2023 is the first year of generative AI, and the world has entered an AI-driven production revolution. The advent of generative AI technology has been hailed as the "most revolutionary" technological progress, and future industrial development is a key node to seize the global innovation highland, reconstruct the global innovation pattern, and reshape the global economic pattern.
The development of Chinese artificial intelligence
From the perspective of policy layout, since 2019, mainland AI-related policies have always closely followed technology and industrial development, and have gone through four stages: extensive pilot, framework construction, industrial development, and scenario-based implementation, effectively promoting artificial intelligence from emerging technologies to standardized applications. In 2023, generative AI will bring major breakthroughs and new hope to the field of artificial intelligence, and the policy layout around generative AI will also be rapidly rolled out, from consolidating the foundation of data resources and computing power, to quickly guiding the standardization of generative AI, to industrial support, forming a powerful combination.
From the perspective of the capital market, the concept of AIGC is hot, and nearly forty percent of the investment events point to the newly established AIGC company in 2023. At the same time, the original AI technology track remained active, among which machine learning was the most widely used and most prominent, followed by computer vision and NLP. It is worth noting that AIGC-related financing accounted for 28% of all financing events in the AI industry in 2023, and penetrated into all tracks except robots, becoming the biggest keyword for AI financing of the year.
From the perspective of technological development, the large model library catalyzes the industrialization of AI and supports the innovation and development of various AI subdivision technology tracks. Large models are good at generalized reasoning capabilities, and small models still have a market with high maturity and high cost performance. Large and small models are the core landing point of current industrial applications, and in the future, multimodal models and MO architectures will jointly expand the industrial space of large models.
From the perspective of industrial development, Chinese artificial intelligence enterprises are actively seizing the opportunity of application exploration and seizing the opportunity in the wave of new technologies, and generative artificial intelligence is further accelerating the penetration into the content industry. The state is also actively carrying out the construction of intelligent computing centers at the infrastructure level, promoting the establishment of artificial intelligence data standards, and encouraging the development of open-source datasets.
In 2023, the scale of Chinese's artificial intelligence industry has reached 213.7 billion yuan, and it is expected that the scale of Chinese's artificial intelligence industry will reach 811 billion yuan by 2028, with a five-year compound growth rate of 30.6%. Compared with the original scale of the artificial intelligence industry when the large model has not yet had the ability to explode, iResearch predicts that the industrial added value brought by the large model may account for 32.9% in 2028.
The development process of Chinese artificial intelligence
Generative AI Industry Insights: In 2023, the world will enter an AI-driven production revolution, and generative technology is an opportunity of the times. The technical architecture of the pre-trained large model optimizes the training and understanding output of the underlying model under the multi-modal path, so that decision-making AI and generative AI can jointly build the development of eight major industries. The volume of domestic large models has increased, and the industry large models have entered a period of explosive growth, among which medical care and finance are typical landing fields.
From the perspective of model modality, the text modality of generative AI applications has reached a high level of application maturity, and code, speech, and images have a commercial foundation.
From the perspective of commercial applications, the state has adopted a "registration system" for the online supervision of large models, and more than 40 large models have been put into operation with "certificates". Starting from the B-end scenario, it is necessary to gradually penetrate and polish, open up the business logic to achieve the closed-loop application of more scenarios, and integrate the star continuation curve to empower. The C-end scenario needs to meet the hardware device conditions and large model capability adaptation from the supply side, and there will be a stepped energy explosion after the software and hardware ecology matures.
Insights at the edge of the AI industry: The emergence of large models has accelerated the evolution of AI capabilities from the cloud to multiple intelligences at the edge. At present, the large model is tried at the edge from the two dimensions of computing power management and scenario optimization, partially replacing and taking over the computing power scheduling authority and capability of the original cloud computing center, greatly reducing the time loss caused by cloud transmission. At the same time, large models can replace the original small models used for prediction, decision-making, discrimination, generation and other tasks at the edge, improve the generalization ability and use effect of scenarios, improve RO1, and are comprehensively upgrading and reconstructing the autonomous driving technology stack.
The integration of AI and terminals is accelerating, and large models on the terminal side are first implemented in scenarios such as mobile phones and smart cockpits. From a hardware perspective, AI reshaping the operating system is the key to unlocking the potential of large models.
The development process of Chinese artificial intelligence
From a societal perspective, the main risk that deserves attention is the possible impact of AI technology on user mentality, user privacy, security ethics and so on.
From the perspective of enterprise application, the defects of artificial intelligence technology itself have a more obvious impact on enterprise application, and any link of artificial intelligence framework, data, algorithms, and models may bring weaknesses to the system.
Based on the above discussion on the risks of AI development, the future development of the AI industry needs to be continuously improved and guided from three levels: technical research, industry standards and regulations, and legal supervision.
In terms of technical research, improve the accuracy and transparency of algorithms and prevent bias and unfairness.
In terms of industry standards, unified norms and ethical guidelines should be established to ensure that the application of AI is in line with ethical and social values.
In terms of legal supervision, it is necessary to formulate and revise relevant laws and regulations to protect personal privacy and prevent abuse and infringement of rights, so as to ensure the stable, high-quality and benign development of the mainland's artificial intelligence industry.
✦
end
✦
Source: iResearch
Editor in charge: Wu Nana
Comment|Lv Shuang