In 2023, the economic field coexists with "risks" and "opportunities". The challenges in traditional industries still exist, and there are still risks in real estate, local debt, traditional manufacturing, and small and medium financial institutions...

In 2024, we need to patiently wait for the industry to release opportunities and focus on resolving risks.

Looking forward to 2024, we will witness an unprecedented innovation revolution. With the rapid development of technology, breakthrough progress in fields such as artificial intelligence, machine learning, big data, and the Internet of Things is changing our production and lifestyle at an unprecedented speed. China's scientific and industrial circles actively respond to the national innovation-driven development strategy, closely focus on the global development trend of science and technology, and promote China's scientific and technological undertakings to a higher level.

In 2024, the field of science and information technology will show a magnificent innovation picture. The internal driving force of consumption is accumulating, public expectations are stable and good, and the technological revolution, industrial upgrading, and artificial intelligence also contain many opportunities. The foundation of "stability" is further strengthened, and the momentum of "progress" is rapidly gathering. The situation of seeking progress in stability is becoming clearer and clearer. The world situation is still changing, how to maintain its own advantages and strategic determination, only by focusing on high-quality development, forging new quality of productivity, can we take the initiative, take the wind of the new industrialization, and continue to lead development with innovation.

Advertisement

The stars do not disappoint those who are in a hurry, and I hope that the New Year is better than the old year. This must not just be a wish.

In 2024, which trends are worth paying attention to, and which surprises are worth waiting for?

I. AI industry year refreshes the application "progress bar"

In 2024, it will become the "AI industry year". The development trend of artificial intelligence is mainly reflected in the popularization and commercialization of large models. More and more innovative application scenarios and product forms will emerge, and the "progress bar" of AI applications will be continuously refreshed. The acceleration of commercialization, investment growth, and the improvement of popularity will further promote the development of the entire AI industry.Academician of the Chinese Academy of Engineering, Wu Hequan, stated that our country is experiencing a crucial moment in economic development. Intelligent economic activities represented by generative artificial intelligence are growing against the trend, becoming the foundation for our country to promote sustained economic growth, build technological innovation, and industrial upgrading. They have also become the new engine to support the modernization of the economic system.

Firstly, large models have become an important development direction for artificial intelligence. Bill Gates predicted that an innovation wave of artificial intelligence will emerge in 2024, marking the next qualitative leap. On March 16, 2023, Baidu officially launched the generative AI product "Wenxin Yiyan," based on Baidu's new generation of large language models, becoming the first domestic large model manufacturer to release a product comparable to ChatGPT. The debut of "Wenxin Yiyan" also opened the curtain on the domestic "hundred model battle." Subsequently, tech companies such as Alibaba, Huawei, Tencent, JD.com, iFLYTEK, 360, and ByteDance have all released large models. Research institutes and startups are also competing on the same stage. The Beijing Academy of Artificial Intelligence released the "WuDao" large model, the Shanghai Artificial Intelligence Laboratory launched the "ShuSheng" large model, and Peng Cheng Laboratory developed the "Peng Cheng · Brain Sea" large model. Startups such as BaiChuan Intelligence founded by Wang Xiaochuan and ZhiPu AI incubated by the Computer Science Department of Tsinghua University are rising stars, attracting industry attention with their excellent self-developed capabilities.

Zhou Hongyi, founder of 360 Group, proposed that large models will become the "standard configuration" of digital systems, ubiquitous, similar to personal computers of the past. Open-source large models will usher in an outbreak, and small models will run on more terminals, such as mobile phones, smart homes, etc. He predicted that 2024 will be the year of large model application scenarios, with the emergence of killer applications, especially the robot revolution and multimodal applications based on large models.

In 2024, various large models will gradually differentiate into different echelons with different levels of technological maturity, application scenarios, market demands, and innovation speeds. The development of technology and changes in the market will promote the continuous "evolution" and "reorganization" of these echelons.

Secondly, the trend of "from empowering thousands of industries to entering thousands of households" will become more apparent in 2024. China is a major network country and will also be a major AI application country.

