ZTE AI Salon Sidenote: AI can empower thousands of industries, see how ZTE makes efforts?

Cui Li, Chief Development Officer of ZTE

Recently, the "AI 'Xing' Vision Salon" hosted by ZTE was successfully held in Beijing. The salon brought together a number of industry leaders and industry experts to discuss hot topics such as the past and present of AI technology development, the latest progress, application trends, and new paths of industrial integration.

Ms. Cui Li, Chief Development Officer of ZTE, delivered a keynote speech at the salon. She said that in the past year or so, AI technology has undergone rapid iteration, which has aroused widespread attention and collision of different views in the industry. ZTE hopes to share the company's thinking and underlying logic in the process of AI technology follow-up, research practice, and productization through this salon, and discuss how to better deal with related challenges and avoid possible risks and problems.

The Past Life of AI: The Third Spring in 70 Years

Cui Li first led everyone to look back at the history of AI development. She pointed out that the concept of AI can be traced back to the 60s, when scientists tried to create intelligent machines by mimicking the way neurons in the human brain worked. However, due to a lack of in-depth understanding of how the human brain works, this direction quickly bogged down. Subsequently, scientists turned to another way of thinking, that is, by teaching machines what they know about human knowledge and rules, and letting them do the work for humans, this is the birth of the expert system.

However, the development of the expert system has not been without its challenges. Between the 70s and the 90s, AI experienced two cold winters. The first winter was due to the disappointment of the neurons simulating the human brain, and the high expectations led to the subsequent disappointment, and the second was the high expectations and disappointment of the expert system, despite this, scientists have not given up on the exploration of AI. With the continuous advancement of technology, AI has gradually come out of the cold winter and ushered in new development opportunities.

Behind every critical breakthrough in AI technology is the unremitting efforts of scientists and industry leaders for years or even decades. Even ChatGPT is not accidental, but based on previous generations of GPT, and it is also based on the joint efforts of data, algorithm optimization, and computing power enhancement.

Cui Li mentioned that in recent years, with the rise of deep learning technology, AI has been revitalized again. In particular, the proposal of the transformer model solves the big problems of AI in processing natural language, so that AI can enter a new spring. In addition, the introduction of image processing technologies such as DALL-E that combine the advantages of diffusion has further promoted the development of AI technology.

Looking back at the past two AI winters, the most important reason is excessive disappointment caused by high expectations. Cui Li said that as spring comes, we should look at AI more pragmatically and calmly to help it develop healthily.

The Future of AI: Pragmatic and inclusive can help develop for good

When talking about the future trend of AI, Cui Li said that although AI technology has made significant progress, there are still many challenges.

She pointed out that the current challenges are mainly threefold: global risks, digital ethics issues, and resource, efficiency and monetization issues. In terms of global risks, the global economy has been affected by the epidemic and other factors, and the pressure on recovery has increased, and countries have had a chain reaction in the process of solving their own problems. Human exploration and cognition are approaching the critical line, and it is necessary to further expand the boundaries of cognition. On the issue of digital ethics, with the acceleration of digitalization and intelligence, information explosion and information scarcity coexist, and information authenticity has become a major challenge. In terms of resources, efficiency, and monetization, with the deepening of digitalization and networking, the amount of data has surged, and the demand for computing resources has increased, especially after the introduction of AI technology, the demand for computing power and energy has reached an unprecedented height.

At present, the development of AI technology mainly relies on large-scale data and computing power, but with the continuous increase of data, the demand for computing power and energy is also increasing, and the cost of AI technology is also increasing. Therefore, how to maximize output, improve resource utilization efficiency, and accelerate monetization under the constraints of existing resources is a major challenge faced by current AI technologies.

In addition, Cui Li also mentioned the regulatory issue of AI technology. She said that because of the black-box nature of AI technology, all the possibilities that it can create make regulation very complicated. Therefore, how to ensure the compliant use of AI technology and avoid its abuse or misuse is also an urgent problem to be solved.

Despite the challenges, Cui Li remains confident in the future of AI technology. She believes that with the continuous advancement of technology and the continuous expansion of application scenarios, AI technology will play a more important role in the future. In particular, AI technology will play an irreplaceable role in promoting industrial upgrading, improving production efficiency, and improving user experience.

