The spiral evolution of artificial intelligence
The history of artificial intelligence has been 66 years since 1956. The history of artificial intelligence in the world can be divided into three stages. From 1956 to 1976 is the first stage of development of artificial intelligence, from 1976 to 2006 is the second stage of development, and from 2006 to the present is the third stage of artificial intelligence. a stage of development.
Three waves of artificial intelligence development
Typical AI tasks and applications include machine definition proofs and machine translation. Machine translation is one of the main courses of artificial intelligence, and it also includes machine learning, expert systems, robots and intelligent control, which are all areas of artificial intelligence research. These research categories have also led to three schools of artificial intelligence: symbolism, connectionism, and behaviorism.
There are also literatures that say there are four schools, symbolism, connectionism, behaviorism (also known as evolutionism) and statisticsism ( that is, statistical machine learning)
1956-1976 was the first wave of artificial intelligence . At that time, symbolism prevailed and functionalism occupied the mainstream. Its iconic cornerstone was the introduction of symbolic methods into statistical methods for semantic processing, and knowledge-based methods appeared. Human-computer interaction becomes possible. During the first wave of artificial intelligence, logical operations, deductive reasoning, syllogism, and Prolog logic became languages and became representative methods and methods of the times. In 1958, Simon & Newell put forward the famous prediction in the early days of AI: within ten years, computers will become chess champions; within ten years, computers will discover and prove meaningful mathematical theorems; within ten years, computers will be able to compose and compose music; Implement most psychological theories. But in 1968, apart from a breakthrough in the field of machine definition proofs, the other three predictions did not come true. During that period, there were also negative remarks on artificial intelligence, which brought great damage to the field of artificial intelligence and brought the development of artificial intelligence into a trough.
Until 1973, the United Kingdom published a report that divided the automaton, robot, and central nervous system of artificial intelligence into these three categories for review, and the result of the judgment was that although automata and the central nervous system have research value, the progress It's disappointing, and the robot has no research value. Therefore, in 1974, the United Kingdom and the United States canceled the funding for research and development of artificial intelligence, which ushered in a cold winter for artificial intelligence .
In 1975, Paul Werbos proposed the BP algorithm, which made the learning of multi-layer artificial neuron network possible. This makes artificial intelligence slowly usher in the spring. The neural network builds a set of algorithms, eliminating the need for manual adjustment operations and reducing the error rate between input and output. The artificial neural network has also become the second wave of artificial intelligence. The Japanese government also launched a large-scale logical reasoning attempt represented by the fifth-generation machine, and raised logical reasoning to the level of knowledge engineering, and experts in some fields began to establish rules for artificial neural networks and use these rules for reasoning. However, due to the fact that the machine runs much faster than the reasoning, the construction of the model is not ideal, and it failed after ten years of development. While search engines have gradually begun to rise, Stanford University used experts to build Cyc, an encyclopedia of knowledge, which gradually declined in the late 1990s. At that time, it was also proposed that artificial intelligence cannot be expressed manually by experts, but must rely on automatic learning by machines, and artificial intelligence ushered in the second cold winter .
Since 2006, artificial intelligence has achieved the third development . In essence, there is no essential difference in methodology between the second wave and the third wave, but great improvements have been made in terms of hardware technology, which can support multiple computers to perform calculations on computing resources at the same time. At the same time, the development of data has gradually matured, and people use data training to solve problems. The third wave is also a joint breakthrough of the combination of "deep learning algorithm + big data".
The rapid rise of artificial intelligence was actually Li Feifei's proposal to build a database of 10 million pictures. At that time, the cost of marking a picture was 5 US dollars, and the cost of 50 million US dollars brought huge troubles. The form of online crowdsourcing promotes programmers to continue to join in this work, gradually forming a kind of competition. This has promoted the evolution of image label efficiency verification algorithms, and ImageNet has promoted the technical development of face recognition. The third wave makes artificial intelligence popular, with clear application scenarios, big data + computing power support, and faster algorithm evolution.
The Troika of Artificial Intelligence: Data, Computing Power and Algorithms
The development of artificial intelligence has been ups and downs so far. It is a spiral development in itself. In the future, it will conduct rounds of drills in areas including computer science, electronics, and automation. In terms of application, algorithms based on deep neural networks will gradually be widely used. The future of artificial intelligence also expects more new algorithms and theories to appear.
In the past 60 years of development, AI has encountered two cold winters. In recent years, with the successful application of deep learning, it has become hot again. Now that deep learning may be facing a bottleneck period, some people have begun to promote the third cold winter. Eventually, though, artificial intelligence will return to the path of steady development. The so-called cold winter theory is only a transitional period for a technology.