Editor's note: Artificial intelligence is undoubtedly the tech buzzword of 2023. As a strategic emerging technology, artificial intelligence (AI) has increasingly become an important driving force for technological innovation, industrial upgrading and productivity improvement. What do you think of the AI boom? How to deal with the relationship between development and governance? How to promote the application and orderly development of related industries? To this end, this edition has launched a special plan for the benefit of readers.
Reporter Cui Shuang.
Artificial intelligence is undoubtedly the hottest digital tech keyword in 2023. On the one hand, countries have accelerated the layout of the artificial intelligence industry. Recently, Russia unveiled a new national strategic plan for the development of artificial intelligence, and Singapore released the national artificial intelligence strategy 20。On the other hand, the governance system for AI is also becoming more and more perfect. The U.S. introduced the Executive Order on Secure, Secure, and Trustworthy AI, and the European Union reached an agreement on the Artificial Intelligence Act. At the first Global AI Security Summit held at the end of 2023, 28 countries, including China, the United States, and the United Kingdom, and the European Union signed the Bletchley Declaration, which aims to promote global governance of AI.
It can be seen that in the era of digital economy, the layout of artificial intelligence has become the consensus of all countries. However, how to lay out and how to develop still requires the wisdom of the world, scientific research institutes, and the industry. What will happen to AI technology in 2024? How can AI be better regulated in the face of rapid technological development?
A key year for industrial applications.
Under the traditional credit business model, after the bank lends money to some small farms, the staff needs to do post-loan inspection and monitoring every other day, which is time-consuming and laborious. In some areas of the west, it may even take two or three days for a single inspection to be conducted. Geng Bo, head of the R&D center of the Agricultural Bank of China, said at the first artificial intelligence application conference held recently.
The high operating costs have been greatly reduced after the large-scale application of artificial intelligence technology. "We installed some cameras on small farms, collected data through these cameras, and then used artificial intelligence to do the counting. Geng Bo introduced, "Not only counting, artificial intelligence can also do discrimination." For example, whether this cow is Lao Wang's or Lao Li's. Geng Bo introduced that the application of artificial intelligence has improved the efficiency of post-loan management, reduced the cost of post-loan management, and enabled some remote areas to enjoy smart financial services.
This is just the tip of the iceberg when artificial intelligence technology is applied to the ground. "2024 will be a critical year for the implementation of artificial intelligence technology in industrial applications. Zuo Min, vice president of Beijing Technology and Business University, judged. In his view, with the rapid progress of large models in the past year, the application potential of artificial intelligence technology has been widely recognized. At the same time, there is a realization that building larger and larger models in the field of AI is not necessarily the end goal. Translating AI technologies into concrete, viable solutions and products is also critical to driving the AI industry and realizing business value.
What is the direction of artificial intelligence, especially large models, that should be worth paying attention to in the coming period? Du Zhengping, executive vice president of China Mobile Communications Federation, gave his own answer: the engineering mechanism of intelligent emergence, the new technology business model, the new intelligent organization and technical architecture.
Intelligence emergence, which refers to the emergence of new behaviors and functions of artificial intelligence that are unrelated to the initial training target. The industry generally believes that the emergence of intelligence is a manifestation of artificial intelligence moving to a higher stage. However, scientists are still unclear about the mechanism and engineering path of intelligence emergence. Du Zhengping mentioned that the Internet company Meta recently said that it plans to spend billions of dollars to buy Nvidia's computer chips for artificial intelligence research and related projects. Meta CEO Mark Zuckerberg also revealed that according to the company's AI future roadmap, the company will build a large-scale computing infrastructure. By the end of 2024, this infrastructure will include 350,000 NVIDIA H100 graphics cards. In Du Zhengping's view, whether a "violent" method like Meta will definitely promote the emergence of intelligence in artificial intelligence, as well as the mechanism and engineering path of intelligent emergence, still need to be explored.
At the same time, Du Zhengping believes that although the development of large models is in full swing, it still needs to be explored by the whole industry in what aspects it will develop new technology business models. "In the era of model intelligence, how to build a new intelligent organization and technical architecture is also the focus of our attention. Du Zhengping said.
China's economy has shifted from a stage of rapid growth to a stage of high-quality development, and is in a period of tackling key problems in transforming the mode of development, optimizing the economic structure, and transforming and enhancing the driving force. Nie Linhai, chairman of the China E-commerce Merchants Association Alliance, said, "We must deeply grasp the development characteristics of the new generation of artificial intelligence, accelerate the integration and development of artificial intelligence and related industries, and provide new momentum for high-quality development." ”
Promote the deepening of governance rules and practices.
Generative AI technology, represented by large models, has developed rapidly and has initially revealed its disruptive impact on the industry. At the same time, the issue of AI security governance has attracted unprecedented attention. Countries, international organizations, academic and scientific research institutions, industrial organizations, and technology companies have joined it. "The rules and practices of global AI security governance will be further deepened and realized. Talking about the development trend of global artificial intelligence security governance, Hu Naying, senior business director of the Artificial Intelligence Research Center of the China Academy of Information and Communications Technology, said.
According to the research results released by the Security Governance Committee of the Chinese Artificial Intelligence Industry Development Alliance, as artificial intelligence accelerates the empowerment of thousands of industries, the security governance issues of artificial intelligence are expanding from traditional technical security issues such as privacy protection, data security, algorithm black box, and model vulnerabilities to underlying security issues applied to biological, chemical, nuclear energy, critical infrastructure and other fields.
In terms of artificial intelligence security governance solutions, Hu Naying introduced that the world continues to strengthen artificial intelligence security governance. In October 2023, the United Nations established a high-level AI advisory body. The agency aims to regulate AI and address the risks that AI may pose. In the process of strengthening AI security governance, risk orientation has become a common governance path for AI in all countries. For example, China's Interim Measures for the Administration of Generative AI Services actively address the risks arising from generative AI in terms of training data, algorithm models, generated content, and personal information rights and interests. The European Union's Artificial Intelligence Act regulates AI at different levels and regulates risks. The United Nations AI Advisory Body also said in its interim report that the development of an AI risk management framework will be the focus of the next step.
In terms of specific measures, in the context of rapid technological development, the use of artificial intelligence to govern artificial intelligence has become the focus of industry exploration. Among them, value alignment methods such as extensible supervision, explainability research, and model evaluation are the main exploration directions of the industry. For example, in May 2023, AI startup Anthropic announced that it would evaluate and train the values of the company's large model through a large model called "Constitutional AI". Artificial intelligence research company OpenAI followed suit and announced the establishment of a super alignment team, investing 20% of computing resources to build an artificial intelligence that solves the alignment problem in 4 years. (Technology**).