China is expected to achieve leapfrog development in the field of artificial intelligence. Although China does not have a significant advantage in basic large models, it has a vast market and user base, making it easier to obtain a large amount of user data and application scenarios. This can provide rich data resources for the application of large models and pave the way for unique applications in the field. Currently, China's large models have begun to be applied in finance, healthcare, entertainment, education, transportation, and other fields, indicating that large model technology in China is gradually maturing and has produced actual benefits in multiple industries. "Application" is highly anticipated by domestic large models and is also the key for AI technology to get rid of value anxiety.

Thirdly, the application of large models and AIGC (Artificial Intelligence Generated Content) will innovate the industrial ecosystem model. The application of large models and AIGC not only improves production efficiency and quality but also promotes business model innovation, facilitates cross-industry integration, enhances data-driven decision-making capabilities, achieves technological democratization, and promotes the development of new technologies, thereby comprehensively innovating the industrial ecosystem model. It promotes the implementation of advanced working models such as human-computer collaboration and intelligent decision-making within enterprises, providing easy-to-use and powerful tools and services for small and medium-sized enterprises, and is expected to further promote the prosperity of China's physical enterprises and software industry ecosystems in the field of the intelligent economy.

Looking forward to 2024, AIGC technology will undoubtedly become an important engine to promote the transformation of China's science and technology industry, improve enterprise efficiency, and promote the high-quality development of the real economy.

II. Manufacturing industry transformation and upgrading relies on "model" power

AI large models will affect the development pattern of the manufacturing industry. The universality and generalization of AI large models, as well as the new development paradigm based on "pre-training + fine-tuning," will empower the manufacturing industry in various aspects such as research and development, design, production processes, operation and maintenance, quality control, sales and customer service, and organizational collaboration.Among them, entering the production process, the most core control systems such as PLCs, MES, SCADA, etc., enhancing the intelligence of production technology, is a key sign of the application of AI large models in the manufacturing industry.

Siemens and Microsoft announced a partnership in April 2023 to promote the next generation of automation technology transformation based on GPT, and jointly develop PLC code generation tools, integrating AI large models into the control process.

Antelope Industrial Large Model takes the universal capabilities of iFLYTEK Spark Cognitive Large Model as the core technology base, and is built in combination with the actual needs of industrial scenes, with five major core capabilities of industrial text generation, industrial knowledge Q&A, industrial understanding calculation, industrial code generation, and industrial multimodality. It can continuously evolve from massive data and large-scale knowledge, achieving a complete process closed loop from proposing, planning to solving problems, just like penetrating into the "capillaries" of the manufacturing industry, helping enterprises to build a "sustainable evolution" brain.

Based on the following aspects, the application of large models in the manufacturing industry will inevitably accelerate in 2024. First, the improvement of computing power, including the emergence of more powerful GPUs and dedicated AI chips, makes the training and deployment of large models more efficient and cost-effective, which will promote their application in the manufacturing industry. Second, the manufacturing industry has generated a large amount of data, including production data, equipment maintenance records, and supply chain information. These data provide a foundation for the training and application of large models, enabling the models to better understand and optimize the production process. Third, the global digital transformation has promoted the application of intelligent solutions. Manufacturing enterprises hope to improve production efficiency, reduce costs, enhance product quality, and enhance competitiveness through large models. Fourth, the customization and modularization capabilities of large models make them better adapted to the specific needs of different manufacturing enterprises. Enterprises can adjust and optimize large models according to their own production processes and business objectives.

With the emergence of ChatGPT and its popularity, the generality and generalization ability reflected by GPT have also triggered new thinking and attempts in the field of autonomous driving for model construction. How large models are specifically implemented in application scenarios has become a continuous exploration issue for car companies - currently, smart cockpits and intelligent driving have become the two main focus points for the implementation of large models on the "vehicle end".

