> ekurtz99{at}WhoKnowsWhere.com wrote 
>>The fundamental mistake here is the implcit assumption that advances
>>in machine understanding of language are problems of storage and cpu
>>speed; they are not; if that's all they were, machines would be able
>>to converse at speeds slower than a human but in a realistic way.
>>Obviously they cannot.
> 
William Morse wrote:
> I realize that advances in language are more than problems of storage and 
> cpu speed. In particular, the classic computer architecture is well 
> suited to logic problems but ill suited to the kind of processing 
> involved in language. But in any case your statement is incorrect. Ants 
> cannot converse at speeds slower than a human - they cannot converse at 
> all. Even chimps can only converse haltingly after much specific 
> training. 

In the case of ants the reason is obvious; but it may that the chimp has 
enough brain matter to support language ability; but it doesn't have the 
necessary brain organization. Language is not learned in the way that 
algebra is learned; we are born with the capacity to acquire it from our 
social group.

It is likely that it takes a very high degree of complexity to
> begin to converse - and current computers are nowhere close to that 
> level. 

You are once again confusing hardware with software complexity. Faster 
machines aren't necessarily more complex. Do you have a background in CS?

> So it is too early to tell if advances in cpu speed and storage 
> will eventually allow computer speech even with the current limited 
> understanding of how language is constituted.

This doesn't follow at all; suppose I believe I have figured out how to 
implement conversational ability in software. I implement my design in a 
program; I then run the program by asking it sensible real-world 
questions; if the machine that is running the software is slow compared 
with the ideal for my system it may take (say) a month to respond, but 
the response will be no different from what I would have got on a much 
faster system, say in seconds; I can then resond to the response, wait 
another month etc etc. Conversation does not have to be real time. As 
long as the responses indicate that the system understood the questions 
and was able to construct human-like answers, I can claim to have 
cracked the problem.

If you cannot see the point I am making, then you do not understand this 
issue.

>>It is one of the conceits of the AI profession, or at least of its lay
>>apologists, that Moore's law will solve every current AI problem. It 
>>won't. There is the small matter of the software...
> 
> 
> But the software is also evolving, albeit at a much slower rate than the 
> hardware.

This is the Discovery Channel version of AI history.

>  
> Yours,
> 
> Bill Morse
>
---
þ RIMEGate(tm)/RGXPost V1.14 at BBSWORLD * Info{at}bbsworld.com

---
 * RIMEGate(tm)V10.2áÿ* RelayNet(tm) NNTP Gateway * MoonDog BBS
 * RgateImp.MoonDog.BBS at 7/22/04 12:20:01 PM