Behind the scenes at Labs
Showing results for 
Search instead for 
Did you mean: 

Your Device Isn’t As Smart As You Think It Is



1635 x 313 v2_abbreviated_edit.jpg


Professor Leon Chua explains the problems with flash storage and chip intelligence in the second installment of his 12-part lecture series, From Memristors and Cellular Nonlinear Networks to the Edge of Chaos. You can watch the replay of his second lecture, but here are some highlights from his talk about “Everything You Wish to Know about Memristors but are Afraid to Ask”.


Our world runs on a data storage system that is unsustainable for current battery technology, and doesn’t guarantee that our data will still be there in ten or twenty years. 


Flash storage may seem like an advanced technology to us now. It’s used to store information in almost every aspect of computing, from phones and tablets to laptops and servers. However, this technology may not be as dependable as people think. Contrary to popular belief, flash storage is isn’t actually very good permanent storage. Flash memory loses about 20% of its information per decade. “That’s why you shouldn’t store your family photos on a flash drive,” Chua explains.


Professor Chua points out some other inherent problems with the current model, including how flash can’t be scaled below 20 nanometers without causing problems. With flash, there is a limit to how small memory storage can get. Flash also fails fairly easily after switching, sometimes in as little as 10,000 rewrites of its memory. “Flash memory is the reason you have to charge your iPhone every night,” Chua says, explaining that the constant flash storage retrieval we use to run apps and other data on our phones puts enormous pressure on batteries to keep up. One fundamental cause of this energy inefficiency comes from the lack of endurance seen in the current technology behind flash operations, the floating gate transistor.




So what can replace it?


Chua’s research hypothesizes that the future of information storage lies in a new kind of circuit element called a Memristor. The Memristor creates a unique kind of circuit, one that Chua proved can’t be built using the three fundamental circuit elements: the resistor, inductor and capacitor. A Memristor can be scaled down to just one nanometer, which is the size of one sugar molecule. This means that memory storage on Memristors can be much more densely packed than flash storage, but that’s only one potential use. Memristors may also change the way information is processed, as they function more similarly to the human brain than any other kind of circuit.


There is new evidence that Memristors react to stimuli in a similar way to living things. A recent study in China used Memristors to emulate the stimulus/response dynamic famously measured by Pavlov’s dog experiment. They found that in very simple circuits, Memristors change their responses to stimuli over time, responding to positive elements and learning to stop responding to a useless stimulus. Studies like this suggest that Memristors show signs of true learning, following the psychological model of habituation. Learning to recognize and ignore boring stimuli is a basic but fundamental behavior, and it is exhibited by living creatures as simple as sea anemone or sea snails. While a computer program could artificially simulate these behaviors, it would be a lot slower and require many times more resources.


“There is no such thing as a smart device right now,” says Chua, claiming that conventional memory processes in today’s technology don’t exhibit true signs of intelligence. Chua believes that future computers will use Memristor technology to be more brain-like in their design. Chua claims there is nothing in chip technology that will rival the Memristor “for the next hundred years.” Memristors were turned into solid state memory by Stan Williams, HP Senior Fellow at HP Labs, after realizing their potential in computing, but there’s still a long way to go. While the idea of a computer chip having true intelligence sounds farfetched, Chua compares the phenomenon to how an amoeba can respond to temperature changes even though it is not an advanced creature, saying, “an amoeba doesn’t have a brain, but it has a mind.” Similarly, a Memristor may not have a brain, but it can process information more intelligently than any other kind of chip, and it may represent the future of computing.


About the Lecture Series


Professor Chua’s lecture is part of a 12-part lecture series taking place every Tuesday (until November 24) at HP’s headquarters at 3000 Hanover Street, Building 20 - Auditorium, Palo Alto, CA, 94304. Register for the entire lecture series and join us in person or on the web.


Register for the next lecture on Tuesday, September 22 at 10:30 am – 12:00 pm PT: 10 things you didn’t know about Memristors.

0 Kudos
About the Author