IBM display brain chip built-in 1 million digital neurons (video) t420s

IBM to display the "brain" chip built 1 million digital neurons digital news (Yi Jing) Tencent developed a computer, let it make decisions like humans, with human like intelligence, this IBM is to consider the matter, now it is a big step towards the goal. IBM developed a chip called TrueNorth, which can simulate the function of the human brain. IBM is testing the chip to prove how fast and energy efficient it is compared to existing computers. The results are impressive. IBM said, TrueNorth can call the depth of learning technology, through the correlation analysis and the possibility to make decisions, like the human brain. The new computer consumes less energy than other computers with the same purpose. IBM said in a blog, TrueNorth chip has learning and computing ability, it creates a may, we can be embedded in all intelligence computing devices, including networking, intelligent robots, cars, mobile phone, cloud computing and supercomputing. Earlier this year, IBM demonstrated the chip, which was put into a computer called NS16e, which was developed by the human brain. The computer can recognize pictures, sounds and patterns, which are processed by the neural network of the processing unit. The human brain has 100 billion neurons, they can through trillions of nodes (synapses) communicate with each other. Visual cortex can identify visual information, and the other is responsible for motor function. Like the brain, NS16e has a number of neurons, but the number is much less, the system includes 16 TrueNorth chips. Each TrueNorth chip has 1 million neurons and 256 million synapses, which communicate with each other through the circuit. NS16e installed a redesigned memory, computing and communications subsystems, data processing, energy consumption is much lower. IBM has said that TrueNorth processors can be classified by the speed of 1200-2600 frames per second to image data, the energy consumption of only 25-275 mw. The processor can also recognize image patterns that are generated with 50-100 cameras, 24 frames per second. If the chip is installed in the phone, it can also have this ability, a few days without charge. Compared with today’s servers, the new chip energy consumption is much lower, today’s servers rely on traditional chips, such as GPU, CPU, FPGA, companies use these components to identify pictures and voice. Facebook, Google, Microsoft, Baidu depth learning technology for learning, by parsing and pictures, voice recognition related to the answer to learn. These deep learning systems are generally driven by GPU with a power of more than 150 watts. TrueNorth has an algorithm and depth learning model, including the identification model, the past can be associated with the existing data. IBM developed the corresponding algorithm for different depth learning models, but the chip can be applied to existing systems, such as MatConvNet. In essence, developers can use Ma.相关的主题文章: