PDA

View Full Version : Memristors: why deus ex technology will soon be possible



Ghostface
3rd Dec 2008, 02:31
'Missing link' memristor created: Rewrite the textbooks? (http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=207403521&printable=true)





Also here's a site I found with some technical information, I'm not sure where the origional source is from.

The intormation on this site is informative, although i must warn you that the site badly designed and links can be malfunctional or link you to NSFW stuff.

here's the site for anyone interested in technical discussions related to memristors

perhaps this information should be copied and pasted onto a more secure site

http://groups.google.com/group/memristor-computer-programming

GmanPro
3rd Dec 2008, 03:21
Very interesting. I wonder what the implications really are. What kinds of technology are now going to be unlocked with this discovery?

Ghostface
3rd Dec 2008, 03:44
Very interesting. I wonder what the implications really are. What kinds of technology are now going to be unlocked with this discovery?

Here's a youtube vid of memristor technology being explained at HP labs

http://www.youtube.com/watch?v=kUOekeiqihc

http://www.youtube.com/watch?v=wZAHG3COYYA&feature=related

K^2
3rd Dec 2008, 06:05
'Missing link' memristor created: Rewrite the textbooks? (http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=207403521&printable=true)
That's a very bad article. The whole "Rewrite the textbooks?" part is just a poor attempt at sensationalism.

Take this for example:

"Electronic theorists have been using the wrong pair of variables all these years--voltage and charge. The missing part of electronic theory was that the fundamental pair of variables is flux and charge," said Chua. "The situation is analogous to what is called "Aristotle's Law of Motion, which was wrong, because he said that force must be proportional to velocity. That misled people for 2000 years until Newton came along and pointed out that Aristotle was using the wrong variables. Newton said that force is proportional to acceleration--the change in velocity. This is exactly the situation with electronic circuit theory today. All electronic textbooks have been teaching using the wrong variables--voltage and charge--explaining away inaccuracies as anomalies. What they should have been teaching is the relationship between changes in voltage, or flux, and charge."
The reason that electronics deals with voltage is because it is the difference of electric potentials at two points, and the gradient in the potential is what drives the current through a conductor. Therefore, dynamics of every circuit can be described by deducing the voltage across each element as some function of other parameters. The only exception is due to Hall Effect. All of this is based directly on Maxwell's laws.

Memristors are no exception. Their V/I diagram show hysteresis. But that tells you nothing. Under right conditions, a capacitor V/I diagram will show hysteresis. In fact, if sinusoidally driven, V/I graph for capacitor is a circle. All this tells you is that you shouldn't be looking for V(I). It should be a function of some other parameters. In case of meristors, there is an internal parameter.

Finally, on the search for the 4th element. I don't see what the memristors have to do with it. In impedance representation, the Resistor, Capacitor, and Inductor do in fact form a set that would be complete with a 4th element. However, that element would have to have current flow in the opposite direction to the applied voltage. Memristors still allow current flow in direction of applied potential difference, so they are obviously not that 4th element.

All that aside, memristors are an interesting idea. What I'd like to know is how susceptible they are to bit rot.

GmanPro
3rd Dec 2008, 06:35
Send HP an email I guess lol... :D

K^2
3rd Dec 2008, 08:44
Guy they interviewed, one who said all that nonsense, doesn't actually work for HP. I'm sure the guys who actually built memristors know what they were doing.

user-9
3rd Dec 2008, 12:10
Personally I believe that they've already got alot of this stuff and much more beyond the realms of our imagination behind the closed doors of the military industrial complex. Give you a couple of examples that you can research for yourself...

John Lear, the son of the Lear Jet inventor. John Lear is in the Guiness book of records for numerous flying records, flew for the CIA and is a UFO KNOWLEDGE. He claims that the military already have planes that can go anywhere in the world in 3.5 hours - Have a look at him on youtube.

The Disclosure Project. An organisation of ex-military whistle-blowers who want to expose what they've witnessed with regards to the UFO/EBE cover-up. I watched the D Project film of the press conference at the national press club. There was a guy on there claiming that they have a device that they can pass over you (handheld) that can fix your health - like in Star Trek.

Not quite nano augs but imagine what else they're up to... :cool:

Ghostface
3rd Dec 2008, 16:46
That's a very bad article. The whole "Rewrite the textbooks?" part is just a poor attempt at sensationalism.
..

Finally, on the search for the 4th element. I don't see what the memristors have to do with it. In impedance representation, the Resistor, Capacitor, and Inductor do in fact form a set that would be complete with a 4th element. However, that element would have to have current flow in the opposite direction to the applied voltage. Memristors still allow current flow in direction of applied potential difference, so they are obviously not that 4th element.

All that aside, memristors are an interesting idea. What I'd like to know is how susceptible they are to bit rot.

I'll have to comment later when i have more time to have an in-depth conversation with you, but that was a good post.


Guy they interviewed, one who said all that nonsense, doesn't actually work for HP. I'm sure the guys who actually built memristors know what they were doing.

Did you see the 2 minute video? Or the 16 minute one? Because they're two different people talking and one of them does actually work for HP.


Philip J. Kuekes, computer architect on quantum science research team at HP Labs shows me the lab where they discovered Memristors.


Here's a link to the original article from wired on this new discovery.

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


Indeed, Chua's original idea was that the resistance of a memristor would depend upon how much charge has gone through the device. In other words, you can flow the charge in one direction and the resistance will increase. If you push the charge in the opposite direction it will decrease. Put simply, the resistance of the devices at any point in time is a function of history of the device –- or how much charge went through it either forwards or backwards. That simple idea, now that it has been proven, will have profound effect on computing and computer science.



Indeed, Leon himself noted the similarity between his own predictions of the properties for a memristor and what was then known about synapses in the brain. One of his suggestions was that you could perhaps do some type of neuronal computing using memristors. HP Labs thinks that's actually a very good idea.

"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.

Ok here is a post I found on the wired site in the original article, hopefully you can give me your input on this.


In response to "Charl", there definitely is reason to calling it a "basic" element. You have the three "old" ones:
v=Ri
i=Cv'
v=Li'
and a "new" one:
v'=Mi'
When M is a constant, the "new" one acts exactly like a resistor. The "memristor" action really occurs when M is not a constant (e.g., when M depends on i'). That gives the device "hysteresis" in the i-v curve. Nearly any device that shows hysteresis in the i-v curve can be thought of as a kind of "memrestive" element.
Remember that springs and masses are not good analogs for capacitors and inductors, respectively. However, a spring-mass system is a good analogy for an LC circuit, where you can think of the inertia (the mass) as inductance and the spring stiffness as the capacitance.
Mechanical analogies for LC circuits work when you equate (for example) speed with voltage and force with current. This analogy works because voltage*current = power and speed*force = power. So, gear trains are impedance transformations, and transformers are gear trains.
If you need an analogy for a memristor, you need to think about *CHANGE* in speed being related to *CHANGE* in torque in such a way that depends on *CHANGE* in speed (or torque) and not absolute speed or torque. Any MECHANICAL system that displays HYSTERESIS (there are TONS) work this way.

Also are you an electronics engineer? I'd be interested to understand how this development would effect both programming and design as well as operating systems for computers. There would be a lot of advantaged to building this type of analog system from what I hear, especially in brain-computer interfaces and artificial neural-nets. From what I've read, this new device will allow for more adaptable design. If one is programming with a general purpose computer.


Conventionally electronic systems may broadly be defined into two distinct classes - hardware systems and software systems. Hardware-based electronic systems are formed using specialized circuitry and are typically faster than software-based electronics but lack adaptability. In contrast, software-based systems, which are based on programs run on a general purpose microprocessor, are more adaptable but lack the speed of specialized hardware.

