Okay, so would it be accurate to paraphrase you as suggesting an AGI MVP should look like a big linear algebra solver? That's an interesting idea if so, and possibly a promising one. The main concern I'd have would be to make sure there's enough of a market for it, i.e. enough problems people care about, that are naturally expressed in that sort of form.
On Tue, Dec 17, 2013 at 6:24 PM, Steve Richfield <[email protected]>wrote: > Russell, > > On Tue, Dec 17, 2013 at 3:21 AM, Russell Wallace < > [email protected]> wrote: > >> If you are trying to design an AGI, what is your design going to be a big >> version of? >> > > Re-mentioning something from the past, but with new emphasis... > > When I was ~15 I discovered numerical series and their ability to predict > future elements, which it appeared to me at that time to be what math was > all about. I set about creating formulas that in effect merged several > different series together and predicted future elements, in the sophomoric > hope that the world was a LOT better organized than it is. This also > paralleled an "intelligence" test that I had taken, which consisted of > lists of numbers and the problem of guessing the next number. It looked to > me at that time like "simple" (page-long) formulas might actually be > judged as being "intelligent"!!! My effort failed for several reasons, some > of which may be repairable, but the dream continues - with AGI being > perhaps the ultimate extension. > > My primary discovery in this early failed effort was that linear algebra > could in effect make decisions, and make "proportional decisions" akin to > Bayesian math WITHOUT resorting to ANY conditional methodology!!! > > Of course, my early effort was only one-dimensional - lists of numbers as > input, with the problem being to predict the next. However, I don't see any > fundamental reason why the concept couldn't be extended to any number of > dimensions. > > Imagine a system of elements (perhaps like biological neurons, but NOT > like people now think biological neurons work) with the capability of > predicting their inputs into the future (with obvious uncertainties and > errors) via substantially linear, or at least continuous means. Some inputs > would be "free" and reflect our intentions. Solving problems would consist > of manipulating the free inputs until a desired result was achieved, > perhaps using methods akin to genetic algorithms. > > In short, maybe our present "elements" of neurons, Bayesian logic, etc., > are just too damn simple for the task at hand. > > OK, I have proposed above one prospective class of super-elements. Does > anyone here have ideas for other super-elements? > > Steve > ======================= > >> On Tue, Dec 17, 2013 at 9:48 AM, Steve Richfield < >> [email protected]> wrote: >> >>> WAY back in the 1960s, a friend of mine, Carl Nicolai, now only known as >>> the inventor of non-deterministic crypto, was promoting the use of >>> pseudo-random carrier waves for spread spectrum communication - to go MUCH >>> farther with less power. As his ideas started to catch on, one UCSF EE >>> professor even put up a web site to debunk such craziness!!! >>> >>> Carl was SO disconnected from the mainstream that he couldn't get his >>> articles published, and his patent promptly received a Secrecy Order. >>> >>> Of course, Carl's thoughts are now accepted as routine communications >>> methods. >>> >>> Now, AGI seems to be caught in a similar sort of "trap" as RF >>> communications once was, where present dogma and "leaders" are leading it >>> to a dead end. >>> >>> Organizationally, AGI appears to be on a suicide mission, thinking it to >>> be SO well targeted that it can enjoy the luxury of dismissing articles >>> about radically different approaches as being "off topic" for conference >>> presentations. Guys like Carl would NEVER have been acceptable to any >>> conference. >>> >>> I think we have all hoped that guys like Jim and Tintner would see where >>> the King has put his clothes. when we can all see quite clearly that the >>> King is wearing them. This has led to an insane level of tolerance, because >>> we all know that all it may take is one CAREFULLY calculated question to >>> blow AGI wide open. Yet - apparently, the question remains unasked. >>> >>> For example, maybe we just haven't figured out yet where in an AGI to >>> put the random number generator? We know that we NEED randomness in game >>> theory and genetic algorithms, so I doubt that an AGI could be guilt >>> without using a random number generator. >>> >>> OK, half a decade of discussions has failed to find the Holy Grail of >>> AGI. Now, how about taking a step back and working on some good QUESTIONS? >>> >>> Any thoughts? >>> >>> Steve >>> >>> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >>> <https://www.listbox.com/member/archive/rss/303/1658954-f53d1a3f> | >>> 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/303/10443978-6f4c28ac> | >> 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> > <https://www.listbox.com/member/archive/rss/303/1658954-f53d1a3f> | > 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
