Imitation of the human brain neural chip “is on the rise

cloud network hunting note: perhaps you’ve heard people say: in the near future, we will be able to produce like human machine, it can not only learning, reasoning, and even can be like a human emotion to deal with the problem. You must feel wishful thinking, but the day really not far away.

for some, the “brain” chip is what they dream; If we can develop to work like the brain of a computer chip, which, citing the development of artificial intelligence to provide complete intelligent countless possibilities.

but for others, they think “brain” chip and the idea of flying cars is not practical, “brain” chip, but they are one of the many things of interest.

chip, also known as the “brain” nerve type chip, the chip of the truth lies in somewhere in the middle of.

the primary drivers of neural network technology is an inconvenient truth: famous physicist and mathematician John von neumann created a set of driving the development of the information age structure (referred to as the von neumann architecture), the structure is also very good explains the physical factors of the development of neural network technology.

you may have seen these effects: have you noticed that your new cell phone is not advanced than your old cell phone? This is due to the application of the von neumann architecture chip performance bottleneck problem, and thus designer must be in the chips now achieve a good balance between functionality and power consumption. Otherwise you will have a very powerful mobile phone, but its battery only for 12 minutes.

, by contrast, are equipped with a “brain” of the chip computer function is very powerful, and if you use it to read will only consume 20 watts of electricity. If reach such a high energy efficiency, this will lead to machine intelligence into a new era. So what are you hesitate?

the neural network is not a new idea. Research on neural network can be traced back to the 1940 study of computer chip technology.

now, a lot of focus on image recognition function of the Internet giant is added to the competition neural network software, like Google and Facebook are related to entrepreneurship company invested a lot of material resources. However, since these chips used in expensive supercomputers, and their potential is always limited. Neural network is applied to neural network software chip. Differs from other chips, using neural network chip autonomous system structure, the one of the key factors for the neural network to attract people.

so, we need to use how long to get these neural network chip?

in fact, since 1993, there is a class of neural type chip is available for use. In 1993, a small independent team with an idea came to IBM, the group wants to develop a neural network chip, they call it ZISC (zero order type computer), the chip become the world’s first commercial neural type chip.

related experience before the independent team, they have at CERN for the super proton synchrotron, particle accelerators and the large hadron collider (LHC) build neural network pattern recognition software.

the independent teams in the von neumann system neural network the inherent limitations of upset, they eventually came to the conclusion that build neural hardware is the unique function of neural network and the best way to IBM’s expertise, thus equipped with 36 artificial neuron ZISC36 chip was born. IBM and General Vision company from 1993 to 2001, after eight years have been selling this kind of chips and launch ZISC78 chip (78 neurons), in 2001, IBM quit business ZISC chip manufacturing and research and development.

the General Vision companies choose to continue to business ZISC chip manufacturing and research and development, because they believe that the technology has many unexploited application. After 5 years of project work, they use their professional knowledge to work in the neural patterns of development; They managed to raise foot enough money to new product development ZISC chips. In 2007, the General Vision launch of Cognitive Memory 1000, namely CM1K: this is a made up of 1024 artificial neurons work in parallel and the nerve type of chip, it only 0.5 watts of power consumption, and the ability to recognize and respond to a few delicate time various data model (image/code/text/any). Looks like the gambling on neural type chip can be, because after nerve pattern research process accelerated.

in 2008, is only the CM1K chip developed after a year. The defense advanced research programme (DARPA) announced SyNAPSE plan, namely neurological adaptive plastic can be miniaturized electronic system. The defense advanced research projects agency authorized IBM and HRL laboratories engaged in neural type chip research and development.

in 2009, it is reported that a European team is developing with a 200000 neurons and 50 million nerve synaptic connections of the chip, thanks to Europe called FACETS (fast simulation) in an instant. Called FACETS, in turn, led to the new European plan, the program was launched in 2014 known as the human brain project.

many universities began to (or to update old plan) plans to study nerve type chip, and the interest in continuous fermentation.

in 2012, Intel announced it is developing with a new structure of the neural type chip, and qualcomm in 2013 acquired some new neural network startup. Just recently, in August 2014, IBM released the TrueNorth chip, this chip is equipped with 1 million neurons and 256 million programming synapses, this is the most powerful in the world and one of the most complex chips. This chip has its own custom programming language, all of this comes from the defense advanced research projects agency (darpa) of SyNAPSE.

TrueNorth chip is an impressive technology, it bodes well for some more lofty dreams about neural network, but unfortunately because of cost, TrueNorth chips could not be applied to your next tablets. But it does prove a point: we are entering type neural network chip can work age, type of neural network chip is no longer stay in the field of pure science fiction.

why it took neural pattern for so many years to succeed?

to some extent, the success of Moore’s law means that the engineer is no longer needed neural network structure, when I heard that you may be surprised, but slow nerve type chip application problem is not due to technical reasons, but stems from a lack of faith. Neuropathic chip back into the world, this also makes the “brain” chip is open, it is also an IQ can reach 8 chip why still can cause people interested. This is a kind of chip can be taught something, yes you heard me, was the church, rather than coding. You can teach it to recognize faces and code, then you can teach it in a few microseconds identify what it learned from large amounts of data. Of course the chips easily and almost all modern electronic integration.

neural pattern technology has the potential to change everything. Maybe one day we have the ability to download to the brain nerve type chip, so that we can achieve a sense of eternal life. Maybe this day have to wait a long time, very but before that, more practical applications have gotten just around the corner.

these practical applications include: applied to eeg/automatic ecg monitor, when detected an irregular heartbeat, will give alarm. Can be applied to mobile phones, it can identify the photo on the face of friends, and automatically send them a copy; Used in pet door, the pet door can only allow your pet to; Used in automobiles, the car could identify its owner and automatically adjust the Settings; Can be applied to robots or unmanned aerial vehicle (uav)… The application examples abound.

through the brain nerve type chip, we can make the things we “know”, we use can also create some neural pattern, also can open a real era of “smart” technology. We frankly: actually your smartphone is not “smart”, but it is a carefully conceived tool. We say something smart: such as “smart”, it must first have the ability to identify the input data and pattern recognition, which are the nerve type chip is very good at something. You could even say that nerve type chip cognitive ability.

Source: TC