Here's a link to the quote above: http://groups.google.com/group/memristor-computer-programming/web/memristors-and-artificial-intelligence

K^2
3rd Dec 2008, 18:05
Ok, the V'=MI' way of putting it makes more sense with what they are saying, but it still isn't a very good way of describing what is happening. First of all, V'=MI' is an extremely special case of voltage hysteresis. Secondly, V'=MI' does not imply that V is a function of I at all. It is a functional of it, yes, but not a function. So saying that R depends on I' is silly. Note also that in a sinusoidally driven circuit, V'=MI' is no different than V=RI, with M=R. The impedance is exactly the same.

As I think I said, the idea is interesting. And the fact that the V'=MI' might have pointed out the idea is great as well. But it in no way implies that we should think of the circuits differently. The existing tools work because they reflect the physics of the carriers. A new component is not going to change it. You'd need new physics all together to warrant that. There is no fundamentally new physics here. Just a new way of putting things together.

Again, I'm not trying to undermine the engineering importance of such element. I think a lot of great things will come from it. And it might change how we view computer architecture, but not how we view electronic circuits in general.

Also are you an electronics engineer?
I'm a high energy physicist. I occasionally teach electronics lab, and I like to tinker with the stuff, but I have no engineering education.

Psychopomp
3rd Dec 2008, 19:12
Being the type of person who prefers Philosophy to SCIENCE(!), MY BRAIN ******* HURTS.

K^2
3rd Dec 2008, 21:02
Science is a branch of Philosophy.

GmanPro
4th Dec 2008, 00:55
^^
Well said. :D

Ghostface
5th Dec 2008, 20:32
I'm a high energy physicist. I occasionally teach electronics lab, and I like to tinker with the stuff, but I have no engineering education.

Your point makes sense , I'm quite busy so i cant contribute much right now. Good to have someone knowledgeable in tech on this board. I guess you'll be interested in the new development about using the force of light to move nanomachines.

http://www.azonano.com/news.asp?newsID=8906

One of my biggest interests are in the cognitive sciences. Including brain implants and more human interaction with technology.

K^2
5th Dec 2008, 23:22
See, this is why I don't like these popular media news articles on that kind of stuff. They over-dramatize everything. "Force of light" has been used for various purposes for a long time. Laser tweezers, laser cooling, and many other methods.

The machines that these guys built is very interesting, but it isn't interesting because of the "force of light". It is interesting because it would allow direct interface between a photonic processor and a nanomachine, which would be extremely useful in nanorobotics.

GmanPro
6th Dec 2008, 00:27
They have to over-dramatize the article, or else it wouldn't appeal to the casual reader. :rasp:

v.dog
6th Dec 2008, 20:34
Even worse are the articles about 'water powered cars'.:mad2:

For a memresistor to work like a synapse, it would have to hold a preference for a resistance over time. That's doable, I see several issues, however; the stored states would need to be at levels discrete enough not to be mistaken for noise in the system.

Then there's questions over how long would it take to fix a state, and whether it's truly permanent. I'm thinking in terms of backing up, restoring, and wiping a memresistor/synaptic array.

K^2
6th Dec 2008, 21:13
Well, DRAM can only hold data for a few seconds, because each bit is essentially a capacitor attached to a base of a MOS FET. So the data in DRM memory has to be refreshed by the memory controller. But as long as the required refresh time is much, much greater than read/write times, it is not that big of a deal. So even if memristors can only hold data for a few seconds, it is perfectly fine, since the neural network needs to operate on time scales of a few ms.

You are right about the noise, but I figure if they are planning to build memory out of this stuff, they have that issue covered, or at least are confident enough that they can get it resolved.

Ghostface
6th Dec 2008, 22:17
See, this is why I don't like these popular media news articles on that kind of stuff. They over-dramatize everything. "

lol you also have to think about it from a business perspective. If you create anything you would hope they do the same for you, or you'd have a hard time finding investors. Most businessmen aren't as visionary as these scientists, therefore things have to be simplified somewhat.

