A powerful new stamp-sized chip has been introduced, using a process that mimics the performance of the human brain to provide supercomputers. The neurosynaptic chip is a breakthrough that opens up a new range of possibilities for calculating self-driving artificial intelligence systems that can be installed on smartphones, scientists say. Researchers from IBM and Cornell University and their collaborators from all over the world said that they have taken a new method design and moved toward a system called "cognitive computing." "We have taken inspiration from the design of this brain for the cerebral cortex," said the brain-inspired calculations of IBM's chief scientist, Dharmendra Modha, referring to the command center of the brain. HENAN JINHE INDUSTRY CO.,LTD , https://www.hncarbonblack.com
He said that existing computers track their descendants back to machines that essentially belong to the "Continuous Digital Computation Calculator" performing math or "left brain" tasks, but no one else in the 1940s. The new chip is called TrueNorth's work, mimicking the "right brain" function of sensory processing, responding to scenes, smells, and environmental information to "learn" to react in different situations, Modha said on Thursday.
It does this by using a huge "neuron" and "synaptic," similar to how the human brain accomplishes this by collecting information from the body's sensory organs using information. The researchers designed TrueNorth with millions of programmable neurons and 256 million programmable synapses, with 4096 cores and 5.4 billion transistors on a single chip. A key performance is the chip, which operates on the energy of a hearing aid battery equivalent to very low energy consumption. This can be installed on the chip of a car or smartphone without the need to connect to a cloud or other network to perform real-time computing on the supercomputer.
"The sensor will become a computer," Modha said. "You can have better sensory processors without connecting to Wi-Fi or the cloud. This will allow autopilot vehicles to, for example, find problems and handle them even if its data connection is interrupted. "It can see impending accidents." . "Similarly, mobile phones can use odor or visual information to interpret them in real time without the need for network connectivity." After years of working with IBM, we are now getting closer to building a computer that is similar to our brain." Manohar, of Cornell University Technical Research, said.
Researchers say that TrueNorth is better than current supercomputers in some respects, although direct comparisons are impossible because of the different ways in which it operates. But TrueNorth can provide 46 to 400 billion calculations of the second watt of synapses. In contrast, the most energy-efficient supercomputer can provide 4.5 billion "floating point" calculations for a second of a watt. Using Samsung's 28nm process technology chip manufacturing.
"This is an amazing achievement that makes full use of traditional craftsmanship to use commercially available low-power mobile devices to provide chips, simulating human brains by processing the extreme amount of power that senses information," Samsung Electronics' Sean Han said. “This is a huge architectural breakthrough and it is necessary because this industry is moving toward the next generation of cloud computing and big data processing.†Modha said that researchers only produce chips, which may take several years to commercialize applications. But he said that it is possible to change the potential of society, and pointed out that hybrid computers may one day combine the "left brain" machine with the new "right brain" device with even better performance.
Ultra-fast new chip mimics human brain