I have several motivations for chasing after what is admittedly a very non-standard form of representation and one certainly not guaranteed of success!
First is that low-level insect-like controllers are a snap and a breeze to do in it, and I'm guessing that the whole brain developed from such humble beginnings by the same tricks evolution always uses: copy, proliferate, and modify. Second is that I think the absolute bottom line to AGI is autogeny, the ability to extend and modify yourself. To do this I need a substrate that's capable of expressing any new ontology the expanded mind might need. Third is that I just spent a couple of years doing molecular modelling and all the techniques are the hammer I'm holding -- so every new problem looks like a nail. But note that there is a pile of well-known algorithms (incl. free sorftare) for applying big iron to problems posed in this form. As I mentioned, LOTS of physical science and engineering is represented in n-spaces. Marvin Minsky opined to my face that I have physics envy :-) And I'm relatively happy to agree. Fourth is that mapping into n-spaces helps avoid McDermott's "natural stupidity" problem. You can't just label a predicate with an English word and think you've captured the meaning. The key, once again, is to find the transforms that allow you to put the high-level concepts into spaces whose geometric properties reflect real, useful regularities in the reality they're representing. My main research at the moment is to find better ways to do that than simple search -- but because frame-trajectories have a predictive ability, I can at least use that as an evaluation function if I have to search. Josh On Saturday 25 November 2006 13:52, Ben Goertzel wrote: > Hi, > > > On the other hand, somewhat simpler blends can be done by simple > > interpolation or mappings like the analogical quadrature I mentioned. For > > example, you will instantly understand "teddy moose" to be that which is > > to a moose as a teddy bear is to a bear, i.e. a stuffed-animal toy > > caricature. I'm fairly sure I could define a continuous space in which > > such a thing would fall out of the simple geometric formula. > > Sure, I agree.... However, this is not the most interesting kind of > blending.... The question is whether your n-vector representation > makes the hard stuff any easier; making the easy stuff easier is > really not so important if your goal is genuine AGI rather than making > prototypes that look wizzy ;-) > > About Teddy Meese: a well-designed Teddy Moose is almost surely going > to have the big antlers characterizing a male moose, rather than the > head-profile of a female moose; and it would be disappointing if a > Teddy Moose had the head and upper body of a bear and the udders and > hooves of a moose; etc. So obviously a simple blend like this is not > just **any** interpolation, it's an interpolation where the most > salient features of each item being blended are favored, wherever this > is possible without conflict. But I agree that this should be doable > within an n-vector framework without requiring any breakthroughs... > > Ben G > > ----- > This list is sponsored by AGIRI: http://www.agiri.org/email > To unsubscribe or change your options, please go to: > http://v2.listbox.com/member/?list_id=303 ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=303
