Editor's note: In the past year, we have been dizzying with the remarkable achievements made by human society in the fields of AI, embodied robots, and brain-computer interfaces. The door to a new era has opened. Standing at the beginning of the Year of the Dragon, we would like to ask, where will the new qualitative productivity brought about by scientific and technological progress take us? What kind of picture of human society will be opened up by optical chips, flying cars, and evolving AI? Here, please join Silver Persimmon Finance to listen to the voices of scientists at the forefront of technology.
From the changes of the times such as stone, bronze, iron, steam, electricity, and information, all of them demonstrate the persistent efforts of human beings to promote social progress through the continuous invention and use of epoch-making tools. At present, as human society is moving from the information age to the era of digital intelligence, it is a strategic technology leading a new round of scientific and technological revolution and industrial transformationArtificial intelligence embodies the "head goose" effect, which not only innovatively allocates production factors, but also empowers the deep transformation and upgrading of industries, optimizes the combination of laborers, labor materials and labor objects, improves total factor productivity, and accelerates the formation of new quality productivity.
Wu Fei, Director of the Institute of Artificial Intelligence, School of Computer Science, Zhejiang UniversityArtificial intelligence (AI) is a general-purpose enabling technology and is considered to be the main front for the development of new qualitative productivity. Just as general-purpose technologies such as steam engines, electricity, computers, semiconductors, and the Internet have appeared in human history, AI must be combined with scenarios or industries in the form of tools, products, or platforms to empower thousands of industries and promote social progress. Unlike steam technology, which is used as a tool to improve human physical strength, artificial intelligence technology constructs artifacts (such as applications or robots) that can complete certain tasks independently or in collaboration with humansEven human beings themselves find it difficult to accomplish the task aloneto become a worker or a part of the workerAmplified human "intelligence".
Early AI was able to solve only toy-like problems, not real-world tasks, and thus fell into the middle of AI winter twice. Today, driven by the data-driven innovation model of "data is the fuel, the model is the engine, and the scene is the space", artificial intelligence can perceive and recognize the massive data that depicts human work and life like never before, understand the environment of individual groups, and then form decisions and take actionsPromote the transition from "Internet +" to "data elements", so that data, a means of production, can generate great value and be used by human beings.
Since the end of 2022, generative AI has shown strong content synthesis capabilities, driving breakthroughs in areas such as language generation and conversational AINatural language interaction is used to complete many tasks that are difficult to reach with traditional artificial intelligenceIt provides a way to explore the realization of general artificial intelligence, which has attracted wide attention from various industries and is known as the "iPhone moment of artificial intelligence". Among the top 10 people of 2023 listed by the British journal Nature, in addition to the customary selection of ten figures from major scientific events around the world, ChatGPT, a product of generative artificial intelligence, is also included.
At the beginning of the second industrial revolution in the electric age, Marx pointed out that machines are different from ordinary production tools, and are "human brain organs created by human hands", which is not only a powerful organism that is alien to itself, but also a "universal intelligence" accumulated and shared by the whole society.
January 9, 2007Steve Jobs released the first generation of iPhoneThe organic integration of iPods and mobile Internet equipment has promoted the development of the mobile InternetMobile phones have become the main entrance to information traffic。In the process of moving from the Internet era to the digital and intelligent era, scientists are wondering whether a new information flow portal will emerge.
As a large language model, ChatGPT combines various applications in the form of plug-ins in the form of natural language interaction, and becomes a traffic entrance linking the three-dimensional space of human society, information space, and physical world. The interaction between traditional humans and machines needs to be realized through application software, but in the era of large language models, the interaction between humans and machines can be realized through natural language.
In order for large language models to complete tasks better,The prompt language that describes the task becomes even more important. The clearer the description of the task and the more specific the desired outcome, the better the answers synthesized by the large language model. In other words, people and artifacts need to work together to accomplish the taskFor example, Fields Medal winner Tao Zhexuan believes that instead of trying to make a large language model answer math problems directly, let it play the role of a collaboratorAsk it for policy recommendations；Microsoft released Copilot, an artificial intelligence chatbotIt is pointed out that the new form of work in the future is that users should not pay attention to how AI affects their work, but apply AI technology more efficiently in their work.
The success of large language models in the field of chat has aroused the interest of researchers in applying large language models to professional fieldsA number of large-scale models for vertical fields such as education, justice, finance and film and television production have been developedFor example, the large models of Zhihai series of vertical fields such as Sanle, Luwen and Jinpan developed by Zhejiang University. The vertical domain intelligent pedestal model builds deep learning on top of human expertise in different disciplines, opening up a new model of data-driven and knowledge-led AI researchIt presents the characteristics of "universal pedestal, vertical empowerment".to promote scientific discovery, social governance, judicial innovation, and educational exploration towards a new era of empowerment.
Large language models are also driving a paradigm shift in basic science research. The two basic paradigms in the field of traditional basic science are experimental science, which focuses on data observation, and theoretical science, which focuses on discovering the basic principles of the physical world. However, for high-cost scientific experiments (such as the core of controlled nuclear fusion, the core of which is to make the "tokamak" ring plasma controllable nuclear fusion, etc.), or for scientific research based on assumptions (such as mathematical conjecture proof and synthesis of unknown substances, etc.), it is often difficult to carry out experiments under the traditional scientific research paradigm for a large number of complex tasksThe combination of large language models and expertise in different disciplines has become a powerful tool in the hands of scientists, enabling the generation of hypotheses, the design of experiments, the calculation of results, the interpretation of mechanisms and other steps, reshaping the research paradigm of scientific discovery and engineering innovation, forming a new research pattern of AI+X, and exploring previously untouched knowledge horizons and fields.
The future will be an era of co-evolution between humans and artificial intelligencePeople and man-made objects will go hand in hand, work together and complement each other. There is a cloud in "The Book of Rites: The Mean", "Those who are sincere are not self-contained, so they become things." Becoming oneself, benevolence; Something is known". In the process of synergy between man and man-made objects into their own things,Human beings are the master switches and determiners of the height, breadth, and depth of artificial intelligence, as well as the coordinator of people and artificials.
The author of this article is Wu Fei, director of the Institute of Artificial Intelligence, Zhejiang University
The order of publication of this column is in alphabetical order of scientists' names).