background
Artificial intelligence is the current cutting-edge technology in the field of artificial intelligence and has attracted much attention. In 2022, OpenAI released ChatGPT, a model that made an important breakthrough at the application level, with the number of users exceeding 100 million in just two months, becoming the fastest growing consumer-grade application in history. Under this trend, many technology companies around the world have increased their investment in artificial intelligence research and development, continuously launched important results, promoted the innovation and commercialization of artificial intelligence, and also driven the rapid development of related industrial chains.
Report release
Thanks to the guidance of the Internet Society of China and the China Software Industry Association, the Tianjin Artificial Intelligence Society, Zhiding Technology and Zhiding Think Tank jointly released the "2023 Global Artificial Intelligence Industry Research Report". From a global perspective, the report comprehensively sorts out the industry overview, infrastructure, algorithm models, scenario applications, opportunities and challenges of artificial intelligence, comprehensively shows the development of the artificial intelligence industry, and provides reference for government departments, industry practitioners, educators and the public.
1. Artificial intelligence concept and development stage
Generative AI refers to a new production method that uses artificial intelligence technology to generate content, including professionally generated content (PGC) and user-generated content (UGC) after automatic generation content.
AI relies on massive training data and large-scale pre-trained models to automatically generate text, audio, images, video, and cross-modal information.
Since OpenAI released ChatGPT in 2022, AI has become rapidly popular, and many technology companies have successively launched AI models, products, and infrastructure and services.
2. The driving force for the development of the artificial intelligence industry
The rapid growth of the AI industry is driven by several key drivers:
- The scale of global data continues to expand, providing a large number of data resources for AI models to train. IDC predicts that the global data scale will reach 175 zettabytes by 2025.
- The introduction of high-performance AI chips provides important computing support for large-scale pre-trained models. At present, NVIDIA Tensor Core GPU chips are widely used in the training of large-scale artificial intelligence models.
- The AI computing cluster provides large-scale computing power resources for AI training, improving the efficiency of computing power utilization and data processing capabilities. NVIDIA DGX SuperPOD, Baidu Intelligent Cloud High Performance Computing Cluster EHC, Tencent's new generation High Performance Computing Cluster HCC, etc. are typical AI computing clusters.
- AI cloud services provide platform support for the development of AI models to reduce development costs and product development cycles. Amazon SageMaker and Baidu Flypaddle EasyDL are typical AI cloud service platforms.
3. Overview of AI model development
3.1 Language classes generate mainstream models
OpenAI has released a series of language generative pre-training models, including GPT-1, GPT-2, GPT-3, ChatGPT, and the latest GPT-4.
These models are based on the Transformer architecture and are continuously iteratively optimized, achieving important breakthroughs in language generation. Among them, GPT-4 has more powerful multimodal capabilities, supports graphic and text multimodal input and generates response text, and has excellent response capabilities.
3.2 Image class generation mainstream model
An important research result in image generation is the Diffusion Model, which produces high-quality images through forward and reverse processes. Compared with traditional adversarial training methods, Diffusion Model can achieve high-quality image generation without large amounts of data.
Diffusion Model has obvious advantages in image generation, especially in terms of data requirements and generation results.
4. Application of artificial intelligence scenarios
Artificial intelligence has a wide range of applications in different fields, mainly including text generation, image generation, audio generation and video generation. Here is an overview of the main applications in these areas:
- Text generation: Including applications such as content continuation, text style migration, summary/title generation, and whole text generation, it is expected to have broader applications in the field of personalized text generation and real-time text interaction.
- Image generation: Includes image attribute editing, image partial generation, and end-to-end image generation. Image editing tools have been widely used, and creative image generation and functional image generation also show great room for development.
- Audio generation: mainly includes speech synthesis and music creation. Speech synthesis includes the fields of specific speech (TTS) and speech cloning, and music creation involves lyrics, composition, and arrangement.
- Video generation: Includes video attribute editing, video auto editing, and video part generation. Video attribute editing has been widely used in the field of video creation, and the technology of automatic video editing and video part generation is still in the experimental stage.
- Digital human: Digital human generation is divided into digital human video generation and digital human real-time interaction. The former is currently widely used in the field, and the latter is mainly used in real-time interactive functions such as intelligent customer service.
5. AI opportunities and challenges
5.1 Employment Impact and New Career Opportunities
AI presents both opportunities and challenges for the job market. On the one hand, intelligent automation of artificial intelligence can improve work efficiency and reduce operating costs, and may replace some traditional jobs. According to Goldman Sachs, artificial intelligence has the ability to replace a quarter of jobs. On the other hand, AI will also create new career opportunities, such as emerging positions such as "Prompt Engineer."
5.2 Copyright Distribution of Product
The issue of copyright attribution of AI works is still controversial. Since the law only recognizes the copyright of persons with subject identity, it is currently considered that the developer and owner of AI software have subject identity, so the copyright of AI works is mainly assigned to software owners and users.