Generative Artificial Intelligence
Definition
Generative artificial intelligence is a branch of artificial intelligence that involves technologies that generate text, images, sound, video, code, and other content based on algorithms, models, and rules. This technology can generate content with a certain level of logic and coherence by relying on pre-trained multimodal foundation models and using relevant information input by users. Unlike traditional artificial intelligence, generative artificial intelligence can not only process input data but also learn and simulate the inherent laws of things, autonomously creating new content.——Excerpt from “Learning Guidance on the Decision of the 20th National Congress of the Communist Party of China”
Source of the Original Decision
Improve the network governance system. Deepen the reform of the network management system, integrate the functions of network content construction and management, and promote the integrated management of news publicity and online public opinion. Improve the development and management mechanisms of generative artificial intelligence. Strengthen the rule of law in cyberspace, improve the long-term mechanism for network ecological governance, and establish a system for protecting minors online.
Expert Interpretation
XU Haoran Professor, Department of Scientific Socialism, Central Party School (National Academy of Governance)
How to Understand “Generative Artificial Intelligence”
If artificial intelligence is categorized by foundational models, it mainly includes decision-making AI and generative AI. Decision-making AI primarily learns the conditional probability distribution from the data, which indicates the probability of a set of samples belonging to a specific category, allowing machines to judge, analyze, and predict new scenarios. Generative AI, on the other hand, mainly learns the joint probability distribution from the data, which presents a certain probability distribution of vectors composed of multiple variables, allowing machines to summarize existing data and create imitative and synthetic content using techniques such as deep learning.
Generative artificial intelligence is also known as AIGC (AI Generated Content), and the forms of content it can generate are quite diverse, including text, images, audio, and video. Its operating principle mainly involves training deep neural networks through large amounts of data input. In 2014, Ian Goodfellow left OpenAI and returned to Google, where he formed a research team specifically to explore “generative models” aimed at creatively generating elements of the real world such as images, photos, and audio-visual content. This led to the birth of Generative Adversarial Networks (GANs), marking the opening of a new path for innovation in artificial intelligence. GANs primarily generate new data by training two neural networks: one is the generator, which generates new data based on input random noise; the other is the discriminator, which judges whether a data point is real, i.e., whether it comes from the original dataset. Through training these two networks, the generator learns how to produce realistic data, while the discriminator learns how to distinguish between real and generated data. The training characteristic of GANs is adversarial competition, and through continuous training and evaluation, generative artificial intelligence can produce natural outputs such as text, images, and sounds.
Data is the fundamental nutrient for generative artificial intelligence, and its intelligent applications are primarily based on generative systems that utilize large databases and corpora. The value highlights of generative artificial intelligence extend beyond the application itself; it can analyze vast amounts of data for machine learning. It can not only extract data from a static database and corpus but also establish a lasting interactive relationship with the entity asking questions and engaging in conversation. Each interaction with the dialogue subject promotes machine learning and growth. Generative artificial intelligence is not merely intelligent because of its application program; rather, it absorbs enough databases and corpora. It masters not individual information, but panoramic information or long-form continuous text; it learns not individual numbers and sentences, but the overall context of the system.
Algorithms are the logical structure of generative artificial intelligence. Artificial intelligence essentially employs neural network algorithms that decompose problems into numerous nodes (neurons) and allow them to interconnect automatically. The generative pre-trained transformer based on neural network architecture is very adept at utilizing the generative principles of basic neuron connections to learn and imitate any object it interacts with, reflecting this in its expressions through language. The most fundamental principle is known as the “Naive Bayes algorithm,” which can discover subtle correlations in vast continuous corpora and databases. The more interactions there are with generative artificial intelligence, and the more complex the requests made, the more beneficial it is for the generative neural network algorithm.
Computing power is the driving force behind generative artificial intelligence. “Computing power determines algorithms”; the explosive development of generative artificial intelligence is not only due to the machine learning algorithms based on neural networks but also the enhancement of computing power. From the current industry development trends, the parameter volume of large models is growing exponentially, the demand for computing power is surging, the iteration cycle is accelerating, and it presents distinct characteristics of “trillions of parameters, ten thousand card clusters, and multimodal.” Looking ahead to future industrial development, exploring innovative brain-like computing and quantum computing technologies with high energy efficiency opens new pathways for enhancing computing power. For example, brain-like computing chips that mimic the structure of neurons not only build hardware but also employ new technologies such as spiking neural networks to achieve real-time online learning, and can even integrate various different modalities.
Artificial intelligence has evolved from machine automation, primarily reflecting the technological extension of human brain functions, which Marx referred to as manufacturing “organs of the human mind.” The technological trajectory of generative artificial intelligence shapes a new type of relationship of “human-machine symbiosis.” Compared to traditional artificial intelligence, generative artificial intelligence is not just a “complex giant system” that is “self-updating, self-improving, and self-evolving,” but also establishes a “request-response” relationship of interaction between humans and technology. At this point, machines no longer simply and passively execute human commands; instead, they form “a mode of mutual cooperation and enhancement” with humans. In this new relationship, human-machine interaction shifts “from machine-centered to human-centered,” “breaking down the boundaries between people, machines, and information resources, and reshaping the paradigm of generating and using information resources.” The development of generative artificial intelligence may serve as an opportunity to build a human community with a shared future based on intelligent technology.
However, in the face of potential risks posed by generative artificial intelligence, such as intellectual property disputes, data privacy threats, ethical usage challenges, continuation of technological bias, and structural unemployment issues, we must adhere to legal and regulatory guidance, supporting industry organizations, enterprises, educational and research institutions, public cultural institutions, and relevant professional agencies in collaborating on technological innovation, data resource construction, transformative applications, and risk prevention in generative artificial intelligence. Departments such as cybersecurity, development and reform, education, science and technology, industry and information technology, public security, broadcasting and television, and news publishing should strengthen the management of generative artificial intelligence services according to their respective responsibilities. Relevant national authorities should refine and innovate scientific regulatory methods that align with the characteristics of generative artificial intelligence technology and its applications in relevant industries and fields, and develop corresponding classification and grading regulatory rules or guidelines.
Source: Central Party School of the Communist Party of China (National Academy of Governance)