For many scientists in this field, one of the ultimate goals of artificial intelligence is to use the machine to achieve the full function of the human brain, and the smallest cell unit of the human brain, the neuron, may be the best starting point.
On August 3rd, US local time, IBM officially announced their latest achievement - the first human neuron, which can be used to manufacture high-density, low-power cognitive learning chips.
The IBM Zurich Research Center produced the world's first artificial nanoscale random phase change neuron. IBM has built an array of 500 neurons and has the array signalled to mimic the way the human brain works.
This technological breakthrough is important because phase-change neurons have properties that are incomparable to neurons made from traditional materials—the size can be as small as nanometers. In addition, its signal transmission speed is very fast and power consumption is very low. More importantly, phase-change neurons are random, which means that the output of multiple phase-change neurons will be slightly different under the same input signal, which is the nature of biological neurons.
The first author of artificial neurons paper: Thomas Fatuma (Tomas Tuma)
IBM phase-change neurons consist of an input (a dendrites like biological neurons), a neurofilm (a bilayer that resembles biological neurons), a signal generator (a neuron-like body like a biological neuron), and an output (similar to a living organism). The axon of neurons constitutes. There is also a feedback loop between the signal generator and the input to enhance certain types of input signals.
Artificial Neuron R&D Team, Image Source: IBM
The nerve membrane is the key to the entire neuron. In biological nerve cells, the nerve film acts as a liquid film whose physical mechanism is similar to resistance and capacitance: it prevents current from passing directly, but at the same time absorbs energy. When the energy is absorbed to a certain extent, it emits its own signal. This signal is transmitted along the axons and is received by other neurons. Then repeat this process.
In neurons made by IBM, the liquid film is replaced by a small piece of neurofilm. The nerve film is made of a ruthenium composite (also known as GST material), which is also the main functional material for rewritable Blu-ray discs. The ruthenium composite is a phase change material, that is, it can exist in two states: a crystalline state and an amorphous state. By providing energy through a laser or current, the two states can transition to each other. In different states, the physical properties of the phase change material are quite different: the tantalum composite is non-conductive in the amorphous state and conductive in the crystalline state.
In artificial neurons, the ruthenium film is initially amorphous. As the signal arrives, the film gradually becomes crystalline, that is, gradually becomes conductive. Eventually, a current is passed through the membrane to create a signal that is emitted through the output of the neuron. After a certain period of time, the ruthenium film returned to an amorphous state. This process is repeated.
Comparison of biological neurons and artificial neurons, image source: IBM
Biological neurons are random (StochasTIc) due to the presence of various noises in the organism. IBM researchers say that artificial neurons also exhibit random characteristics, because the neuron's film has a slightly different state after each reset, so the subsequent crystallization process is slightly different. Therefore, scientists cannot know exactly what signal each artificial neuron will emit.
So what is the significance of artificial neurons?
First of all, artificial neurons use mature materials, after several billion work without damage (long life), the body is positive and small (it is reported to be 90 nanometers, but from the following figure should be around 300 nanometers, and the paper It indicates that the future is expected to reach 14 nanometers). Therefore, this is a very good device.
Artificial neural network. The silver square in the figure is the amplified phase change neuron, which is not equipped with an industry standard input and output interface. Image source: IBM
Second, artificial neurons work very much like biological neurons. When a large number of artificial neurons form a parallel computer, it may be able to make decisions and process sensory information like humans. IBM said that their artificial neuron technology complements the memristor, another artificial neural component currently under development.
Currently, IBM has created a 10 by 10 array of neurons, combining five small arrays into a large array of 500 neurons that can be processed in a manner similar to the way the human brain works. In fact, artificial neurons have shown the same "collective coding" characteristics as human neurons. In addition, its signal processing capabilities have exceeded the limits set by the Nyquist-Shannon sampling theorem.
Editor's Note: Collective Coding: Each neuron has 2 states that can represent 1 bit of information, then N neurons can represent 2N bits of information. When the number of neurons is large enough, the amount of information that can be expressed will be extremely alarming.
IBM researchers plan to build a single chip containing thousands of phase-change neurons and write software that takes advantage of the random nature of phase-change neuron chips.
We specialize in cable assembly overmolding. Our customized molded cables are built precisly following customers' spec.Experienced proposal can be offered to customers for evaluation by considering stable quality and competitive price.
we have internal design, prototype, and manufacturing the widest range of molded harnesses, also a diversified line of strain / flex reliefs and grommets.
Molded Cables,Stranded Round Molded Cables,Molded Patch Cord Cable,4 Pole Waterproof Molded Cable,Waterproof Connectors
ETOP WIREHARNESS LIMITED , http://www.wireharness-assembling.com