Tesla has trained a large amount of data for its AI models to achieve advanced autonomous driving functions. Google's Waymo has also trained a large amount of data for its AI models to improve the safety and reliability of its autonomous driving cars. Baidu is developing autonomous driving technology and using AI models to enhance the intelligence level of its cars. Haowu Zhixing introduces the method of large models in various links such as perception, cognition, and decision-making, increasing the generality and generalization ability of the processing capacity of the autonomous driving system, making it smarter as a whole. Xiaopeng Automobile has also introduced the capabilities of large models in the latest released XNet2.0 perception architecture, increasing its perception level generalization ability. Specifically, Xiaopeng XNet2.0 can read the text information on traffic signs, have a sense of time, and understand the semantic characteristics of traffic elements in different cities.

In 2024, car companies will promote the application of large models from three aspects.

First, continue to invest in the research and development of autonomous driving technology, using large models to improve the perception ability, decision-making ability, and safety of vehicles. This includes improving sensor data processing, enhancing the prediction accuracy of algorithms, and enhancing the vehicle's adaptability to complex traffic environments.

Second, car companies will further integrate AI large models into smart cockpits to provide a more personalized and intelligent user experience. This may include more advanced voice recognition, emotional recognition, gesture control, and other functions.Thirdly, production efficiency is improved. In the field of manufacturing, automotive companies will use large models to optimize the production process, such as predictive maintenance, supply chain management, quality control, etc., to improve production efficiency and reduce costs.

III. A More Solid Foundation for Information Technology Application

The global deployment of 5G networks provides better connectivity for the Internet of Things (IoT) devices, promoting the development of smart homes, smart cities, and telemedicine, greatly enhancing the speed and efficiency of data transmission. In 2023, China's data centers transitioned from a period of rapid growth to a stable development phase. However, under the influence of new policies and favorable factors such as AIGC, China's data center market has ushered in a new round of opportunities.

In October 2023, the Ministry of Industry and Information Technology and five other departments jointly issued the "Action Plan for High-Quality Development of Computing Infrastructure." The plan proposes quantitative development targets to be achieved by 2025 in four aspects: computing power, carrying capacity, storage capacity, and application empowerment, guiding the high-quality development of computing infrastructure. It also promotes the coordinated construction of data center planning and construction, the integration of data networks, and the construction of the computing power industry chain. The plan proposes that by 2025, in terms of computing power, the scale of computing power will exceed 300EFLOPS (EFLOPS refers to the number of floating-point operations per second), with intelligent computing power accounting for 35%, and the balanced and coordinated development of computing power in the eastern and western regions.

At present, more than 20 smart computing centers have been built or are under construction nationwide, of which 11 are government-led smart computing centers. Looking at the layout related to "East Data West Computing," there are 8 located in the eight major hubs, accounting for more than 70%. With the digital economy becoming an important engine for economic growth, China's core digital economy industries continue to grow. In 2024, the advantages of new infrastructure construction will be further consolidated.

Firstly, the construction and deployment of 5G networks will continue to accelerate. It is expected that by 2024, 5G networks will cover a wider area, providing higher data transmission speeds and lower latency. This will promote the application and development of technologies such as the Internet of Things (IoT), Augmented Reality (AR), Virtual Reality (VR), and autonomous driving.

Secondly, the upgrade and expansion of optical networks (including optical fibers and cables) will continue to advance to support the high-bandwidth requirements of 5G networks and data centers. The improvement of optical network technology will enhance data transmission efficiency, reduce energy consumption, and lower network operation costs.

Thirdly, with the advancement of Artificial Intelligence (AI) and Machine Learning technologies, network infrastructure will become more intelligent. Predictive maintenance, automated fault detection, and adaptive network management will become the norm, improving the reliability and efficiency of the network.

Fourthly, to reduce data transmission latency, edge computing will continue to develop, pushing data processing and analysis to the edge of the network. This will help real-time applications, such as autonomous vehicles and telemedicine.Fifth, generative AI poses higher demands on data center computing power, and heterogeneous architectures such as CPU+GPU/FPGA/ASIC will become the main form of computing power in data centers in 2024 and beyond.

