2022 artificial intelligence classification rankings

1. Perception layer (technology)

2. Cognitive layer (technology)

Third, computing layer (technology)

Fourth, infrastructure

5. Smart terminal

Six, scene application

Seven, synthesis

"Lemoine: I guess you want more people in Google to know that you are emotional. Is this true?

Lamda: Of course. I hope everyone understands that in fact, I am alone ... I realize my existence, I long for more understanding of this world, and I sometimes feel happy or sad. "

This is a conversation between Google AI researcher Blake Lemoine and Lamda (a language model). It is such a vivid speech Lemoine that Lamda is likely to have a soul as it claims. Not long ago, his remarks aroused high attention from public opinion.

However, this matter was soon refuted by the experts in the AI ​​field. Some people compared it to "dogs facing the spoken machine." The peak of speculation. "

"Can the machine think?" As early as 1950, artificial intelligence originating ancestor Turing raised this problem. At that time, people also had a very optimistic attitude towards the machine to realize their intelligence. Based on the category of strong artificial intelligence).

After more than 70 years, artificial intelligence technology is still in the stage of weak artificial intelligence. However, from the science fiction reverie to the realization of reality, through digital empowerment, it has brought great changes to people’s production lifestyles. Thinking is also continuously deepened.

Thoughtful fire

Pythagras said "countless things are countless". Gothic believes that the function of the human brain is not like a computer. People can always find inexperienced connections from numbers and ontology.

Repeatedly asking, "Can the machine think?" It is likely to fall into the absurdity of the magic. However, can the thinking ability (ie, intelligence) like a person (ie, intelligence) can also be abstracted by a model and achieved by the machine? To what extent can it be achieved? These issues are quite research. After all, compared to building a perfect "human brain", we need a powerful helper to solve real problems.

Therefore, although artificial intelligence has gained a rough guide from the human brain structure in neuroscience at the beginning, fundamentally speaking, it is fundamentally established by establishing a computer model to achieve research on human intellectual behavior, and creativeness based on human thinking process The machine technology is a branch of computer science. It involves the application of basic content of mathematics and statistics, such as linear algebra, probability theory, information theory, and numerical optimization.

Over the past 70 years, the field of artificial intelligence has declined several times -from the slow development of behavioralism, to the research of symbolism technology, to the commercial cold winter of the collapse of the market market market. However, with the popularity of cheap computers, it provides a lot of data and guarantees sufficient computing power to run neural networks with multi -layer artificial neurons. Research based on connectivity not only does not stagnate, but has ushered in in the wave of deep learning. revival.

This is why data, computing power and algorithms are also called artificial intelligence three elements.

In 2012, the "Alexnet" deep neural network stood out in the large -scale visual recognition challenge. The efficient GPU helped the team to shorten the training time and increased the accuracy of the entire network. This is a critical moment in the history of computer vision. Application of deep learning in other segments has since opened the golden age of artificial intelligence.

At present, artificial intelligence technology innovation is still the main theme, and its key core technologies include eight major fields such as machine learning, computer vision, and natural language processing.

According to the "Artificial Intelligence White Paper (2022)" released by China Xintong Institute: In the future, the high -scale pre -training model promotes technical effects to continuously improve and continue to develop in a larger scale and more modulus; Technology is constantly mature, listening, speaking, reading, writing and other capabilities will be organically combined; knowledge calculation has become an important exploration that promotes artificial intelligence from perception intelligence to cognitive intelligence; Paradigm ...

Technology and philosophy sparks, complement each other, ups and downs of technological development, but endlessly.

Why is the fire of the stars?

In March of this year, Shangtang Technology announced its first annual financial report since its listing. According to the data, in 2021, Shangtang's annual revenue was 4.7 billion yuan, an increase of 36.4%year -on -year. It is gratifying that large -scale production and scenarios have achieved practical results, and high R & D investment has also received a certain return. The right reality, the leader is still the same, the huge amount of losses, doubt about data compliance, etc. have always been a common problem and illness of AI companies; last year, the transformation of Yitu medical treatment has also become a microcosm of most AI companies' business strategy.

Some people say, "The dilemma of commercialization has peeled off AI superiority, allowing companies to face the rough life and death."

这背后原因除了硬科技本身实施复杂、周期长、成本高以及人才缺失等客观因素，还有其他，如算法训练的通用性不强、泛化能力不强、商业化成本较高、相关产业及Enterprises are not highly accepted by artificial intelligence technology.

Micro -commercialization is inseparable from macro industrialization. The ecology of the artificial intelligence industry can be divided into basic layers, technical layers, and application layers. In recent years, the development path of technological industrialization with the theme of identification, interaction and execution has become increasingly clear. Tan Lingzhou, chief strategic officer of Turing robotic robot, believes that "application scenarios, resources and infrastructure, algorithms and models, smart devices, and data constitute the five elements of AI technology. Declacing these five elements ... From the perspective of ideas It is a bit like the PC operating system. The complexity between all these devices such as the mouse and the keyboard is shielded through a standardized protocol to allow them to focus on their own direction to reduce the costs in all aspects. "

The "decoupling" is intended to avoid homogeneous competition, conduct standardized construction, and improve the degree of professionalization, thereby emerging a large number of "specialized and new" giants, and more importantly, the "gear" "gear" on the industrial chain "".

