[Ultron]

https://wtvox.com/emerging-tech/the-first-exascale-computer/


I have always disliked comparisons of raw computer power with the power of the human mind, but what really bugs me on this topic is why are there not any graphene-based superconductor microprocessors out there yet? I have not read about any such experiments therefore I am not sure if any exist.


I realize the price of graphene is insane - I visited a lab early this year and they spend a lot of money and time just to make some low-purity graphene while they wouldn't even let us see the high-end material they had received from abroad. But what else other then the price is stopping these experiments?


What also makes me think a lot is, can we mimic intelligence (at least that of a dog or cat) with a regular personal computer? I personally believe we could even simulate human intelligence with very little processing power however with a specialized architecture - extra units to process sensory input (organic beings posses a lot of sensors trust me) and a high-end memory r/w system unique at design and purpose.

[ranch vermin]

Hey ultron havent seen you around in a while.   

You said something about computation not being the most important thing - as an exacomputer is only 100 times faster than a petascale, so its not getting very far very quick.

I think reversing ram. (cam hardware) is the way to go,  and is only 1000 times the size of ram, transistor wise.  (you multiply the ram by the keysize you want.)  so its completely doable now by even first upcomers to micro tech.

Otherwise your stuck with ram, and writing tree optimizations.  (what im doing now.)   

A cam is basicly a pattern matching setup, so it can basicly do what a neural net does in a basic linear like blocked out form - - and get the equivilent of what a deep net can do in 1 cycle!

[ranch vermin]

Here man, heres an excellent document on how to make computers, and its homegrown.

http://www.rle.mit.edu/media/currents/2-1.pdf

even a 5 micrometre transistor is still very good.

[Ultron]

I was away for quite a while... Trying to slowly crawl back into my A.I. research but I also have some exams to score on so I can qualify for studying robotics and machine intelligence next year. The thing is, the current exams are unrelated to the specific topic. But I'l be watching over you guys and hopefully finish my hardware platform for A.I. testing during winter (the cold helps me think better).


Anyhow - yes, I have stated on multiple occasions that I believe raw computer power / processor speed is barely significant towards building smarter machines and eventually A.I. I always say that tactic beats power therefore a better system architecture might help more then a juicier machine. Also I have stated that A.I. (somewhat less advanced then ourselves) could be programmed within a decent commercial computer - we simply lack the keys to the most basic functionality. These keys are somewhat more of a philosophical problem because they don't require more research or fancy algorithms yet simply a fresh and wider view on intelligence consciousness in general.


P.S. I live the PDF and I read it in more detail when I get the chance. I'd gladly discuss transistors and future alternatives/improvements when they stop taking over my whole desk.