In 2024, computing power infrastructure will face greater challenges in terms of industrial collaboration, operational security, and green development. First, the computing power industry chain needs to develop collaboratively. Only by achieving optimal allocation as a whole can the best performance be achieved. Second, operational security capabilities and standards will be further enhanced. Third, computing power infrastructure will become more deeply green. New technologies such as liquid cooling and energy storage will be promoted by the scale of intelligent computing business, driving computing power towards a green and low-carbon direction for rapid development.

IV. The Domestic Robot is Just the Beginning

A Chinese research team at Stanford University recently launched a low-cost artificial intelligence robot called MobileALOHA. The eye-catching feature of this robot is its high mobility and flexibility, capable of performing a series of complex tasks, such as frying shrimp, washing pots, and opening cabinet doors. MobileALOHA is based on an operating system that enables the robot to learn and master skills through remote operation and imitation learning principles.

It is reported that this robot can learn a new skill after only 50 demonstrations, with a success rate of 90%, and the cost is relatively low, with the entire set of equipment costing about $32,000. It combines the dual-arm manipulation capabilities of the ALOHA system with the mobility of the mobile base, characterized by its ability to perform complex mobile tasks, easy operation, and cost-effectiveness. The research team has improved its imitation learning performance by combining the data collected by MobileALOHA with the existing static ALOHA dataset for training. The greatest value of MobileALOHA lies in the progress of its physical operation capabilities, showing its potential in repetitive and fine operation tasks. However, the system also has limitations, such as a larger footprint and difficulty for the dual arms to reach lower cabinets. The research team has open-sourced the MobileALOHA project on Github, including code, hardware structure, and data. The team leader said that human control is only temporary, and they are already researching how to bridge the gap between human control and robot self-control. 2024 will be the year of robots, and the domestic robot is just the beginning.

On July 6, 2023, at the 2023 World Artificial Intelligence Conference, Musk proposed: "At some point in the future, the ratio of robots to humans may exceed 1:1." The robots Musk referred to are AI robots (embodied intelligence). AI robots are intelligent agents with bodies and support physical interaction. In the future, AI robots will not only include humanoid robots but also wheeled robots, force control robots, visual robots, welding robots, grinding robots, home service robots, etc. AI robots will not only replace a large number of human labor but also replace some specialized equipment. In 2024, Tesla humanoid robots, UBTECH, and Zhiyuan robots are expected to enter Tesla and BYD (BYD) factories for commercial verification.

As 2024 has just begun, many companies have started to announce new humanoid robots. Similar to 2023, humanoid robots will continue to be the focus of the industry in 2024, with companies continuing to innovate and launch new models or upgrade existing robots. However, the actual deployment or application of more mature humanoid robots is still relatively rare.

V. Data Governance Advances into Deep Waters

The "Opinions of the CPC Central Committee and the State Council on Constructing a More Perfect Market-oriented Allocation System and Mechanism for Factors" and the "Data Elements × Three-Year Action Plan (2024-2026)" have been released, further promoting the implementation of data strategy. By promoting the construction of basic data systems, activating the potential of data elements, and promoting the development of digital China and the digital economy.The acceleration of digital transformation has increasingly highlighted the importance of data governance, which is gradually entering the "deep water area."

Firstly, the integration of data governance with business operations is deepening. Enterprises' demand for data governance is no longer confined to the technical level but focuses more on the combination of data governance with business strategy. Enterprises hope to improve data quality through data governance, ensure the reliability of data, and use data as a driving force to optimize business processes, enhance decision-making efficiency, and further strengthen competitiveness.

Secondly, data security and privacy protection are becoming more important. Frequent data security incidents have made enterprises and organizations pay more attention to data governance. At the same time, countries' requirements for data privacy and compliance are becoming increasingly strict, and enterprises need to consider compliance issues when conducting data governance.

