The "soul" of artificial intelligence technology is not to "make machines like people", but to realize intelligent decision-making through digital intelligence.
OpenAI released a new model, GPT-4o, which takes multi-modal real-time interaction capabilities to a further step. With the rapid development of artificial intelligence technology, the impact on society is becoming increasingly significant, and the application scenarios are constantly expanding. At the same time, its development is also facing challenges in terms of safety and reliability, and the quiet change of employment structure has put forward new requirements for talent training.
On May 14, the second session of the "Peking University Guanghua Scholars Salon" was held at the Guanghua School of Management of Peking University, with the theme of "Opportunities and Challenges of Artificial Intelligence". Weng Xi, professor of the Department of Applied Economics of Guanghua School of Management, Peking University and convener of Guanghua Artificial Intelligence and Social Sciences Interdisciplinary Horizontal Development Platform, and Meng Juanjuan, professor of the Department of Applied Economics of Guanghua School of Management, Peking University and director of the Center for Behavioral Science and Policy Intervention of Peking University, shared the theme.
In the sharing with the theme of "The "Soul" of Artificial Intelligence Technology", Weng Xi said that in order to look at the development of artificial intelligence technology from the source, it is necessary to clarify the traditional philosophical questions of "who am I", "where do I come from" and "where are I going", and fall into the field of artificial intelligence: what is the "soul" of artificial intelligence technology, who is empowered by artificial intelligence technology, and where is the key to the future development of artificial intelligence technology.
Weng Xi pointed out that due to the influence of thinking, people often pay more attention to the scientific content of artificial intelligence, resulting in the inability to see the core content of artificial intelligence as a technology. The "soul" of artificial intelligence technology is not to "make machines like people", but to realize intelligent decision-making through digital intelligence. At the same time, artificial intelligence is a general-purpose technology, and the future will definitely empower the general public to make intelligent decisions. The key to the development of AI technology in the future lies in the transformation of education.
"Artificial intelligence as a production tool for the masses, to some extent, will reduce some inequalities and disparities. In the short term, many users may be somewhat reluctant to use AI technology, but with the further popularization of technology and products, the pain points will be slowly resolved in the process of development. Weng Xi said.
In his view, with the development of actions such as "artificial intelligence +" and "data element ×", the gap between the development level of domestic artificial intelligence technology and foreign countries is rapidly narrowing. China's main advantage in developing AI lies in its strong ability to apply AI in specific vertical industries and fields, such as finance and marketing. In addition, Chinese society is more inclusive of the development and application of AI, which also contributes to the wider application of AI among the public.
Weng Xi said that at present, many fields in China are applying artificial intelligence for digital and intelligent decision-making, and have achieved certain results, and benchmark cases in key areas such as industrial manufacturing, financial services, and transportation have been promoted. In addition, the three major elements of artificial intelligence are computing power, algorithms and data, and the potential of data is gradually being released as the mainland includes data as a new factor of production. Therefore, the mainland has great potential and space to realize the expansion and application of artificial intelligence technology in the future, and encouraging the application of artificial intelligence in the industry is the key to further promoting the development of the artificial intelligence industry.
"At present, although the wave of artificial intelligence has surged, the future development of artificial intelligence is also encountering bottlenecks, such as computing power bottlenecks and data shortages. In the long run, demand creates supply, and domestic computing power is expected to develop rapidly under the stimulation of all parties. In terms of data shortage, the National Data Bureau has been promoting the construction of AI large model training datasets, and high-quality annotated datasets need to be conquered in the future. Weng Xi said.
More critically, the current cognitive bottlenecks and talent shortage hurdles need to be addressed in contrast to the above-mentioned constraints. Weng Xi proposed that in order to better adapt to the development of artificial intelligence technology in the future, it is necessary to promote educational reform, and the emerging professional courses should respond to the needs of reality, rather than simply changing the name of traditional courses, and at the same time focus on cultivating compound talents who are familiar with computer and management knowledge and vertical industries. This requires the education system to keep up with the times and adjust the curriculum content in a timely manner through continuous dialogue with key enterprises or institutions in the value chain.
Finally, Weng Xi also pointed out that the integration of artificial intelligence represented by large language models into education will inevitably cause students to rely on the use of it, which is not a problem of the technology itself. In order to better stimulate students' innovation ability and creative nature, it is recommended to give full play to the advantages of the mainland system, integrate artificial intelligence into education at all ages as soon as possible, and guide students to develop the ability and habit of using new production tools healthily.
