Silicon has supported the computer work for nearly half a century, both graphics and digital operation, all of the information processing together by millions of tiny logic gate, but these circuits are made of silicon.
Moore's law pointed out that in the number of silicon transistors in microprocessor will double every two years, but this may not always apply, because can carry the number of transistors on a chip is limited.
Semiconductor industry association (SIA) is expected to the current trend, computer energy demand in 2040 will be more than the world's total energy supply.
Researchers around the world are developing the huge amounts of data to be processed silicon replacement of new computing systems. These systems to chip smaller than now, energy efficiency is higher.
JulieGrollier leadership team is the French UMPhylab laboratory design can nano-devices in imitation of a human brain work. Her team operations with magnetic particles, especially in pattern recognition.
When magnetic particles is very small, they will be in a state of flux, magnetic field began to irregular oscillation. By applying electric current, the team can use such oscillations to complete basic operation. If the computing device to expand the scale, Grollie believes that the technology can complete recognition technology faster than the existing model.
The technology will also reduce energy consumption. Grollier said the spontaneous magnetic oscillator is working one percent of the energy is needed for the silica based computing systems, the device size is only one over ten thousand.
Founded in Paris last December IgorCarron LightOn company found another alternative to silicon chips, is - light.
How about the future LightOn computer specific works, Carron does not talk more, but they will build a optical system can deal with large data sets, so that the machine learning algorithms to more easily use these data. This system USES is referred to as random projection (randomprojection) mathematical methods. Random projection method is put forward in 1984, but because of a large amount of calculation, the silicon chip incompetent. Now Carron and his colleagues found the run to complete all operation method.
The new method of processing data and using data to study what kind of role will play? Carron think, if you don't rely on large processor machine learning, the wearable computing devices can be rapid development. It can also make the emerging technologies of computer into everyday objects "Internet of things" more powerful. These wearable devices are no longer needed will return to the data center for a large amount of data, but can realize real-time data processing.
Grollier and Carron invented equipment is not the only alternative computing technology. A Stanford university research team developed a chip contained 178 transistors, the transistor consists of carbon nanotubes. Carbon nanotubes unique electrical properties make them relative to the silicon crystal is a more efficient switching devices. Earlier this year, Israel's ben-gurion university and the Georgia institute of technology researchers use DNA to create the world's smallest diode, the diode is the basic components of electronic computers.
So far, can handle huge amounts of data of high performance silicon chip computer continues to make significant progress in the field of machine learning. But the silicon chip computer performance can never keep exponential growth. To fully processing and use of data around the world, we need the ubiquitous smart learning devices. Companies like Facebook and Google are currently only touches the fur of these data. Carron said: "the big Internet companies every day in the accumulation of vast amounts of data, but they haven't fully excavate its value."
