Humanity has entered a new era of comprehensive intelligence, where artificial intelligence (AI) represents a historic opportunity. For institutions, enterprises, and individuals across society, the ability to utilize AI has become a core competency.
1. The Revolutionary Significance of Large Models
The breakthrough growth of generative artificial intelligence (AIGC), particularly the explosion of large models like GPT, is a hallmark event of comprehensive intelligence, having unprecedented impacts on the global economy and society. Its revolutionary nature is reflected in three major characteristics:
Firstly, the large model GPT serves as a “super knowledge base,” concentrating vast amounts of human knowledge and data, including various networks, libraries, museums, and Wikipedia, making it a true “strongest brain.”
Secondly, the large model GPT introduces a new mechanism for human-computer interaction. It communicates in natural language, replacing the previous necessity of understanding programming languages for interaction. This represents a significant breakthrough.
Thirdly, the large model GPT possesses creative capabilities. With over a hundred billion parameters, it generates an “emergent” effect, achieving creativity through reasoning to autonomously generate intelligent content, thus becoming a form of general artificial intelligence.
In the past six months, general artificial intelligence has exhibited explosive growth, manifesting in three phenomena: firstly, a tenfold increase in AI job positions, with many entering the field for employment and entrepreneurship; secondly, the emergence of tens of thousands of “digital nomads” serving AI; and thirdly, a surge in demand for digital humans, with millions of digital laborers entering the market. These figures indicate that large models have reached a historic “turning point.”
2. Paradigm Shift in Knowledge Innovation
Artificial intelligence is transforming the paradigm of knowledge innovation. Historically, human knowledge innovation has gone through four major paradigms: the first paradigm is theoretical thinking, where scientific theories are summarized, such as Newton’s law of universal gravitation and Einstein’s theory of relativity; the second paradigm is scientific experimentation, where new technologies are invented through scientific experiments, as seen with Edison and Faraday’s inventions of electricity and electromagnetism; the third paradigm is experiential simulation, which involves imitating and extending research based on historical and real-world experiences, many contemporary knowledge innovations follow this paradigm.
We are now creating the fourth paradigm, which is the artificial intelligence paradigm, known as “AI for science.” The content of knowledge represents a new demand for humanity, and generative artificial intelligence (AIGC), especially large models like GPT, provides new tools for knowledge content creation. Traditional knowledge content production has been carried out manually by individuals, whereas the contemporary AI paradigm involves collaboration between humans and machines. Initially, the large model provides foundational content creation, followed by human-guided content creation, achieving collaborative innovation. For example, in fashion design, traditional designers could only create one outfit a day, while generative AI can design 1000 outfits in a day, significantly enhancing design efficiency and reducing costs, showcasing the tremendous advantages of collaborative human-machine knowledge content creation.
Artificial intelligence has become a general technology, with AI being a core competitive advantage. Learning and enhancing AI capabilities is the way to victory.
3. Building AI Capabilities
As a new core competency, artificial intelligence (AI) requires the development of three major capabilities: large model prompting capability, small scene fine-tuning capability, and human-machine co-creation capability.
Firstly, large model prompting capability is crucial. The application of large model GPT is highly significant, with 50% of the work in AI content creation being prompting tasks: firstly, demand prompting, which involves clearly articulating needs to the model so it can understand the essence of the requirements. This often requires repeated questioning to achieve satisfactory results. Asking questions is a feedback mechanism that makes the model “smarter”. Therefore, the ability to ask questions is a new skill, and the level of questioning determines the effectiveness of the answers. In this sense, asking questions is more important than answering them. Secondly, the thinking chain prompt is a demonstration of instructions to the model, allowing it to gradually reason and generate outputs through human examples of questions, intermediate reasoning steps, and final answers, enhancing its complex reasoning abilities.
The insights provided by the large model are not standard answers but rather suggestions that are systematic and open-ended, broadening thinking and enabling analogical reasoning. For example, if someone asks, “A child is drawing on a train,” the model does not provide a single answer but generates three images: one showing the child drawing on the train’s platform, another on the roof of the train, and another on the train’s exterior. Such insights are invaluable.
Secondly, the small scene fine-tuning capability is essential. The application of large models typically follows a “large model + small scene” approach. Given the vast differences across industries, industry-specific large models should emerge accordingly. Therefore, the actual application should be “generic large model + industry-specific large model + application scene.” The large model is the result of pre-training on massive data, and fine-tuning with feature data from the application scene plays a decisive role.
In small scene applications, most scenarios can directly employ generative artificial intelligence (AIGC) technology to solve specific problems. For instance, McDonald’s has been operating in China for over a decade, maintaining stable product prices despite rising costs in labor, materials, and energy. To address profitability challenges, McDonald’s applies AIGC technology to audit its operational and management processes annually. As objective conditions continuously change, McDonald’s breaks down all processes into granular components, utilizing AIGC technology to fine-tune data based on changes, implementing ongoing optimization through algorithms, thereby achieving profitability without raising prices.
Thirdly, the human-machine co-creation capability represents a new paradigm of knowledge content innovation, where humans and AI collaboratively create value. Initially, humans propose ideas, and then AI executes the creation, meaning humans provide the ideas while AI does the work. Human-machine co-creation has formed the “80/20 rule”, where 80% of basic intelligence is completed by AI and 20% of core intelligence is guided by humans. Furthermore, humans only need to handle the initial and final stages of work: “1” refers to posing questions and needs, while “100” involves verifying and supplementing the results. Human-machine collaboration sees AI providing basic intelligence, which significantly helps beginners by lowering technical barriers, achieving “technological equality”; at the same time, it greatly enhances professionals by allowing them to focus their energy on improving creative levels.
Digital humans represent a new technology for human-machine co-creation. By combining natural and digital humans, they assist and reinforce each other, achieving shared intelligence. Digital humans will become intelligent assistants, and the new entrepreneurial model involves entrepreneurs co-founding ventures with digital humans. One person can represent a company, where the entrepreneur proposes ideas and the digital human implements the creation, organizing supply chains and marketing online. The new individual entrepreneur is a super-intelligent individual. Humans are carbon-based intelligences, while AI is silicon-based intelligence; the integration of carbon-based and silicon-based intelligences is the inevitable path for future development.
The key to building AI capabilities lies in human transformation and upgrading. AI will not replace humans; for every job that AI eliminates, 2.6 new job positions will be created, fundamentally hinging on enhancing human learning capabilities. Currently, there is a global shortage of AI talent, particularly in three hot areas: AI prompt engineers, AI data trainers, and senior AI creative specialists, all commanding annual salaries exceeding one million dollars. Existing professionals must evolve into composite talents, proficient in both technical expertise and AI technology, as AI capabilities have become the most essential competency.
The penetration of artificial intelligence spans numerous industries and has become today’s most powerful productivity force. According to research by the McKinsey Institute, the future scale of the intelligent economy will be 300 times that of the current industrial economy, underscoring the limitless prospects for AI development.
Winning in AI means winning the future!
Author Profile: Qian Zhixin, a digital economy expert, senior engineer, professor at Nanjing University, doctoral supervisor, founding president of the Jiangsu Enterprise Development Engineering Association, and former secretary and director of the Jiangsu Provincial Development and Reform Commission. He has long engaged in macroeconomics and enterprise management, primarily researching digital economy, metaverse, new economy, new business models, and industrial finance, with nearly 30 published works on these topics. He teaches the latest technology and economic knowledge through his WeChat public account “Professor Qian’s Classroom” and Weibo account “Professor Qian’s Sayings.”
