New hope for Artificial Intelligence, The Memristor

http://blog.wired.com/gadgets/2008/04/scientists-prov.html

Here also

And here

And here

And google "memristor" for more links.

Researchers at HP Labs have built the first working prototypes of an important new electronic component that may lead to instant-on PCs as well as analog computers that process information the way the human brain does.

The technology could even change the way computers function. HP believes that memristor circuits will eventually power machines that mimic the human brain.

Such machines could learn from previous experience and retain information for future use in associating events and recognising patterns.

Read this!!

"Building an analog computer in which you don't use 1s and 0s and instead use essentially all shades of gray in between is one of the things we're already working on," says Williams. These computers could do the types of things that digital computers aren't very good at –- like making decisions, determining that one thing is larger than another, or even learning.

While a lot of researchers are currently trying to write a computer code that simulates brain function on a standard machine, they have to use huge machines with enormous processing power to simulate only tiny portions of the brain.

Williams and his team say they can now take a different approach: "Instead of writing a computer program to simulate a brain or simulate some brain function, we're actually looking to build some hardware based upon memristors that emulates brain-like functions," says Williams.

Such hardware could be used to improve things like facial recognition technology, and enable an appliance to essentially learn from experience, he says. In principle, this should also be thousands or millions of times more efficient than running a program on a digital computer.

Basically this new circuit will make possible to build a computer which can work like the human brain, then we can create artificial intelligence.

I read that in around 3 years they will make flash memories and SSD from these memristors and in around 10-15 years depending on the market they will make processors based on memristors ,processors which will be able to work like the human brain.

The computing speed of processors will not skyrocket, it's the different way in which this computing takes place with memristors, it's made in more than just 0 and 1 sequences and that's why the memristors are so interesting.

Memristors will be the base for the non-linear computing technology, and this non-linear computing technology will be the base for building androids.

They can build processors based on memristors right now but it's up to the circuit design teams to start working with them.

I hope these processors based on memristors will hit the market while we are still young

ooo, I like this... the problem in the past to current day is that it was hard to make values in-between 0 and 1. But if this is correct, the given values could be any given number between 0 and 1.

The main reason this method was dropped long ago was that a "current" (volt) for a 010101 sequence looks like this... __--__--__-- <-- that dosen't look great, but basically there is a clear difference. Trying to place sub-values was hard and mistakes were made almost every time.

So yea, i'll be keeping an eye on this new development.

MattDragonTamer wrote:

The main reason this method was dropped long ago was that a "current" (volt) for a 010101 sequence looks like this... __--__--__-- <-- that dosen't look great, but basically there is a clear difference. Trying to place sub-values was hard and mistakes were made almost every time.

Yes soon processors will not be formed only from switches( transistors) which all they do is to switch the light on and off or 0 and 1 ,they will be able to switch between more values making possible to compute a lot more complex programs.

The memristor is not ready yet for that kind of purpose it may probably come to reprogram software al over. Programmers struggle with parrallel coding. The point is the problem to a.i is plasticity its the changing factor that makes a data result different every day how do you track that a memory manager for a memristor would change state so fast that the communication between the a.i software and the memristor storage shall reformat the data the a.i has builded up. Its different with a harddrive storage you save it retrieve it put it back. Memristors probably have to map a structure of somekind. Maybe if and if its good enough for somekind of dynamic variable storage it can support any form of a.i program.

Kal-el wrote:
Yes soon processors will not be formed only from switches( transistors) which all they do is to switch the light on and off or 0 and 1 ,they will be able to switch between more values making possible to compute a lot more complex programs.

Well, what I was trying to say was that (if my years in the computer science program in University serves me right) the reason that the "grey" values between 0 and 1 were dropped a while back is that the information stored as "grey" values were almost always lost. They would usually default or be misread as either a 0 or 1 value.

Computer scientists have tried to toy with "grey" methods as an alternative for the binary 0's and 1's for many years and have failed over and over. It’s just too unreliable. But there have been some advances with solid state hard drives and their storage methods using “grey” values. Using very sensitive multi-translating “switches”, they can store “grey” values properly. But again, very unreliable, a slight change in power levels and all the data erases or gets modified. Not to mention the extra processing information needed for every level of “grey” values used.


So I know it’s a great area to explore, but I really can’t see it being used for a long time yet. It’s more of a research project subject the anything else at this point. The 0 and 1 method used today has been by far the most stable and reliable way of storing data and transferring information. We are just starting to hit big changes to the processing methods with AMD’s “CUBE” (a processor with a core system comparable to 16 quad cores. (That’s 64 single cores running at 2.6GHz) Their only problem ATM is to try and keep it cool…lol… even with the chilling space like environment.

But anywho, if they do find a way to make it as reliable as the current binary system, then I’m all in with it.