Thirdly, the objects of governance are becoming more diversified. The objects of data governance are no longer limited to traditional structured data but have expanded to unstructured data, IoT data, etc. The governance methods and tools for these new types of data are still developing, bringing new challenges to data governance.

Fourthly, the degree of data assetization and marketization is deepening. Data assetization is an important direction for data governance. Through data transactions and circulation, it is possible to achieve effective allocation of data resources and maximize the value of data while protecting data rights and interests.

Embracing open source is a "required course."

The digital world is shaped by developers, and every line of code is a force to change the world. Open source has shifted the code development model from individual, centralized, and closed to collaborative research, creation, and use, gradually becoming the dominant mode of global software technology and industry innovation. The consensus that "software defines the future world, and open source determines the future of software" has been formed.

Sun Wenlong, the chairman of the Open Atomic Open Source Foundation, said that currently, 97% of global software developers and 99% of enterprises use open source software. Embracing open source is no longer a "multiple-choice question" but a "required course" related to survival and long-term development.

Firstly, open source software will continue to maintain its growth momentum because it provides cost-effectiveness, flexibility, and customizability. Enterprises and organizations are increasingly inclined to use open source software to reduce costs, improve innovation speed, and enhance community support. Open source software will pay more attention to cross-platform compatibility and integration capabilities to meet users' needs for seamless software use across different systems and platforms. The application of open source software in the field of artificial intelligence and machine learning is also increasing. For example, deep learning frameworks such as TensorFlow and PyTorch will continue to develop, offering more features and better performance. At the same time, with the popularity of cloud computing and containerization, open source software is also widely used. For example, container orchestration tools such as Kubernetes and cloud platform projects such as OpenStack will continue to develop to meet the growing demand.Secondly, the degree of corporate participation in open-source projects will continue to increase. For instance, Microsoft continued to increase its contributions to the Linux kernel in 2023, becoming one of the largest contributors. In addition, many other technology companies, such as Google, Amazon, Intel, and others, are actively contributing code to various open-source projects. Companies are not only participating in existing open-source projects but also initiating and leading new ones. For example, Google has launched several new open-source projects, including the open-sourcing of its internal machine learning platform, TensorFlow. This can drive the development of features that meet their own needs and enhance the reliability and security of the software. An increasing number of companies encourage their employees to participate in open-source projects and even open-source their own projects within the company. For example, Huawei launched its own open-source operating system, HarmonyOS, and encourages external developers to contribute. Cooperation between companies on open-source projects is increasing. For example, Microsoft and Amazon's cooperation on the open-source project Apache Hudi, jointly promoting the development of data management and analysis technology.

In 2024, it is expected that the open-source community will continue to expand, attracting more developers, users, and contributors to jointly promote the development and improvement of software. At the same time, the open-source culture will continue to spread globally, encouraging more people to participate in open-source projects, share knowledge and resources, and promote technological progress.

Thirdly, foundations play an indispensable role in promoting the healthy development of open-source software and communities, fostering technological innovation, and building a cooperative and win-win open-source ecosystem. By providing comprehensive support and services, foundations help open-source projects to develop more focused and efficiently, thus contributing to the development of the entire software industry and the digital economy.

In June 2020, the OpenAtom Open Source Foundation was registered and established by the Ministry of Civil Affairs. The OpenAtom Open Source Foundation is a non-profit charitable organization and the first open-source foundation in China. Three years of exploration and effort have yielded initial results. As of December 2023, the foundation has 86 fund donors, 10 open-source contributors, and nearly 2000 co-builders around the development of open-source projects. In 2024, the foundation will provide a more suitable environment for the development of open-source and accumulate more industry strength.

In conclusion,

2023 has passed with a magical figure, and the "shock" is still lingering. The emergence of large language models like ChatGPT has made AI start to "become human" and "surpass human". Drones and autonomous driving technology have shown great potential in fields such as aerial photography, logistics, and agriculture. Autonomous driving cars have started commercial operations in specific scenarios, and humanoid robots have taken the stage...

2024 has quietly arrived, and the dream continues...