Let the machine work like a human brain - this is not an illusion, there are scientists on the way to achieve it.
According to foreign media reports, the School of Computer Science at the University of Manchester in the United Kingdom "activated" the world's largest "brain" - a brain-like supercomputer SpiNNaker. According to the official website of the University of Manchester, this computer has 1 million processor cores, which can perform 200 trillion operations per second. The way to process information is similar to that of the human brain. SpiNNaker designer and computer engineering professor Steve Fober said that this type of brain supercomputer "reconstructed the way traditional computers work."
Simulate the information processing of biological brains
"SpiNNaker is called a brain-like supercomputer because it mimics the way the biological brain processes information, and the processing speed and scale are far superior to similar models, but the architecture is significantly different from the traditional supercomputer. I prefer to refer to this kind of machine that mimics the biological brain as a 'class brain machine'." Huang Tiejun, director of the Department of Computer Science and Technology at Peking University, said in an interview with the Science and Technology Daily that supercomputers often refer to higher performance and scale. Large traditional computers, while brain-like machines refer to intelligent machines that borrow and simulate the structure of biological brain nervous systems and information processing processes, rather than traditional computers that simply perform computational tasks.
Qin Zhigang, president of the California University of Science and Technology, also holds the same view. "Class brains work differently than traditional computers." He said that personal computers often have only one central processing unit (CPU). The CPU is powerful enough to handle multiple tasks, but in this mode, tasks can only be It is processed continuously, that is, one is processed before the next one can be processed. But the prototype of a brain-like machine - the way the biological brain works is not the case. According to statistics, there are about 100 billion neurons in the human brain. These neurons are the most basic structural and functional units of the human brain. Each neuron can be thought of as a simplified version of the CPU. Its computational power is not as good as the CPU of the computer, but it is more numerous, and each neuron can complete the task independently. In short, the brain can be thought of as a machine consisting of multiple CPUs running simultaneously, with efficient multitasking information processing capabilities.
Although conventional supercomputers also have a large number of CPUs, these CPUs can only perform simple parallel work. In contrast, the network structure formed by the interconnection of biological neurons is much more complicated.
So, how does a brain-like machine mimic the information exchange and processing between neurons in a biological brain?
Each neuron in the brain is connected to other neurons through thousands of synapses, forming a neural network system that can perceive, comprehensively process, and feed back information. After the external signal is sensed, the upstream neurons can "send" the signal to a number of downstream neurons in the form of nerve impulses, which in turn pass the pulse signal to more neurons. The process of transmitting these nerve impulse signals between neurons is actually the process by which the brain processes information.
"The brain-like brain simulates a biological neural network through a large-scale neuromorphic chip, and a large number of electron or photon neurons and synaptic arrays are integrated on each chip." Huang Tiejun introduced that, unlike biological neurons, the electronic version of neurons is connected. The status can be implemented in software. Brain-like brain processing information is also transmitted by means of transmitting neural impulse signals, but it is usually not directly connected to the biological neural network, but routing is used to improve flexibility. A typical approach is to package the pulse signal and then use the "delivery address" on the package to achieve accurate delivery to downstream neurons; the downstream neurons receive a large number of packets, and then generate new ones based on their processing characteristics. The pulse, then "deliver" the information, and reciprocate. When the computational accuracy of this kind of brain-like machine reaches a certain level, it can produce certain functions that only the biological brain has, and even advanced intelligence such as "inspiration".
More energy efficient, can promote the development of related disciplines
In 1981, American biologist Gerald Edelman proposed the theory of "comprehensive neuromodeling", which became a pioneer in the field of simulated biological brains. For more than 30 years, people have spent a lot of energy and financial resources researching brain-like algorithms, models and supporting hardware facilities. Take SpiNNaker as an example. The project was built in 2006 and has cost about 15 million euros so far.
So, why can brain-like machines make people so fascinated and put a lot of effort into it? “The brain-like machine has the advantages of high intelligence and low energy consumption that cannot be compared with mainstream artificial intelligence models (such as deep neural networks).†Liu Chenglin, deputy director of the Institute of Automation, Chinese Academy of Sciences, said that deep neural networks often only have a single task. Intelligence, such as picture recognition, speech recognition, etc., lacks the ability to comprehensively process information in different scenarios, which is one of the bottlenecks that restrict its future development. Brain-like machines benefit from the “innate advantage†of imitating the brain, and their ability in comprehensive perception and reasoning is more prominent.
Brain-like machines have significant advantages in terms of low energy consumption. The energy consumed by the human brain to perform computational tasks is much lower than current general-purpose computers. As the European Union pointed out in its Human Brain Project report, there is currently no artificial system that can match the low energy consumption of the human brain when dealing with equal tasks. The energy consumed by the human brain is typically around 20 watts, while the power consumption of a typical laptop is about 100 watts. This gap is more apparent in artificial intelligence based on artificial neural networks. In 2016, when the Alpha dog played against Li Shishi, a master of Go, the artificial intelligence program consumed 1 megawatt of power, nearly 50,000 times that of the human brain. "At present, the energy consumption of brain-like brains cannot be reduced to the level of the human brain, but it is more energy-efficient than artificial neural networks." Liu Chenglin said.
On the other hand, brain-like research can promote the development of brain science and neuroscience research. The brain is the most complicated organ in the human body. Its neural structure and operating mechanism still have many unclear places. It is difficult to directly observe it by means of imaging. Liu Chenglin pointed out that brain-like brains, as a computational model for simulating the brain, can generate brain-like activities and intelligent behaviors through calculations, which in turn can provide useful inspiration for the study of brain neural structures and functions. At the same time, advances in brain science and neuroscience will also promote the development of brain-like machines to a higher level of intelligence.
The first real-time simulated human brain machine or appeared 4 years later
"Current brain-based research is still in its infancy, and its learning and creative ability is far less than the human brain. But with the further development of related technologies, it is undeniable that brain-like machines do have the potential to reach or even surpass the human brain." Qin Zhigang Say.
Huang Tiejun describes the blueprint of the future brain-like machine: when the precision of the neuromorphic devices and chips develops to a certain stage, the information processing speed is several orders of magnitude faster than the human brain, and there is no human brain skeleton in the shape. Structural limitations...
How far are we going to go when we develop such a "computer"? Huang Tiejun said that according to the "Human Brain Plan" launched by the European Union, the first machine that simulates the human brain in real time will appear in 2022. After about 20 years, the brain-like machine with the size and the human brain can accurately simulate the human brain function. Or will be available.
When a brain-like machine appears, what changes will it bring to our lives? "The emergence of brain-like machines will inevitably bring about tremendous changes in people's lifestyles, especially in learning styles. Brain-like machines can greatly reduce the work of human repetitiveness, and at the same time it will become one of the sources of innovation inspiration." Said.
Liu Chenglin believes that a robot equipped with a brain-like machine may be functionally the same as a real person. It will think, judge, and learn, provide more intimate services, and replace high-intelligence work for people, greatly improve work efficiency and promote social economy. development of. However, the development and widespread use of high-intelligent machines in the future may also bring about negative effects such as unemployment and misuse. Relevant ethics and risk research should be gradually carried out, and the construction of relevant laws and regulations should be synchronized.
Engine Spare Parts,Large Engine Spare Parts,Car Engine Spare Parts,Diesel Engine Spare Parts
NINGBO CITY YINZHOU RUICAN MACHINERY CO.,LTD , https://www.china-sandcasting.com