At 01:30 PM 3/14/2002 -0500, Bagotronix Tech Support wrote:
> > We will be talking to our computers in a few years, do we want to have to
> > say "space" or depend on some AI routine to insert the spaces it thinks we
> > want? Speech has neither spaces nor case. It does have, so to speak,
> > italics. But I don't want a filename to depend on whether or not the speak
> > is angry, for example.
>
>This prediction is so old it is laughable.  It never seems to work out as
>good as it needs to be.

It's getting better. Yes, it is still not good enough, which is why I put 
this at some not clearly defined time in the future. But I don't think it 
will be long. I expect to see it in my lifetime, which right now would be 
estimated at twenty years on the average.

>   I have doubts that computer speech recognition will
>ever be practical until people learn to speak better.

People generally know how to speak better already, when we consider that 
the listener might have some difficulty. I think it is pretty obvious, 
however, that if human beings can decipher speech with reasonable 
reliability, so can computers. Our brains do not run on magic.

>   Everyone has times
>when they mumble, get tongue-tied, use the wrong tense, run their words
>together, etc.

I do not see speech as replacing keyboards. "The hand is quicker than the 
eye." No, the place for speech in CAD is as an *additional* and parallel 
input method.

The eye is also quicker than the ear, but there is also a minor place for 
sound in presenting information to the user.

>   I used to think that the intercom systems at the drive-thru
>burger joint were awful.  But then when I got the pay window, I couldn't
>understand what the person was saying even when it wasn't being said through
>the intercom.  I am from the South, and us folks in the South speak slowly
>and hear slowly.  We are not all mentally slow, however.  I sometimes have
>to ask folks from other areas of the U.S. to slow down when they talk.
>Especially anyone from New York.  And I have to mentally translate their
>pronunciation into the correct word meaning ("I'll cawell you" into "I'll
>call you", for example).  Pronunciation can vary greatly even in a small
>region.  For example, some of us Southern folk say "gree-its", but others
>(like me) say "grits".  How can a computer understand what someone is
>saying, when I can't even understand them?

Today, it can do fairly well when trained. You could understand them if you 
were trained. So can computers.

Right now, you have a tremendous advantage because you can anticipate 
meaning and fill in for unclear words from context. Good speech recognition 
will require a great deal of AI indeed; to recognise speech well, the 
program must essentially understand the world.

But in a narrow field, for computers to understand a range of commands is 
not so difficult.

>   I've got more "effective" MIPS,
>MOPS, TFLOPS, etc. and a better pattern recognizer than a computer.

Don't count on that lasting forever. Moore's law, after all. Sure, switch 
density will reach a limit, but then there are three dimensions, and the 
problem with AI now is not really computing power, but the programming problem.

But we do not have to be as smart as Mr. Bagget to understand speech. 
Idiots savant can do it, that is, recognize speech without understanding 
it. For a CAD system to understand its command set and the names of files, 
I would think, is already feasible. I can tell you, having carpal tunnel 
syndrome, largely from trying to be quick with both mouse and keyboard 
simultaneously, there have been many times I wished I could just tell my 
computer what to do.... I'd use the mouse to position the cursor and the 
voice to replace hotkeys. To do that, the computer would only have to 
recognize a very small word space. Filenames, however, would be more 
difficult, but because existing pathnames are by definition available to 
the system as a relatively small word space, it might not be as difficult 
as it would seem at first. It would not be necessary for a computer to 
recognize everyone's speech, just that if its user, and training could be 
made part of normal use, i.e., as one types hotkeys or filenames, one 
might, for a time, also say their names.






Abdulrahman Lomax
Easthampton, Massachusetts USA

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to