The memristor could be the key to the production of artificial neural networks capable of processing and interpreting all kinds of visual and auditory information, emulating the logical functioning of the brain
An electronic circuit called memristor could be the key to producing a new generation of neuromorphic chips, capable of mimicking the functioning of a biological brain and allow tablets or smartphones to thinking like humans. In principle it is a analog memory circuit designed to learn simple patterns emulating the logical functioning of a brain, but in an improved and more complex version could make computers interpret and learn autonomously with each visual or auditory stimulus of their environment.
The origin of memristors dates back to 1971, when the professor of electronics at the University of California at Berkeley, Mr. Liuhuabing, predicted its operation mathematically. But it wasn’t until 2008 when a group of researchers from Hewlett-Packard they demonstrated Chua’s theory with the development of the first prototype of this electronic circuit.
To this end, Hewlett-Packard researchers together with experts from the University of California and of the Stony Brook University, both in the United States, created a simple circuit whose electrical resistance encoded the current experienced in a kind of analog memory that allowed learn and interpret simple black and white patterns. Obviously, it was a device with certain technical limitations, but its technology could be scaled to create much larger and more powerful devices.
This is not the first time you have tried to create artificial neural networks inspired by biological brains, IBM it has been trying for some time to market a preliminary version of these neuromorphic chips. But silicon transistors and digital circuits in conventional computers are not compatible with the simplified functioning of a human brain’s synapse.
In this regard, Dmitri Strukov, the researcher at the University of California who participated in the design of the memristor, assures that the large number of digital circuits and transistors that are required to emulate a single biological synapse of the human brain, the viability of artificial neural networks based on conventional devices is extremely difficult.
On the other hand, with the development of memristor technology, the design process of these artificial neural networks could be simplified, since each of these memory circuits can represent up to a hundreds of biological synapses. An aspect that could ensure the technical-economic development with neuromorphic chips of low cost, able to process and interpret all kinds of visual, auditory information, etc.
In the conclusions recently published in the prestigious journal Nature, Strukov says that this is a preliminary demonstration, but with recent improvements could produce a neural network chip integrated exclusively by memristors, to create reliable devices much more powerful and complex as tablets, smart phone and even devices data storage.
Images / via Nature and Pixabay
