A new K-line cycle may be starting, and the future of artificial intelligence is approaching.

Artificial Intelligence (AI) is a key driving force in the new round of technological revolution and industrial transformation.

From ChatGPT to DeepSeek, generative artificial intelligence accelerates and empowers thousands of industries, becoming an important engine for nurturing and developing new productive forces. This has led representatives and members attending the National Two Sessions to reflect, "the world may enter a new cycle of prosperity".

Kondratieff cycle, an economic cycle theory, often takes the breakthrough of core technology as the starting point. It refers to the economic fluctuations over a period of 50 to 60 years, during which the economy experiences stages of recovery, prosperity, recession, and depression.

In the Fourier Intelligent Robotics Laboratory, engineers are training and debugging general robots, with every detail and key joint undergoing countless tests. / Wen Wei Po journalist Yuan Jing photo

In the eyes of the representatives and committee members, the dawn of a new round of Kangbo cycle is emerging. Taking humanoid robots as an example, by the end of last year, there were nearly 100 humanoid robot enterprises nationwide, with more than 30 enterprises announcing the start of commercial production this year. At this critical moment, how to seize the opportunity and take advantage of the situation?

Members of the committee believe that it is necessary to proactively address the current practical challenges in the field of AI, namely, the situation where technology outpaces standards, products outpace certification, and applications outpace regulations. This can be achieved through systematic design, forward-looking planning, institutional support, promoting accelerated application landing, facilitating human-machine interaction and virtual-real integration, and accelerating the move towards general AI, thereby redefining the labor force and production tools in the intelligent era, and promoting the development of new productive forces.

System Design

From the laboratory to the CCTV Spring Festival Gala stage, humanoid robot mass production is imminent. However, many representatives and committee members have expressed concerns in their research notes: collaboration is difficult.

Phenomenal applications emerge one after another, but representatives and committee members at the grassroots level keenly capture the other side of the thriving technological ecosystem - the lack of communication protocols and interface standards, leading to difficulties in collaboration between hardware, algorithms, and scenarios. From motion control precision to environmental perception capabilities, there is a lack of quantifiable metrics.

"Industrial data sharing is difficult, and there is low-level redundant construction of large models in vertical areas." Zhang Fan, a member of the National People's Congress and Minister of Innovation and Technology of China Electrical Equipment, believes that AI new technologies have unearthed many new opportunities in the consumer end, but have not yet spawned high-value application scenarios at the manufacturing end.

The picture shows the humanoid robot 'Qinglong' undergoing tests of identifying and grasping objects of different shapes and weights. / Wen Wei Po journalist Yuan Jing photo

National Committee member of the Chinese People's Political Consultative Conference and Director of the Shanghai Economic and Information Committee Zhang Ying also found that some 'pseudo-intelligent' products are entering the market. Some companies are using the concept of intelligence to package and hype ordinary products in order to gain favor with capital. The reason behind this is that existing testing institutions mostly focus on traditional industrial robots, lacking specialized evaluation of humanoid robots with artificial intelligence.

The rapidly growing emerging industry urgently needs regulation. Recently, the world's first international standard for elderly care robots, led by China, was released. The standard elaborates on the functions and performance classification of elderly care robots, guiding practitioners to return to rationality and address the practical problems of technological bottlenecks, cost constraints, and differences in market demand.

"Accelerate the formulation of industry classification system, communication protocols and other basic standards." Zhang Ying believes that standardizing data collection and usage standards, clarifying human-computer interaction safety standards and ethical guidelines, can only improve industrial scale efficiency through software and hardware collaboration.

Some small and medium-sized enterprises also reflected that there is a lack of training and verification resources, and a shortage of high-cost simulation testing environments and physical "training schools", making it difficult for them to complete reliability verification and restricting the speed of technological iteration.

Recently, the first heterogeneous humanoid robot training ground in the country was launched in Shanghai, and it is expected to accumulate 10 million pieces of data this year. This training ground, built by the National Local Joint Construction Humanoid Robot Innovation Center, is expected to break through bottlenecks such as difficult standardization, and inability to migrate and reuse data across platforms.

This further inspired Zhang Ying. "Can we build a national-level evaluation and certification 'examination room' to promote humanoid robots 'certified employment'?" She called for strengthening the evaluation and certification mechanism at the national level, conducting comprehensive scientific evaluations to ensure the quality and safety of products on the market.

Strategic Layout

The importance of computing power to the AI industry goes without saying. From quickly processing user requests to real-time analysis and decision-making, only high computing power can ensure the system's responsiveness and accuracy.

During the investigation, Zhang Fan found that computing resources are scattered across the country and urgently need overall planning. It is suggested to give full play to the leading role of enterprises in the industry, establish AI joint innovation centers, and jointly build, share, and use large industry vertical models.

