Problem is that even if fonix' solution does turn out to be the one which wins in the end who's going to tell Jerry?
Bob, I know your right about the general point that there are almost always at least two ways to, is it "skin a cat", but I don't think this is going to be like choosing between rye, whole wheat, basic white, pumpernickel etc etc, breads. At the moment there is an opportunity for some chipsets with NN architecture to fill niches, but in the future it might be more like choosing between Windows and OS/2.
Just suggesting metaphors, but if NN architecture prooves to be the thing which does really emulate the way our minds are working, ie , with small amounts of memory and/or processing power (a decreasingly significant advantage, granted, although how many calories of chemical energy does our brain consume to think)and fonix have found a way to make it work at last, then this may be the way. It does'nt necessarily have to be fonix's NN solution which eventually dominates (if one must)but HMM appears to have some basic flaws. Chasing the pattern of a word amongst hundreds of thousands is an inelegant solution (who cares about elegance). I get the sense that many solutions to technology problems never get realized because the idea is never researched despite the need, while a working idea may be fully explored.
So even if fonix tech is not ideal or the only solution, its beginning to get a lot of attention while almost all of the alternatives are still HMM (thankyou Russ Wilcox and notice my almost)
Take care all
Parx |