Based on the above-mentioned AI development trends and current problems, Cui Li put forward suggestions on how to help the healthy, benign and benign development of AI in the future: First, seek truth and be pragmatic and forge ahead, she emphasized that generative AI is still in the early stage of development, with rapid technology iteration and confusion in the market, as well as hallucinations, privacy, security and ethics issues. At present, industry self-discipline is particularly important, leading enterprises should first self-restraint, such as setting up a responsibility framework and self-discipline mechanism, and constantly implement these principles in the process of AI model training and product development and application, and at the same time work together to continue to innovate and improve, solve problems, and make AI more healthy. The second is to be inclusive and science and technology for good. She emphasized that AI is a highly interdisciplinary field that requires deep integration and efficient collaboration with various technologies such as big data, computing power and networks, materials technology, embodied intelligence, and energy transition. From the perspective of ethics and security, it is necessary to ensure that the design and application of AI conform to the universal moral standards and values of mankind, minimize bias and discrimination, and continuously optimize the diversity of data and the fairness of algorithms, etc., which need to include the combination of ethics, law, sociology and other humanities and sciences.

The Present of AI: See how ZTE makes a difference

As the key part of the salon, Cui Li introduced ZTE's empowerment methods and processes in the AI industry in detail. She said that ZTE has always regarded AI technology as one of the main channels of the company's long-term strategy, and has made a comprehensive layout in intelligent computing infrastructure and platform technology, large models and applications, and application ecology.

In terms of intelligent computing infrastructure, ZTE provides full-stack and full-scenario intelligent computing products, including IDC, AI servers, high-performance storage, and RoCE switches, covering AI scenarios such as terminals, edges, and data centers. These products not only have the characteristics of high performance, high reliability and high security, but also provide users with flexible resource allocation and pay-as-you-go service models, reducing the cost of users.

In terms of large models and applications, ZTE has adopted the strategy of "1+N+X" for layout. Among them, "1" represents the basic model library, including self-developed basic models and various state-of-the-art basic models open source; "N" stands for multiple domain models, including R&D models, industrial models, communication models, etc.; "X" represents the application of various scenarios, such as network security scenarios and anti-fraud scenarios. Through this strategy, ZTE can provide an upgrade path for large models from general knowledge experts to domain experts to specific scenarios, and improve the performance and application value of large models in specific fields.

In terms of application ecology, ZTE adheres to the concept of openness and decoupling, and works with industry partners to promote the standardization and modularization of AI technology. By providing open API interfaces and tool chain support, ZTE makes it easier for third-party developers to access and use AI technology, thereby promoting the wide application of AI technology in various industries.

Cui Li emphasized that the future development of AI must be based on convergence, and the construction of intelligent computing is a systematic project, which cannot be achieved by pure GPU stacking alone. ZTE will position itself as an end-to-end intelligent computing solution provider, internally identify intelligent computing as the company's long-term strategic main channel, and comprehensively launch the expansion of intelligent computing through the method of "chip + complete machine + assembled R&D + large model", and externally will rely on the company's long-term deep cultivation in the field of communications, assembly R&D, software, hardware and system engineering practices, etc., and work with industry partners to complement each other, open and win-win, and work together to promote industrial innovation and breakthroughs.

Cui Li also introduced some successful cases of ZTE in the process of AI technology implementation at the event. ZTE not only applies the R&D model and telecom model in its own field and builds a data flywheel, but also carries out practice in many fields such as water conservancy, urban lifeline, industry, and park security. ZTE's Nebula series of large models, including language, vision and multimodality, have been deployed in mobile phones, edges and data centers, and the R&D models have been fully promoted within the company, with 13,00+ daily active users, and the overall R&D efficiency has increased by 10%; The Nebula communication model has also helped operators to commercialize autonomous networks in Shandong Mobile and Zhejiang Mobile. ZTE's Nebula programming model has functions such as intelligent requirements analysis, process design, code generation, completion, annotation, and automatic generation of unit tests, and supports a variety of mainstream programming languages, which can help improve the efficiency of AI in the whole process of R&D and provide developers with a full-link intelligent R&D experience.

It is worth mentioning that in the industrial field, ZTE has applied the industrial model to its own Binjiang intelligent manufacturing base, which has achieved quality and efficiency improvement in multiple production scenarios. In the field of communications, ZTE has also launched the Nebula communication model, which applies AI technology to network planning, management, and operation, improving network quality and user experience. In addition, ZTE has established the Cloud AI Open Intelligent Computing Lab to provide multi-vendor GPU interconnection, optimization, and large model compatibility testing and certification, and actively promote the construction of open intelligent computing infrastructure and application ecosystem in China together with industry partners. Cui Li said that in the future, ZTE will continue to uphold the spirit of openness and cooperation, complement each other with industry partners, and work together to promote industrial innovation and breakthroughs.