Some representatives believe that the sharing of computing power should not be limited to enterprises. Research has found that the bottleneck of computing power is making it increasingly difficult to connect 'industry-education-research'. On one side, there are continuously iterating new technologies in the market, while on the other side, there are relatively closed curriculum settings and learning platforms. It is suggested to break through the computing power dilemma of university scientific research through the integration of industry and education, and at the same time, support various social entities to collaboratively promote the construction of AI data science and high-quality corpora.

Every detail and key joint in the picture must be tested countless times. / Wen Wei Po reporter Yuan Jing photo

Top domestic universities have taken action, each showing their unique strengths. Some are strengthening cooperation with enterprises, some are leasing resources, and some are directly purchasing computing power services. For example, Shanghai Jiao Tong University recently completed the localized deployment of the entire DeepSeek series of models, becoming one of the first domestic universities to achieve the domestic deployment of billion-level large models. This marks a new stage in the construction of AI infrastructure in universities.

"Exploring a new paradigm of scientific research under the blessing of AI." Ding Kuiling, a National People's Congress representative and president of Shanghai Jiao Tong University, explained that relying on the Shanghai Jiao Tong University Kunpeng Ascending Science and Education Innovation Center jointly built with enterprises, it will promote the deep integration of the DeepSeek series models with curriculum teaching, scientific research innovation, etc. For example: using new technologies such as AI to complete knowledge imparting, allowing students to have more space for full interaction and discussion, stimulating creative thinking.

In Ding Kuiling's view, the cultivation of future talents should especially emphasize the ability of underlying innovation and cross-innovation. For universities, this means that all disciplines such as science, engineering, agriculture, medicine, and literature should actively embrace AI. The university has set up an AI comprehensive reform promotion group to promote the application of AI in different disciplines and fields through organized scientific research and teaching. Recently, the university is planning to set up a 10 million yuan "AI+" seed fund to solicit high-quality projects from the whole school, and use project funding as a baton to guide more teachers and students in traditional disciplines to integrate AI technology and open up new possibilities for scientific research.

"I believe that the next DeepSeek-level innovation breakthrough is all around us," Ding Kuiling is full of expectations.

System escort

The explosion of large models, the entry of general AI into the fast lane of R&D, and the development of new technologies have also brought about the other side of the coin - AI rumors, AI infringement, AI fraud and other phenomena occur from time to time.

According to data from the China Judgment Document Network, from 2021 to 2024, the national courts have tried a total of 1659 AI-related dispute cases, the vast majority of which are related to generative AI.

Looking at the world, it has become a consensus to set "basic guardrails" for AI development. At the beginning of last year, the European Union introduced the world's first comprehensive AI regulatory law, taking the lead in establishing a unified and standardized regulatory framework to ensure that the development of technology applications follows the principles of fairness, transparency and trustworthiness.

Lv Hongbing, a member of the National Committee of the Chinese People's Political Consultative Conference (CPPCC) and a partner of Grandall Law Firm (Shanghai), found that the draft AI law has been included in the legislative work plan of the State Council. "Compared with traditional legislation, AI legislation faces uncertainties in social relations and unpredictable risks brought about by the iteration of the technology industry." In his view, it is extremely difficult to formulate a comprehensive AI law, so it is advisable to choose generative AI, which is the most concerning, direct and realistic incision of society, to promote "small, fast, and flexible" legislation as soon as possible, and to introduce administrative regulations as soon as possible.

The picture shows the country's first humanoid robot innovation center jointly built by the state and the local government in Zhangjiang, Pudong. / Photo by Wen Wei Po reporter Yuan Jing

Lü Hongbing believes that this administrative regulation should clarify the rights, obligations, functions, and responsibilities of technology developers, service providers, users, regulators, and the general public, categorizing them separately. For example: technology developers should rigorously screen and classify the data used to train large language models, remove inappropriate textual elements, ensure the legality and healthiness of the data sources; for existing models, apply techniques such as data forgetting to eliminate inappropriate content output; introduce a manual review mechanism to review content marked as sensitive or prohibited by automated systems; establish open research, community cooperation, and clue reporting mechanisms to optimize the screening system for large language models.

Members of the delegation and committee believe that legislation should effectively respond to the key core issue of who owns the copyright in the process of AI-generated content. Lv Hongbing suggested that, in accordance with the original legislative intention of better leveraging the role of data elements, ownership should be played down, and usage rights should be strengthened, with usage principles clearly defined. For example, when it comes to commercial use, technology developers are required to pay remuneration and negotiate with rights holders to resolve the issue.

Managing the present and planning for the long term, AI legislation is timely.

(Article source: Shangguan News)

Source: East Money

Author: Shangguan News