R. Stanley Williams from HP Labs gives a keynote presentation on memristor technology at the UC San Diego Center for Networked System's Winter Research Review 2010.
Of course, Memristors are likely to be affected by what's known at HP as the Inverse Moore's Law, which means that every two years, it will take yet an additional two more years for an actual working prototype to appear. - j q t - (former HP employee : - )
@joquarky zoomed out I don't think the slide would be as readable. There was always plenty of time to see the slides, and if you need more time to look at something, there's always pause/screen caps.
There is many software (computer scientists) that are working on parallel computation and many other unusual computational architectures that are not in the public eye. What may happen with this technology is that it will give a new generation of people coming into the industry a chance and the people who are not willing to change will be swept aside. Programmers of the future will need to be physics based computational engineers so as to get the best out of technological systems.
If you can have a hybrid system that basically can switch between storage and logic on the fly you can then in theory simply increase the number of "cpu's" as you need them in exchange for some free memory. And if you get low on memory you can automatically convert the CPU back thus sacrificing performance but keeping the system running. I like this :)
@2oonhed Hehe, yeah I was a little bit quick there. I felt I had to write it before he started to talk about the real history. At about 01.11, it didn't look like he was going to cred anyone but the in house staff, my mistake. Wchounderful piece of info this clip brings! ;) I really enjoyed it!
@pythro Yeah that was one of the first things they tried: making an memristor version of an FPGA. The problem with FPGAs is that you to devote extra transistors/space to make it programmable, so you have less transistors to do the actual work, making them slower than more specialized processors. With memristors you might not have to make that huge sacrifice.
@ortcloud99 Sorry, the name "Cyberdyne" has already been scooped up by a japanese company that makes Hybrid-Assistive-Limb technology(... or HAL for short :-) ). It's a sort of exeskeleton robo-suit that is intended for disabled people, search and rescue type operations and heavy manual labour in factories where you'd want a machine to do most of the heavy lifting but you need the brains, flexibility and agility of a person.
@drakulva Yeah, about that: Leon Chua, who came up with the theory of memristors, is from the Philippines and part of the Chinese minority in that country.
I honestly expect about a 30 year development time for memristor tech... We're what, almost 10 years in... I ABSOLUTELY believe it's coming. I just think the tech will take more time to develop.
@utubesqueeze "Everything they know is about to be changed but its the electrical engineers who will get all the glory!" That's just giving the glory to the people due to receive it!
I'm a bit late to the party here. Since, this video is 8 years old, and Mr. Williams presented a 10 year timeline. Are memristor chips available today in 2018?
@unskeptable I´m quite relaxed, listing to what the presenters of the memristor says, they keep saying "jit reacts ust like a neuron", so if you build a digital copy of the human brain with memristors, and insert an self learning program. What does it take for the digital brain to become self aware? Just take a simple example, the Boston Dynamics Mule. When it walked over some ice and self corrected the engineers was surprised how life like it responded.
@KawazoeJapan The biggest obstacle currently is the limited lifetime of memristors. At best these devices last for a few million cycles, until the TiO2 degrades so much that there is no hysteresis anymore.
memristor is not passive device. memristor requires additional energy in order to change resistance. (due to free energy barrier between memory states)
Wow. They can do Boolean logic now? I'm no expert, but I think that means you can theoretically create a one-chip-does-all chip. Instead of having bunches of different hardware designs with different strengths and weaknesses, you could just program a memristor chip to behave like any current transistor-based chip, and it would still be extremely efficient. It would be like having a hammer that could transform into a wrench, screwdriver, level, etc. whenever needed, and work just as well.
R. Stanley, you invented Skynet are responsible for the war between humans and AI. Having memristor located near a CPU, and having the CPU transistors replaced with memristors and with infinite bandwidth. We're talking AI!
"winter research review 2010" threw me off b/c I still think of 'winter' as synonymous with 'december' but really the middle is in early february, isn't it.
Memristor is too complicated to be the missing linear circuit element. The "slactor" is more likely a better fit. A "resistor" converts electrical energy into "heat" and dissipates this energy to the outside, it's a device built on "irreversible" phenomena. A "slactor" converts electrical energy into "heat" and stores this energy, it's a device built on "reversible" *ThermoElectric* phenomena. So, a "slactor" is like a "capacitor" or "inductor", the latter two storing "electric" and "magnetic".
@martinjpedersen Yes it seems very realistic but self awareness is science fiction .Memristor is just a charge-dependent resistance dude we cannot build life with it lol 0.0
What is an ironical contradiction in your brilliant speech is that you first refer to the NOT operator as essential (Betrand Russel), then you try to explain it away... The NOT operator is essential!
Memristor is just a kind of resistor with memory rather than new fourth passive element. The relationship between charge and flux represents the resistor. The resistors include linear resistor, nonlinear resistor and memristor. The memristor is memory resistor.
Pandoras box have been open, and cant be closed again. This have both positive and negative potential. On the negative side, soon we will have a real AI, among us..
The memristor is supposed to be the fourth passive deviced as theorized by Chua in 1971. It's completely passive. i.stack.imgur.com/3PvGn.png Resistance is a connection between voltage and current. Capacitance between voltage and charge Inductor between current and magnetic flux One element that connects charge and magnetic flux is missing, if it exists, and that would be the memristor (see picture).However since a few years people have been calling resistive memories "memristors", but that's simply not correct...
+ImaginaryHuman072889 ok, why do you facepalm? provide a physics and electrical-engineering based explanation of why you think memristors will not scale well on a CMOS process.
+jrcenina85 i've never heard of "almost unlimited" being described as an idiom, but even if it were an idiom, why would he use such a meaningless expression to describe how many memristors could fit on a chip? i guess i could understand him saying this if it were a non-technical audience, but seeing as the intended audience is clearly mostly electrical and/or computer engineers, this seems silly. it would be like an english teacher using "aint" during lecture.
+ImaginaryHuman072889 the existence of the specific idiom doesn't change anything. Our language allows for non-literals. Whether that's a good thing or not is arguable. But if you hold someone to something they literally say then the way language operates goes over your head. And an English teacher using "ain't" doesn't bug me. If anything it shows they are confident enough in their command of the language to know that using "aint" doesn't define them.