The report points out that there are still many difficulties in the development of generative AI applications in China, and with the increasingly fierce global competition in science and technology, some countries have implemented strict export controls on artificial intelligence chips in China for the sake of safeguarding their own interests, which has formed a certain obstacle to the development of artificial intelligence-related industries. In recent years, China has successively launched a number of computing chip products with independent and controllable capabilities, but there is still a certain gap compared with the global leading level in terms of computing performance and versatility. Second, there is an imbalance in the distribution of computing infrastructure. The development of generative AI relies on expensive high-end chips, and due to the demand for massive computing power, the energy consumption of data centers is extremely high, which makes the industry mainly concentrated in more economically developed regions. From the perspective of geographical distribution, the computing resources in the eastern coastal areas are relatively concentrated, while the central and western regions are relatively insufficient. From the perspective of enterprise scale, large enterprises are usually able to purchase high-quality computing resources due to their abundant funds, while small and medium-sized enterprises are unable to bear the high cost of hardware, resulting in a lack of computing resources. In addition, with the rapid growth of model applications, the problem of computing power management has become more and more prominent. The demand for large-scale model training and inference has increased dramatically, and the demand for computing power has also increased. As the scale of chips increases, the complexity of O&M also increases significantly. Hardware equipment will inevitably fail during use, and with the expansion of computing power, the probability of failure will also increase. If the problem of computing power management cannot be effectively solved, a series of problems such as training cost and efficiency will arise, which will seriously hinder the progress of enterprises in the field of generative AI applications.
High-quality datasets are relatively scarce
Algorithm optimization: Generative AI requires complex algorithms to support its operation and learning. However, there are still some problems with current algorithms when dealing with complex natural language and image generation tasks. For example, when generating text, grammatical errors, logical incophies, or inaccurate content may occur. In addition, the optimization of algorithms requires a lot of computing resources and time, which is a huge challenge for researchers and developers.
Data quality: The performance of generative AI is highly dependent on the quality of the training data. If the data is noisy, biased, or incomplete, it will directly affect the generation of the model. Obtaining high-quality, large-scale, and representative datasets is a difficult and time-consuming process. At the same time, labeling and cleaning of data requires a lot of work and expertise.
Figure: There are many difficulties in the development of generative AI applications
Industry application scenarios need to be expanded urgently
Although generative AI has achieved remarkable results in some fields, it still faces difficulties in its promotion and application in other fields. Different fields have their own characteristics and needs, which need to be customized for specific scenarios. In addition, some traditional industries may be less receptive to new technologies, which will also affect the scope of applications of generative AI.
Improved user experience: The ultimate goal of generative AI is to provide users with high-quality, useful, and expected generated results. However, current models still fall short in terms of the accuracy and relevance of the content generated, which can lead to a poor user experience. How to better understand user needs and improve the quality and personalization of generated content is the key to improving user experience.
In addition, generative AI involves multi-domain knowledge, which requires compound talents who understand algorithms, machine learning, data processing, ethics and law, etc., and the current shortage of such professionals is relatively scarce, and the training system is not perfect, which limits the development and application of technology. Issues such as ethics and law are also worthy of attention. AI-generated content may be similar to existing copyrighted works, giving rise to copyright infringement disputes, and at the same time, it is difficult to ensure the copyright legitimacy of data sources and the copyright distribution of generated works is also difficult to define, which brings new challenges to intellectual property protection, and requires new legal frameworks and judicial interpretations to clarify relevant rights and responsibilities.
Related:
Analysis of the Application Development Report of Generative Artificial Intelligence (1)
Analysis of the Application Development Report of Generative Artificial Intelligence (2)
Analysis of the Application Development Report of Generative Artificial Intelligence (3)
Analysis of the Application Development Report of Generative Artificial Intelligence (4)
Analysis of the Application Development Report of Generative Artificial Intelligence (5)
Analysis of the Application Development Report of Generative Artificial Intelligence (6)
Analysis of the Application Development Report of Generative Artificial Intelligence (7)
Analysis of the Application Development Report of Generative Artificial Intelligence (8)
Analysis of the Application Development Report of Generative Artificial Intelligence (9)
