Ralph Daugherty wrote:

>     As a PC assembler programmer veteran of the 80's (Z-Soft PC Paintbrush and

>   Melita Electronics call processing software were the two products I worked

> on), excellent algorithms were a must to do what we had to do with the CPU's.

>   But the mantra became speed of programming, not speed of execution, hardware

> will get faster.  And it did, and along with it a completely different set of

> coding principles that negated much of the improvement in hardware.

> 

>     Along that line, has any demonstratable AI software been written in the

> object oriented coding style that is superior to prior coding styles or

> languages?  OO was mentioned at least as frequently as faster hardware as the

> key to breakthroughs in AI.  It was funny that on the /. AI MM thread, SHRDLU

> was still mentioned as the example of great reasoning software, and the

> statement made that nothing better has been seen since.  I found that hard to

> believe, whether true or not.  It is a great achievement, but I read about

> that it seems like decades ago.



Indeed. I agree (even tho myself I have not gotten accross to realizing

the importance of simplicity when I was programing, rather when I was

trying to make something basic work on a 'out of the bin' computer).



As I said most modern software is not really optimized for speed, if AI

programs were to run perfectly optimized in something like ASM, on a

processor that isn't optimized for an average C++ program (Pentium;

older co-processor based systems preform much much better in real

calculative work than Pentiums, yet average desktop utilities will still

work best on a Pentium), the progress on devoloping AI would be a little

faster.



Interestingly we have heard of AI programs all the way to as unoptimal

as Java. You can actualy do archive-searches to see philosophers claim

things that are the solid basis of actual CPU instructions, as

impossible, because they are unsuported in modern high-level programming

languages.



The old knowledge has been given a second chance when parallel

processing was in it's infancy. The unique solutions to running a

simulation or AI thought process over multiple processors had used

actual CPU instructions that were intended to make the maximum use of

the cluster's capabilities. Still surviving projects form that time

appear to be quite rare.



I have heard of some attempts to make neural networks work via ASM,

allas the projects appear to have ended where the GUIs began. How sad.



I stand my ground however, the principle needs to be worked on before a

project is set into code. One must realize exactly how full AI will be

reached, before starting to code. I would say "look everybody I have a

solution right here", but I already know nobody is really interested.

Those that have financial backing for their AI project should consider

themselves lucky...



			Observer aka DustWolf aka CyberLegend aka Jure Sah



C'ya!



--

Cellphone: +38640809676 (SMS enabled)



Don't feel bad about asking/telling me anything, I will always gladly

reply.



Trst je naš, Dunaja ne damo; Solmuna pa tud ne. Za vstop v EU. ;]



The future of AI is in technology integration,

we have prepared everything for you:

http://www.aimetasearch.com/ici/index.htm



MesonAI -- If nobody else wants to do it, why shouldn't we?(TM)