Yes I could agree on some of your saying but 0 and 1 are not the only concept any more but as you said the boolean system is stable, But knowledge this to that Fuzzy Logic is also a proved system which is capable of switching to a degree and there is also the research into quantum computers qubit which is a state of a small constellation.

The way to see a memristor internals is like a cube with a charge and air the state of charge from this cube is defined by the oxidation the more air the dense the charge is. And if the air compressed the lesser the charge the variance between more or less is what the memristor remembers because once it its set it does not change.
In case my explanation does not help.

Ps here is the link to a good article enjoy

srry here the link www.spectrum.ieee.org/dec08/7024/2

MattDragonTamer wrote:
We are just starting to hit big changes to the processing methods with AMD’s “CUBE” (a processor with a core system comparable to 16 quad cores. (That’s 64 single cores running at 2.6GHz) Their only problem ATM is to try and keep it cool…lol… even with the chilling space like environment.

Indeed it will take a while until cip designers start using memristors.

The Amd Cube seems interesting, can you give me a good link about it?

wulf wrote:Yes I could agree on some of your saying but 0 and 1 are not the only concept any more but as you said the boolean system is stable, But knowledge this to that Fuzzy Logic is also a proved system which is capable of switching to a degree and there is also the research into quantum computers qubit which is a state of a small constellation.

Yes quantum computers look promising also, if they indeed become available they will shatter Moore's law which stipulates that at every 1.5 years the processing power of processors doubles.

Nice link about memristors.

Kal-el wrote:
The Amd Cube seems interesting, can you give me a good link about it?

I have been seaching for a while and can't find the post anymore... all I keep getting is
some sort of Digital Cube company info related to AMD...

I'll keep looking.

HI

For some reasons, I find AMD is a bit slower, but the price is a lot cheaper.
What I really need is a cool cpu and fast, but doesn't cost a million dollars.

I think intel C i7, is a good start.
I read a lot of reveiw for it... it seems to be cooler than other cpu.

Perhaps Fuzzy Logic can be applied to Aiko in a sense that she can use it as a base for feelings. Having her own mind, I am not sure how to go about it at this point. I believe just like there are constant chemical changes in the brain that alter a persons mood, feelings, and descision making perhaps something can be applied using fuzzy logic. This is because humans do not make only True and False descisions but grey True and grey False. When you ask someone how do you feel you expect yes or no but sometimes you can get "I don't know" or "ok, kinda".

I don't see this being used as a base for everything in the BRAINS software because reliably the core would still have to be 1's and 0's to have any meaning to the rest of Aiko's software. Therefore I believe that there can always be one section or subroutine for Aiko running in the background controlling artifical chemical imbalances that cause Aiko to change her mood, or though process a little bit. Just like a real person.

AMD can be bit slower in RAW processing power,, but will stand, and excels in using much less power, and creating ( much) less heat then Intel
That's why they are so popular in the professional world.
And, I refuse to work with Intel, for obvious reasons, only have to admit their Pentium Pro 200 with the big cash was a very good product.
Have seen a P4 going way down contra the first Athlons, in speeds and processing power, while busspeed from the P4 was higher.

Further , wouldn't it work better to start all over under 64 bit OS ? And, in a multiprocessor environment, you could split up processes over one or more cores. That is, if the OS let's you arranged that dynamically.
Scalability of the processingpower, will further lower the needed power, and heat.
Why not using watercooling, to cool the processors, and maybe use that, to give heat to the 'skin' ?

Read an article in NATURE date 8 April 2010 headlined on page 873 ''Memristive switches enable 'stateful' logic operations via material implication''
While reading it I saw possibility that could turn around better way's to stabilize A.I research by hardware,

Memristors as you might remember are a new kind of devices that offer memory and logic functions (not know at that time yet as of now does),
One of the main characterization is that it holds more then one state, defining the resistance(A state of pressure) history upon the voltage history (which its remembers). Its like a memory block becomes a train, Traverse by walking through each wagon on the train finding another memory waiting to get out the train. while the other return to its seat letting the other out till another passenger block replace its seat. everyone stays put.

I think this new memory could deliver other way's support building application towards A.I. Because of how it can form logic functions as mean of processing. For example what if you have an A.i system that learns a pattern stores it on a drive and retrieves the same data putting it in memory transfer it partially in the local cache while making comparison of the pattern by the state it learned how to do something.

This process falls into the same system execution space the application used to build up the pattern meaning the program pulls these pattern or builds them all together. What if you had a propitiatory part where it could put the pattern knowing the pattern will logically execute itself. The main program remains control but still adjust the output of the total pattern state. (''does this still makes sense''). The a.i system gets more versatility where it does not need to compare all outputs only so it does less by adding more matches to fit the situation.

Summary.

The data becomes the program while the programs builds up other data or task remaining control of how much of that data logic function can do something.