I wish I knew more about the implications of this development. I'm curious if colleges/universities will be quick to integrate these new developments into their programs. One of the problems with this chip is that most people in the field don't know how to use it to it's full potential.

Speaking of synapses; the main advantage that the brain has over a computer is parallel processing. Computers today have their limits when it comes to that.

This will also make it easier to create BCI's since translating brain signals to the chip will be much easier, especially if its so adaptable. Adaptability is necessary since everyone's brain is different. You'd also be able to have more bang for your buck, since chips of the same size would be more powerful, since they don't need so many transistors. They're even saying its simpler then MOSFET transistors.

For those of you who know about neurology, would it be possible to create a pain suppressor chip by causing electrical interference in specific parts of the brain before pain is registered?

K^2
6th Dec 2008, 23:05
I don't know if you could distinguish between pain-carrying pulses and info-carrying pulses in the brain. Hence most of the modern pain killers slowing down brain function. You can also pump adrenaline to compensate, but that's how you get berserkers. Would be interesting to see if something better can be done with a signal processor built into the brain.

Universities tend to pick up that stuff pretty well, because there are always some grad students that want to play with new stuff. So the profs tend to get involved, and if it is something useful, it tends to make it into curriculum pretty fast.

Ghostface
7th Dec 2008, 00:25
I don't know if you could distinguish between pain-carrying pulses and info-carrying pulses in the brain. Hence most of the modern pain killers slowing down brain function. You can also pump adrenaline to compensate, but that's how you get berserkers. Would be interesting to see if something better can be done with a signal processor built into the brain.

Universities tend to pick up that stuff pretty well, because there are always some grad students that want to play with new stuff. So the profs tend to get involved, and if it is something useful, it tends to make it into curriculum pretty fast.

Good that things get picked up quickly. I'd like to learn more about this.

As for signal processing. I think if you have information about what's going on in the brain it can be done. If the chip is adaptable, it can be made to improved based on the continued stream of information.

Not sure if you've seen my earlier post but here's a vid I posted in the other thread.

http://www.youtube.com/watch?v=Pgmoz4f8LA4

K^2
7th Dec 2008, 01:18
The problem with processing brain signals is that the brain is like a learning compression algorithm, with algorithm being part of the data that is being compressed. If you want a chip that will be capable of sorting through all that, things that are going through the chip will have to become part of consciousness. You won't have the chip controlling the brain anymore, but rather becoming part of it. And rather than control, what you'll really have is better regulation.

What might be slightly more realistic, for starters, is a chip that gets rid of phantom pains after limb amputation. It can supply the brain with stimuli that resemble normal signals from the limb, perhaps, based on sensory input from a prosthetic limb. Studying how these things work will be a big step towards learning to control other impulses in the brain.

Ghostface
7th Dec 2008, 04:20
The problem with processing brain signals is that the brain is like a learning compression algorithm, with algorithm being part of the data that is being compressed. If you want a chip that will be capable of sorting through all that, things that are going through the chip will have to become part of consciousness. You won't have the chip controlling the brain anymore, but rather becoming part of it. And rather than control, what you'll really have is better regulation.

What might be slightly more realistic, for starters, is a chip that gets rid of phantom pains after limb amputation. It can supply the brain with stimuli that resemble normal signals from the limb, perhaps, based on sensory input from a prosthetic limb. Studying how these things work will be a big step towards learning to control other impulses in the brain.

guess you gotta crawl before you walk. If a pain suppressor chip turns out to be too difficult that might be a better place to start. I do know that Parkinson's and depression have been successfully cured using brain implants, so the concept isn't impossible. Implants have also helped people with Alzheimer's. What would also help would be better tools to navigate the delicate areas of the brian, perhaps a robot controlled by the surgeon that mimics his/her hand movements.

By the way what did you think of that youtube link with the nanowire experiments??