What is "computing power"?
In recent years, the concept of "computing power" has frequently appeared in the Internet circle and has become a hot word often hung on the lips of science and technology leaders. But in fact, this is not a new concept, the original meaning of computing power is not complex, as the name suggests, is the ability to calculate. It is the ability of the server of the data center to realize the result output after processing the data.
The human brain is a natural computing server. In life, as small as going to the food market to buy food, as big as solving a math problem, the human brain is providing computing power for our needs through oral calculation and mental calculation at every moment. But such computing power is a little low, especially when it comes to very complex computing problems, which has given birth to generation after generation of computing tools.
First of all, of course, we have to find tools from humans themselves, so the finger became the simplest and most convenient counting tool that has been used ever since, and this is probably the fundamental reason why decimal computing was invented and popularized.
With the continuous development of production practices, man finally invented his own computing tools. At each step of the problem solving process, there is a standardized procedure, which is therefore considered by some historians of mathematics to be structural and "mechanized".
However, similar to the improvement of tools in the mechanical age, with the development of computing software, that is, algorithms, faster and faster, computing hardware, that is, "chip", although it has been improving, but eventually can not keep up with the progress of the algorithm, and eventually replaced by more convenient abacus.
It was also the first major revolution in the history of computing tools. The principle of the abacus is similar to the calculation chip, but it is more practical in operation, especially after mastering the abacus formula and basic skills, the calculation speed is amazing. Around the 15th century, the abacus invented by the Chinese people was introduced to Japan, South Korea and other countries, and gradually introduced to the West through business, and it is still widely used in some areas today.
Modern computers and Moore's Law
From the ancient knot counting to the manual computing era, and then to the mechanical computing era, every change in computing tools has brought about a great leap in computing power, and further promoted the development of human economy and society.
But in the past, people's perception of computing power was not obvious, until the birth of electronic computers, so that people deeply feel the power of computing. With the support of such super computing power, many previously unimaginable problems can be solved in an instant. It is also the emergence of generation after generation of faster and faster electronic computers that makes the importance of computing power really concern people.
How is the computing power of modern computers measured?
Simply put, when a human uses a computer, an input character or message first needs to be converted into an encoding that an electronic computer can handle: a permutation of 0 or 1. The electronic transistor is the basic unit for numerical operations as 0 or 1, at this time the electronic transistor's path can represent the value of 1, the open road represents the value of 0, each 0 or 1 is a bit (the smallest unit of information). This also means that the size of the computing power is positively correlated with the number of transistors, and the more transistors there are, the more information can be represented.
It can be said that the emergence of chips has really promoted the rapid development of modern computers. Following this trend, chip density and computing power will increase exponentially relative to the time cycle. This discovery became known as Moore's Law.
For more than half a century, the development of electronic computers has basically followed Moore's Law. This also marks that the computing power of electronic computers is finally no longer only for a few powerful institutions to serve, but began to truly move towards thousands of households, and then promote the arrival of the information age of the whole people.
Computing challenges in the age of artificial intelligence
According to Moore's Law, computing power doubles roughly every two years. However, such a terrible growth rate can not meet the needs of human computing power, especially after the development of deep learning led the rise of artificial intelligence in 2012.
In this case, how can the computing power be increased to meet the needs of intelligent computing? There are two big areas for breakthrough.
3.1 Aircraft and sea tactics
The first method, to find more people to calculate together, that is, to use high-performance computing and distributed computing to boost power, is now the most important way to improve computing power.
Among them, the core technology of high performance computing is parallel computing, that is, the process of using multiple computing resources to solve computing problems at the same time.
In the era of cloud computing, data centers are the main carriers of computing power. This is also the fundamental reason for the construction of information infrastructure such as big data centers, cloud computing centers, intelligent computing centers, and supercomputing centers across the country.
3.2 Equipment Improvement
The second method is to fundamentally improve the equipment and achieve a breakthrough in the computing mode.
Although high performance computing and distributed computing are the main means to improve computing power at present, there are many restrictive factors, and the improvement of computing power is limited. Therefore, the second largest method to improve computing power has become the pursuit of many cutting-edge scientists, a typical representative of which is quantum computing.
However, there are still other problems: one is to find a suitable and universal physical carrier to maintain the quantum superposition state, and the other is a widely used computing scenario. So far, the various quantum computers reported can only perform a few special scenarios, but the future is very promising.
Although computing power is not the only factor affecting the development of artificial intelligence, it is one of its necessary conditions. The horn has sounded, and in the next decade, if you want to win on the track of artificial intelligence, computing power is probably the first line to be broken.