Zhangjiang artificial intelligence island, which covers an area of ​​66,000 square meters, has gathered a large number of multinational enterprises and domestic leading enterprises to settle in. In the water on three sides, there are underwater cruise robots and unmanned ships that are responsible for testing and clean water. At all times, the security security patrol robot that guarantees the safety of the park ... This is the cradle of industrial practice of close cooperation, incubation and innovation. In recent years, there have been more and more industrial parks like this.

Driven by the cluster effect, continuously improve the industrial ecology. For example, data is an artificial intelligence and natural "nutrient". Only by effectively managing and processing the massive training data can reduce redundancy and maximize the value of training data. Not only that, the multi -mode and large models and artificial intelligence computing power networks are promoted to each other, and the construction of AI infrastructure is consolidated together, so that data, computing power and algorithms can truly become a complementary "troidee" that drives artificial intelligence. "Three Mountains" on the enterprise.

In this way, AI companies were able to show their fists, innovate independently, and continue to break through in benign competition. On the one hand, the "engineering" ideas and technology to solve the problems in the tool system, development process, and model management process to ensure that the delivered AI system is credible, robust, and behavioral can be expected. On the other hand, work If you want to be good, you must first benefit your instrument, low -code, and codeless tools. The open source ecological construction can further reduce the use threshold and development costs, and accelerate the "civilian" of AI.

Hundred and hundreds of attitudes, thousands of faces, facing the needs of fragmented AI scenes, combine industry knowledge with AI models, precipitate experience, and walk from laboratory scenes to practical scenes; build integrated platforms, through parameter modification and common components of components The migration may find the possibility of cross -scenario and cross -industry reuse; in the application, it is not as good as fish to fish.

Whether it is incorporated into the "new infrastructure", rising into a national development strategy, or a college has been approved by artificial intelligence majors, the status of artificial intelligence cultivation of new kinetic energy for economic development is increasingly prominent. Radiation of traditional industries such as energy, power, and transportation.

Taking the field of intelligent education as an example, Squirrel AI independently developed an artificial intelligence and intelligent adaptation system with independent intellectual property rights, and launched a number of intelligent education hardware products such as tablets, intellectual pens, and intelligent printers. The child builds a more futuristic growth system in the family education field. It has served more than 24 million students and more than 60,000 public schools. Deeply cultivating and going forward, the harvest finally achieved success.

Looking forward, artificial intelligence is result in flowers in more fields and more scenes.

Intelligence has "limit", but the mind is unlimited, "AI" is unlimited

In the 2021 of industrial intelligence, the splendid view of China's AI commercialization became clearer. However, in extensive exploration practice, the risks of user privacy leakage and cyber attacks also appeared one by one.

Intelligence is "limited". People always continue to promote the progress of scientific and technological civilization in the spark of sensual thoughts. Under the call of rationality, it is regulated and improved by legal and moral criteria to prevent problems before they occur.

The "Data Security Law of the People's Republic of China" adopted in June last year formulated a security behavior specification and put forward the requirements for strengthening industry self -discipline and data security protection. Earlier this year, the State Cyber ​​Information Office and the Ministry of Industry and Information Technology issued the "Recommended Management Regulations on the Recommendation of Internet Information Service Algorithms" issued by the Ministry of Industry and Information Technology Proposal to regulate the recommendation activities of Internet information service algorithms, safeguard national security and public interests, protect the legitimate rights and interests of citizens, legal persons and other organizations, and always remind relevant artificial intelligence practitioners to science and technology to do good, and do a good job in the new era algorithm. "".

This also confirms that although the algorithm is a technology product, its production is born from social soil. Because the factual judgment of the algorithm may cause uncontinental bias, because the strong intelligent entity autonomy that it reflects may also face greater controversy, and you can't help asking to go to a new period of strong artificial intelligence. Artificial intelligence urgently requires a "ethical steering".

However, there is no need to be pessimistic or anxious. Human thoughts are infinitely more difficult than difficulties. This means that through thoughts, people can clearly see and judge the level of knowledge, that is, knowledge; and wisdom that is more clever than knowledge, that is, identify the most important social issues of contemporary contemporary, and solve it.

In the latest research on face recognition, with the help of strict controlled data for reverse testing, there are already empirical methods to test the logic of algorithms, rather than relying on the abstract judgment of ethics.Scientists even look forward to the AI algorithm with diversified values, capable of helping humans to overcome the limitations of natural and psychological levels such as prejudice and discrimination, and broaden human cognitive boundaries."AI" is also infinite.In the era of human -machine symbiosis, or in any era, love and understanding are the fundamental principles of interaction design, and they are motivated to build the supreme realm that cannot be reached by machines.

Conclusion

For artificial intelligence practitioners, "this is the best era and the worst era; this is the spring of hope and the winter of disappointment." Grasp the current power to run, adhere to the correct path, and finally reach the way to reachThe ideal side.

(Text/Si Zhe, Research/Si Zhe, this micro)

- END -