What is the future trend of AI?Industry leaders say so

Mondo Technology Updated on 2024-01-29

China-Singapore Jingwei, December 15 (Wan Keyi) What will be the development trend of AI technology, which is currently showing infinite potential?At the Tencent Technology Hi Tech Day and 2023 Digital Open Things Conference held on December 14, a group of industry leaders gave their views on the theme of "Intelligence Emerges, Numbers Open Everything".

Huang Chenxia, general manager of Tencent News Operations, believes that the technological progress in 2023 is unusual and can be called the starting point of a new round of industrial revolution. She asked three questions about AI technology: "What obstacles do we need to overcome on the road to AGI?""How will the big model be embedded in the industry?".and "what product possibilities will be brought about by the unprecedented capabilities brought about by AI?"

In response to everyone's ambivalence in the face of new technologies, Huang Chenxia believes that anxiety is understandable, but it should not be rash to act, but to think more deeply. As for the solution of complex problems, Huang Chenxia said that although there are many challenges, in the process of discussing these problems, the emerging information will lead to countless futures.

Huang Chao, Chairman of Kezhi Group and Founder of Digital Kaiwu, said that 2023 is an important year of change in the field of global digital technology, and it is also a year of acceleration for China's digital economy to move forward steadily. Against the backdrop of technological innovation, digitalization is becoming a transformative force in China's economy and society.

In the AI era, GPUs and generalized AI chips have gradually replaced CPUs as the core of computing power development. How can China break the game in GPU R&D and manufacturing?Liao Qiwei, director of the Expert Committee of the State-owned Assets Supervision and Administration Commission of Science and Technology Power Think Tank and professor of the Chinese Academy of Sciences, believes that computing power in a broad sense determines the real competitiveness of a country in the future.

Regarding the disruption brought by AI, Wang Sheng, partner of Innova Angel**, said that a new technological paradigm has its time point from generation to maturity, and next year may be the first year of application. He further explained that the combination of AI and industry is not just beginning, and the large model only broadens the possibility of integration, but this progress will still be gradual if the business model does not change.

In this regard, Fang Han, chairman and CEO of Kunlun Wanwei, added: "Incremental innovation will land relatively quickly on the B side, and disruptive innovation will shine on the C side. He believes that this wave of AI will definitely give birth to new C-end giants from small enterprises.

Regarding the concept of AI Native, which has been widely discussed recently, Fang Han believes that AI Native is a completely false proposition. Wang Sheng believes that AI is a tool and an extension of past applications, and has limited relevance to native.

Talking about which application scenarios are the first to be changed by AI, Chen Xi, vice president of investment at Quzhong Capital, divided AI into three layers: the basic model layer, the middle layer and the application layer. "The basic model layer is the one with the highest barriers, and the application layer has the most development opportunities. Chen Xi said.

In response to the question of why large models did not first appear in China, Yang Dong, distinguished professor of "Changjiang Scholars" of the Ministry of Education and dean of the Institute of Interdisciplinary Sciences of Chinese University, pointed out that in the past mobile Internet era, there was serious data riskism in China. Why can't large models be produced and developed early?The main reason is not in computing power, nor in algorithms, but in the lack of interconnection of data.

Talking about the risks of security and privacy, how to manage generated content and synthetic content, and the illusion of large models, Jiang Chunyu, director of the Big Data and Blockchain Department of the Cloud University Institute of the China Academy of Information and Communications Technology, said that it is necessary to build the protection capability of data security and privacy throughout the life cycle, covering the entire training process.

In response to the problem of illusion, in Jiang Chunyu's view, "the authenticity and accuracy of different fields can be constrained by some rules, content generation requirements, monitoring mechanisms and authenticity assessments, and harmful problems can be automatically detected through content identification and filtering + manual review", but at present, these fields in China are in a blank stage and need to be improved urgently.

According to Jiang Chunyu, the China Academy of Information and Communications Technology is writing a book on artificial intelligence data governance, with the aim of establishing a system of methods and rules in this field.

In addition to the security improvement, Jiang Chunyu also emphasized the value of data quality: without good data, the model's capabilities will definitely be lacking. "The current situation is that the domestic IT development path is to pollute first and then treat, so the quality is generally biased, and there needs to be changes in the evaluation dimensions related to data quality, and the process and engineering capabilities to improve data quality need to be strengthened. ”

At a time when the new technology of AI large model shows great potential, how the entire industry will be changed may be the most concerned issue for ordinary enterprises at present. Talking about how enterprises use AI models, Sun Bin, president and COO of Emotibot, suggested that external purchase and joint construction should be adopted, and if it is not a leading enterprise in the industry or an enterprise that does not have enough IT development strength, it is not recommended to do self-research and development.

Talking about whether the large model is a necessary option for the digital transformation of the manufacturing industry, Yan Tongzhu, chairman of the Beijing Informatization and Industrialization Integration Service Alliance, said that the significance of the large model for different enterprises is also different. "The high-end manufacturing industry itself has relatively intensive knowledge requirements, and the large model plays an optimization role and has great value. However, for the traditional manufacturing industry, the amount of data is not large, and the threshold of the large model is relatively high, and the input-output ratio is not so clear. ”

Bing Jinyou, chief expert of Tencent Cloud's intelligent manufacturing, believes that there are two problems in the acceptance of large models by industrial enterprises, one is the cost problem, and the cost of model training is highThe second is the illusion problem, either the economic output in the industry needs to be 100% correct, and the large model is difficult to achieve.

Although the large model cannot be applied to all industrial scenarios at this stage, there have been many examples of the scenarios it is good at. Bing Jinyou said that the use of large models in customer service, process management, human resources, marketing, content output and design has been widely practiced.

Yan Tongzhu also put forward another perspective on the significance of large models to industry, he said that industrial knowledge and industrial data have been lost in many places because of the resignation of talents, and have not been transformed into enterprise knowledge assets, so many innovations are reinventing the wheel, but the emergence of large models can more effectively transform empirical data into usable knowledge assets. (Zhongxin Jingwei app).

For more exciting content, please pay attention to the official WeChat of JWVIEW***

Related Pages