I am not sure if a super algorithm which was able to generate many different operations for a kind of activity would be of any use. For example, I was thinking that it might be useful in trying to match Hinton's recognition of a handwritten character. The super generator would be able to generate an extensive system of characters and by adding a little morphing latitude the range over the characters would be huge. However, since the solutions have to be defined by the set of individual parameters to the super algorithm that means that it would not be a viable compression method (even though the super algorithm generator that could create an immense number of variations of the character would be.)
The problem space can be too large to be used as a viable method for AGI. The problem is that we need to find good ways to discover and represent solution methods even though the only feasible way to do this will typically overgeneralize (noisy) and undergeneralize (lossy). By using multiple vantages we can better hope to cross-categorize the problem in such a way so that it reduces the overgeneralizations and undergeneralizations. This means that multiple vantages, the different categories that would produce the cross-categorizations that could be more effective for solving the problem, should be given a utility value based on how well they complement each other. In this respect, a super algorithm generator might be useful. If we can make good estimates, based on previous usages of the variants of the analytics super generator, how the variations of the parameters might change the characteristics of the analytical algorithms then could be generated then we might have a concise method for finding good combinations of cross-categorizers (based on the analytical algorithms that were generated) for the problem. (In this case the analytical algorithms would not define every individual that would work as a representative of some solution but simply be used as effective analytical algorithms that could work on a wide variety of cases. So the analytical algorithm super generator is not the same as a character variant super generator that I mentioned above). The reason the super analytical algorithm generator would be useful is only because it might (if feasble) be a concise way of finding the analytical methods that would best complement each other given some particular kind of problem. Jim Bromer On Tue, Feb 5, 2013 at 4:16 PM, Piaget Modeler <[email protected]>wrote: > > Perhaps formulating the problem space is part of the solution: > http://ai.eecs.umich.edu/cogarch0/common/theory/prob.html > > ~PM > > ------------------------------ > From: [email protected] > To: [email protected] > > Subject: Re: [agi] Multiple Vantages Can Be Used to Find Multiple > Observation Objectives > Date: Tue, 5 Feb 2013 10:25:16 +0000 > > > You’ve given two examples of real world problems – “what do those women > really think about me?” / “how to reason about methods of dealing with > ‘inscrutable’ problems”? > > Show how in either (or any real world problem whatsoever) there is or can > be “ a systematic search through possibilities” > > What are the systematic possibilities whenever you have to “read minds” > (as above) – a classification which embraces a vast amount of > psychological RWR. > > What are the systematic possibilities for considering/reading – *what > Obama really intends to do in any foreign policy area”, or Iran intends to > do re Israel, or ......etc > > The whole damn point of RWR is that you don’t have a set of options – you > have to construct options from scratch – and you aren’t going to get > anywhere near a set. What you’re arguing is pretty well the complete > opposite of the truth. > > *From:* Jim Bromer <[email protected]> > *Sent:* Tuesday, February 05, 2013 10:05 AM > *To:* AGI <[email protected]> > *Subject:* Re: [agi] Multiple Vantages Can Be Used to Find Multiple > Observation Objectives > > Mike, lt is a systematic search through possibilities. The > possibilities are limited to reasoned conjectures about the problem. The > preparation does involve using ideas that had been considered before but > the majority of the ideas do not come from a pre-prepared set of > "options". The conjectures do include what I call imaginative projection, > but the use of the imagination in reasoned conjectures is driven by > rational consideration. > Jim Bromer > > > On Tue, Feb 5, 2013 at 4:51 AM, Mike Tintner <[email protected]>wrote: > > Jim, My post explicitly said that a/your guess is a “reasoned > inference.” The key point I’m making – and you’re ignoring – is that it’s a > one-off, one-at-a-time business rather than a narrow AI systematic search > through a pre-prepared set of options (the kind that cause such > “complexity”). And this BTW is the irrefutable truth – no one’s going to > produce an AGI problem where there is a neat set of options. > When you noted the existence of “inscrutable events” – (I would say > “partly invisible events/objects”) – you were onto something big, taking a > big step forward in your AGI thinking. When you started looking for > “reliable” ways of solving problems about them – (trying basically to cling > to the old narrow AI ways) – you took a big step back. Concentrate on the > invisible nature of the subjects of real world problems. There is no > reliable way to deal with them. You just gotta get stuck in and guess, or > if you prefer “hypothesize”/”theorise”. > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/6952829-59a2eca5> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/19999924-5cfde295> | > Modify <https://www.listbox.com/member/?&;> Your Subscription > <http://www.listbox.com> > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/10561250-470149cf> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
