Researchers at the University of Exeter in England and Stanford engineers work to consider the creation of synapses and electronic neurons whose behavior mimics human brain cells.
The English paper published in the journal Advanced Materials shows that it is possible to design an electronic system that processes and stores the information simultaneously, like the human brain. In a classical computer, these two functions are separate. Very schematically, a processor has memory caches to store information and units of calculation to treat them. However, using an alloy composed of germanium, antimony and tellurium similar to layer writable DVD - RW, it is possible to reproduce the same cerebral phenomenon that allows to store and process information at the same time.
These searches are interesting because they show, for the first time this alloy also known as the GST can be used as a storage medium and unit of General, as calculations of the additions, subtractions, multiplications and divisions. English scientists are considering the creation of "artificial neurons." and synapses "This means that an artificial system composed entirely of phase change materials could learn and process information like our own brains". David Wright Professor leading the research States that "we (the authors of the paper, Editor's note) have revealed a technique to potentially develop a new form of computer systems mimicking the brain which could learn, adapt and change over time.Stanford researchers draw exactly the same conclusion in an article published in Nano Letters. They were able to emulate the operation of a human synapse with a nanoélectrique component. Their research also focused on the functioning of a cell GST imitating a neuron and Synapse. Specifically, the molecular structure of this alloy can be crystalline or amorphous. Nevertheless, the alloy can flaunt crystalline or amorphous States qualified. Research speak of "1% step" and compare these changes with variations in light intensity of a light bulb. A synapse GST would thus have more than 100 States. Ipso facto, an electronic neuron to basis of germanium, antimony and tellurium can store more information than a simple 1 or 0 as human neurons.Similarly, the brain learns through repetition, because a synapse is stimulated, more it will be strengthened and it will be inclined to send information. Similarly, the electrical resistance of a neuron GST drops when it passes the amorphous State in the crystalline state, editing the electrical path depending on its use.
We can imagine faster computers, using less energy, and with a power closer to the human brain. Research use tools of production and materials already available in large quantities, which is good sign for the future discussed technologies. Stanford scholars explain that their purpose is not to replace the processors we use, but to design completely different systems and that are now out of reach. To do this, it will first improve the yields and the density of synapses and electronic neurons to be closer to the human brain. Thus, it would be possible to design platforms capable of processing a very large number of data in parallel to serve as a decision-making aid or enable better understanding of the functioning of the human brain system.Should still take a lot of hindsight. The conquest of the world by Skynet will not be tomorrow. The most powerful supercomputer today application eight and a half minutes to five seconds of human brain behaviour and application 140 000 times more energy.
0 Comments:
Post a Comment