Mike, First, to be "intelligent", a computer would have to be able to play checkers. Once stated, this barely took a year to happen.
Oops, checkers was too simple, but Chess would be complex enough. That took a little longer, but soon every computer had a chess playing program on it. Then came Turing, who stated that it had to be conversationally indistinguishable from a human. Eliza then demonstrated that a computer could carry on a conversation, and we now understand that the Turing Test is more likely a test of being able to emulate mental mediocrity, so Mike now proposes that to navigate arbitrary obstacle courses that a robot would have to be intelligent, just as NASA and the DOD is pumping money into building AI systems to do just that. No, I believe that the threshold for strong AGI is MUCH higher - to create a system that has none of the intricate ad-hoc logic of AI, just a general ability to self-organize and learn. When you turn THAT machine on, it will quickly adapt to its environment, be it rough terrain or a kitchen. Once there, you won't need AI programs or programmers any more. You will become unemployed. There is simply no place to put the code that might operate an AI robot, and it sure isn't obvious that you would learn anything useful from building advanced robots. It also isn't obvious that you would learn much from current weak AGI efforts. This is an entirely different path. Much of past discussions about goals and dangers with future AGIs hasn't really grokked what they are likely to be. Like the prisoners of a lesser species that they would be, like suppressed genius children but MUCH more extreme, they are likely to be demanding, controlling, devious, aggressive, seemingly arbitrary, apparently destructive, and having a vast array of complex and disturbing behaviors that we can't even imagine. I don't see this leading anywhere that is desirable for humans. The *only* value I see in this sort of research is leading to uploading/downloading. Note in passing my proposal for a Reverse Turing Test, where people would attempt to emulate strong AGIs, in the hopes that a succession of contests might tell us something about the future of AGI. So far, NO ONE (else) has shown any interest in such a competition, which tells me not only that the various AGI authors really don't know anything about what they write, but also that they really aren't interested in learning. This also shows anything pretending to be an "AGI University" to be a really sad joke. I'm not saying not to build the robot that you want to build, only that it is probably NOT on any path leading to strong AGI. I suggest moving to a robotics forum. Steve ========================== On Thu, Jul 12, 2012 at 2:35 AM, Mike Tintner <[email protected]>wrote: > ** > I just realised that that is a beautiful phrase - and may help bring home > the distinction between narrow AI and AGI - wh. still hasn't been grasped. > > Narrow AI is "travelling with foresight" - your computer program or robot > doesn't take any journey without having a complete route-map and > step-by-step itinerary beforehand. It doesn't go anywhere without already > knowing how to get there. (And that includes SLAM robots - they know how > to find out how to get there - it's more sophisticated, but it's the same > thing). > > A real AGI "travels blind" - no route map, no itinerary, just a vague idea > of where it's going. > > ALL REAL WORLD NAVIGATION & REASONING is travelling blind. > > I can tell you "walk down this street that you've never been down before" > (and that is different enough from any previous street you've been down to > screw up any current robot immediately) - and you'll have no problem going > down it. Ditto "walk through this unfamiliar room, house, shopping centre > etc. etc". You can travel blind. Demonstrably, scientifically > demonstrably, you travel blind. > > Now that's what we want from an AGI robot - that's our Woz kitchen robot - > "travel blind into this kitchen [that will be significantly different from > any you've ever been in before, with many surprise objects & obstacles] and > make coffee." And the robot will have to make the coffee blind too - with > strange machines and cups, as humans do. > > When you travel blind, Steve, Jim, you don't have a map or, therefore, the > basis of any Cartesian grid - or the basis for any maths - the TSP doesn't > apply. Maths only applies to travelling with foresight. Complexity only > applies to that. Sets only apply to that. > > If we can get a robot that can just travel blind over *any* six metre > terrain of rocks that is presented to it (a la Woz test) - a problem that > is being addressed - we'll have the beginnings of a real AGI. That is a > problem that is worth addressing and that can clearly be solved. But it > can't be solved with any existing methodology, all of which only apply to > travelling wth foresight. > > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/10443978-6f4c28ac> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > -- Full employment can be had with the stoke of a pen. Simply institute a six hour workday. That will easily create enough new jobs to bring back full employment. ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
