How does generative artificial intelligence change the world? Zhang Yaqin, Russell and Ceng Yi said this.
On March 27th, at the sub-forum of "AIGC (Generative Artificial Intelligence) Changing the World" of the 2024 annual meeting of Boao Forum for Asia, China Academy of Engineering, Professor Tsinghua University, Zhang Yaqin, Dean of Tsinghua University Intelligent Industry Research Institute (AIR), Stuart Russell, author of "Artificial Intelligence: Modern Methods", professor of computer science department of University of California, Berkeley, researcher of Institute of Automation of China Academy of Sciences, expert of United Nations artificial intelligence senior advisory body, and many other experts.
Experts believe that at present, artificial intelligence is still the "fifth kilometer" of the 42-kilometer marathon, and everyone has a chance. The future must be the integration of artificial intelligence and machine intelligence. Of course, the risk is great, so we must face up to the risk. If it is not guided and controlled, this risk will turn into a risk of survival.
Zhang Yaqin: Artificial intelligence is still the "fifth kilometer" of the 42 km marathon.
At the scene of the sub-forum, Zhang Yaqin talked about the relationship between the big model and the generative AI. He thought that one is technology and the other is ability, because with the pre-trained, super-large-scale and self-monitoring big model, the generative AI (artificial intelligence) can have such high energy.
Regarding the trend of the big model, Zhang Yaqin said, firstly, it is multi-modal, and secondly, the big model of artificial intelligence is moving towards autonomous intelligence, which can make goals, plan tasks, make paths, complete schemes and code itself. The third is edge intelligence. Now the big model is mainly in the cloud, and it will go to mobile phones, PCs, robots, cars and so on in the future. "The most important thing is that the whole world is now moving towards the physical world and the biological world. Therefore, generative AI can not only carry out speech recognition, face recognition and image recognition, but also create and generate, not only text, images and videos, but also codes, drugs, mathematical formulas and physical equations. Therefore, generative AI will bring about tremendous technological and paradigm changes."
However, Zhang Yaqin believes that there are some risks in the change. "Of course, I am an optimist. I think human beings have two kinds of wisdom, one is the wisdom of inventing new technologies, and the other is the wisdom of guiding the development of technologies. Therefore, we can manage while developing, but we must think of governance while innovating and developing."
In view of the future development of AI, Zhang Yaqin believes that elements, computing power, algorithms, data and talents are all important, but in the long run, talents are the most important.
Zhang Yaqin believes that at present, artificial intelligence is still the "fifth kilometer" of the 42km marathon, and everyone has a chance. The future must be the integration of artificial intelligence and machine intelligence. Of course, the risk is great, so we must face up to the risk. If it is not guided and controlled, this risk will turn into a risk of survival.
Stuart Russell: Learning AI system is the direction of efforts.
In the sub-forum, Stuart Russell put forward his views on the development path of AI and the competition and cooperation between China and the United States.
Stuart Russell said that the next generation of GPT5, people may find that its training materials are ten times as much as before, but it also means the end of data, and there is no more data in the universe. At this time, if there is no real universal intelligence, human beings must look for different directions. "I don’t think synthetic data can change this problem, because synthetic data is not a real number, and it will not make the system more capable."
What are the possible new directions? Stuart Russell mentioned that by combining probability theory with programming language, we can use expression language to learn new concepts. So far, this technology has not been developed like the big language model, and more human intervention is needed to create a structural model. "Let AI system learn and produce products like human beings. At the same time, human beings can understand its internal operating mechanism, which is the direction of our efforts."
When talking about the concurrence of AI between China and the United States, Stuart Russell said that competition is benign and the statement of zero-sum game is very wrong. If AI is safe and can benefit human beings, even more powerful than human beings, many people will benefit from it. From this perspective, competition is meaningless. The core issue is how human beings can coexist with technology safely.
Ceng Yi: Super intelligent artificial intelligence should be super altruistic.
In the sub-forum, Ceng Yi talked about the future development direction of AI and his views on AI competition between China and the United States. Ceng Yi said that the current AI seems to be close to the so-called AIGC, but in fact there is no relationship between the two.
Ceng Yi believes that what we must do is to bring the real artificial intelligence to the world and serve the world, and at the same time, to study the possible risks. "Artificial intelligence evolves very fast, and the speed of human learning from history and nature is too slow. If human beings don’t reflect, it is not artificial intelligence but human beings who bring catastrophic risks to human beings, because human beings have not learned how to get along with ecology in the evolution of hundreds of millions of years. When there is intelligence with a higher level of intelligence than human beings in the ecology, that is the greatest risk for human beings."
However, Ceng Yi is still optimistic about the future. He believes that the artificial intelligence of super intelligence should be super altruistic, but the question is whether the artificial intelligence constructed by human beings is super altruistic now, and whether human beings can embed the law of nature evolution into artificial intelligence, so that it can become an extension of natural evolution and pay attention to other life, instead of bringing disastrous consequences to other life and the world. This is the direction that artificial intelligence technology research and future human evolution need to evolve together.
In view of the competition of AI between China and the United States, Ceng Yi believes that neither the United States nor China can really lead the future artificial intelligence. Artificial intelligence is a basic and enabling technology, and the world is big enough for China, the United States and more countries to find opportunities in building infrastructure and applications. The more critical question is how the whole world can unite to construct safer, more reliable and more credible artificial intelligence for human prosperity, which is the goal of more unity in the world.
Beijing News Shell Finance reporter Pan Yichun
Editor Yue Caizhou
Proofread Sissi













