Meng Juanjuan shared the theme of "Human-Computer Interaction: The Impact of Artificial Intelligence on Human Behavior and Social Equity". In her view, artificial intelligence technology is overwhelming, we can't refuse, such an era background, how human beings face artificial intelligence technology, is a question worth exploring.
"We need to understand the impact of AI on human society from the perspective of human-computer interaction." Meng Juan said. In this regard, she first asked a series of questions:
What are the impacts of artificial intelligence on employment, business organizations, information dissemination, and social equity? Answering these macro questions can be approached from a micro perspective, i.e., understanding how humans and AI will interact? The Difference Between Humans and AI: Substitution or Assistance? Human Attitudes Toward AI: Algorithmic Aversion or Algorithmic Preference? Human-AI Interactions: What New Behavioral Problems Are Generated by Algorithms?
Meng Juan pointed out that the adoption of artificial intelligence technology is largely determined by four factors: technology, cost, value and ethics. Specifically, in terms of technology, it is the maturity of technology and the availability of data; In terms of cost, it is the cost of new technology, the cost of personnel replacement, and the cost of management; In terms of value, it is the business value brought by the new technology; At the ethical level, it includes government regulation and social acceptance.
In her opinion, the timeline of AI development is divided into pre-GPT and post-GPT eras. In the pre-GPT era, one of the most classic conclusions is that repetitive, non-cognitive labor will be replaced by AI, while cognitive social interaction, labor related to emotion and emotional intelligence will not be easily replaced. This leads to a higher probability of substitution for low-income people. However, after the birth of GPT, GPT has done a lot of cognition-related things, which may lead to a higher probability of high-income people being replaced by GPT.
It can be seen that the substitution of skills in the post-GPT era is more significant, such as programming, mathematics, comprehension, reading, expression, writing and other abilities in cognitive abilities will be quickly replaced by GPT. However, in some areas, such as scientific research, critical thinking, active learning, and other skills, the degree of substitution is relatively low. Meng Juan said.
So, under what circumstances can't AI completely replace humans? In Meng Juanjuan's view, AI cannot completely replace humans when humans have better interpersonal skills, when humans have private information, and when humans have different goals and preferences than algorithms.
First, when humans have better interpersonal skills, AI can be responsible for data analysis to provide conclusions, while humans can be responsible for communicating information more efficiently. Second, when humans have some private information that artificial intelligence cannot grasp, the value of human decision-making cannot be replaced, because most of the training of artificial intelligence is based on public data. Third, when humans have different goals and preferences than algorithms, humans need to be the final decision-makers. Because everyone's goals are different in different situations, AI may not be able to make decisions that best meet the individual's goals at that time.
She emphasized that AI technology will bring great efficiency improvements, but human attitudes towards AI have influenced the adoption of AI technology. Research shows that the pre-GPT era is usually manifested as "algorithm aversion", while the post-GPT era may reflect a certain "algorithm preference" because artificial intelligence is more human-like. From the perspective of the impact of AI on social equity, the substitution of labor in the labor market will exacerbate inequality to some extent, but enabling or assisting decision-making will reduce inequality by leveling the performance of workers. The latter conclusion shows that in both the pre-GPT and post-GPT eras, AI empowering humans with assisted decision-making will close the gap between workers of different skills. For the commodity market, artificial intelligence provides better goods and services for a wider range of people at a lower cost, and personalized recommendation algorithms can avoid the winner-take-all of merchants to a certain extent, giving small and medium-sized enterprises more opportunities; The adoption of AI can also reduce human bias and promote social equity due to its neutrality.
But at the same time, Meng Juan also pointed out that in the interaction between artificial intelligence and humans, there are also problems such as privacy paradox, ideological polarization, digital addiction, cognitive degradation, and reduced behavioral diversity, which deserve our attention.
At present, artificial intelligence is undoubtedly a hot spot of concern for the whole society, and enterprises and individuals in it are facing unprecedented development opportunities. Guanghua School of Management of Peking University is actively drawing on artificial intelligence methods to study social science problems, exploring the huge impact of artificial intelligence on society and new development opportunities from the perspective of social science, exploring the boundaries of technology and the solution of social problems, and exploring the in-depth application and cutting-edge practice of AI empowering all industries.