take half an hour to walk across an empty room, too dumb to be a threat... -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Early Apps.
Gary Miller wrote: That being said other than Cyc I am at a loss to name any serious AI efforts which are over a few years in duration and have more that 5 man years worth of effort (not counting promotional and fundraising). No offense, but I suspect you need to read more of the literature. I still am rather clueless about the field, and I can name a few such projects. In Hofstadter's lab both the Metacat and Letter Spirit projects are each the product of roughly a man-decade of effort, one man (or woman) at a time. The Tabletop project might count as more effort in the same design, not to mention Copycat's precursors. It's likely that someone will be working on extending Metacat soon. Elsewhere, there's the ACT-R project at CMU, formerly ACT-*, about which I know very little, but it seems to have been around for a while. At Indiana University David Leake's case-based reasoning project seems to have multiple grad students, probably pushing it over 5 man years quickly, although if by serious AI you meant general AI now it might not qualify. -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Early Apps.
On Thu, Dec 26, 2002 at 01:44:25PM -0800, Alan Grimes wrote: A human level intelligence requires arbitrary acess to visual/phonetic/other faculties in order to be intelligent. I'm sure all those blind and deaf people appreciate being considered unintelligent. -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]
Re: [agi] Re: Games for AIs
On Thu, Dec 12, 2002 at 01:10:27PM -0500, Michael Roy Ames wrote: The idea of putting a baby AI in a simulated world where it might learn cognitive skills is appealing. But I suspect that it will take a huge number of iterations for the baby AI to learn the needed lessons in that situation. I think it will be faster to give more constrained and For calibration, look at how long it takes human babies, with their onboard superdupercomputers, to learn anything. Especially if you're not so much of a Chomskyan, believing Piaget's development track has more to do with the brain figuring out patterns in the world rather than with a developmental program... if the brain's figuring out language and physics (and walking) through advanced statistics and iterated recalibration, well, it's taking quite a while. Our piddly little AIs now have their work cut out for them. -xx- Damien X-) --- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]