Re: Off Topic: NLP, le to'e melbi glico bo bangu

[xod <xod@bway.net>]

On Fri, 19 Feb 1999, Christopher Palmer wrote:

> From: Christopher Palmer <reid@pconline.com>
> 
> On Fri, 19 Feb 1999, xod wrote:
> 
> > If you follow Moore's Law, and count the number of neurons it enables us
> > to simulate, you'll see that my prediction, though not certain, is
> > respectable.
> 
> The problem is not one of hardware inadequacy, which is all Moore's Law
> refers to. An 8088 is probably *fast* enough. :P The problem is that we
> don't understand the problem (human language).


Who cares? Little kids don't understand the problem either. Just throw
enough neurons at the problem, and you're good.


> > Extend us off silicon into optical/protein/full nanotech computers, and
> > you'll agree that real-time better-than-human language translation of bad
> > handwriting and drunken accented speech is inevitable.
> 
> ? I can do that already without biomechanical augmentation. :^) Adding
> human brains to hardware is not solving the problem, and would probably
> kill the human anyway. :^)


I wasn't suggesting cyborgs, just really good computer systems that are
better than humans at deciphering text, handwriting, or speech.


> > "English is hard" is meaningless, but "English is harder than Spanish"
> > is both meaningful and measurable. 
> 
> What, exactly, does it mean? And how are you measuring 'meaning' and
> 'difficulty' and 'Spanish' and 'English'?


Take a random sample, divide them up, teach them different languages and
compare their abilities to read a newspaper, follow a soap opera on TV,
discuss their sex lives, etc, after a certain duration of education.



-----
How's my typing? 1 (800) 243-6624