#Fast Moving Eight Talks##2023 Zhongguancun Forum#Silicon Valley has a famous inventor and futurist Ray Kurzweil, more than a decade ago, he proposed a concept of technological singularity in the book "The Singularity is Approaching", which can be roughly summarized as the development of technology becomes faster and faster, after a certain singularity, human beings will no longer be able to understand the development of technology. The advent of ChatGPT predates the discussion of the arrival of the "technological singularity." In just half a year, artificial intelligence technology has developed and iterated rapidly.
Big models were the first to make waves. On May 28, at the "Artificial Intelligence Big Model Development Forum", one of the parallel forums of the 2023 Zhongguancun Forum, Zhao Zhiyun, director of the China Institute of Science and Technology Information, released the "Chinese Intelligent Large Model Map Research Report" showing that 79 large models with a scale of more than 1 billion parameters in China have been released, mainly concentrated in Beijing and Guangdong, including 38 large models in Beijing and 20 large models in Guangdong.
When ChatGPT "breaks the circle" from the field of science and technology, when AI technology moves from the laboratory to the street, it has become a social consensus that human beings are entering the era of artificial intelligence. So what kind of era is the future intelligent era, but also aroused people's thinking. At the Artificial Intelligence Big Model Development Forum, many industry insiders gave their own answers.
Zhou Bowen, Hui Yan Chair Professor and Tenured Professor of Electronic Engineering at Tsinghua University:
AI is increasingly like the "System 2" model of human thinking
Compared to the previous two waves, artificial intelligence has changed in nature. First, this artificial intelligence is driven by generative artificial intelligence, and the artificial intelligence in the past is usually done by "understanding", and does not open up the real AI road, because the judgment of understanding is vague, and generative judgment can force the improvement of understanding; Second, in the past, artificial intelligence attracted widespread attention by defeating humans, but behind generative artificial intelligence, people and AI are collaborative and interactive, and the value of this breakthrough is very huge, and the boundary of its application and the value of the scene have increased by an order of magnitude.
Based on these two points, I think that the future AI may become more and more like "System 2" AI. Human thinking modes include system 1 and system 2, the former is mainly responsible for fast thinking and relying on intuition to make rapid decisions; System 2 is slow thinking and requires a lot of computation and reasoning.
In the past few years, artificial intelligence similar to face recognition has been the logic of System 1, but the development of artificial intelligence, especially after the emergence of GPT-4, AI actually has the ability to systematize, and AI based on System 2 is a clearer direction. This brings us to think about how humans can further interact with AI as it becomes more powerful and able to do more detailed work.
Baidu Chief Technology Officer Wang Haifeng:
Artificial intelligence is becoming the core technology of the fourth industrial revolution
There have been three industrial revolutions in human history, and it has become a general consensus that artificial intelligence is becoming the core technology of the fourth industrial revolution. Throughout every industrial revolution in the past, whether it is mechanical technology, power technology or information technology, these core technologies have one thing in common, that is, very strong versatility, and artificial intelligence has also shown this strong versatility.
The landmark node at which these core technologies have a huge impact on human production and life is the beginning of industrial production, which can be applied very efficiently and then affect life. At this point, standardization, automation and modularity are highly apparent. At present, artificial intelligence technology represented by deep learning also has these characteristics, especially the two layers of framework and model in the four-layer technical architecture of artificial intelligence.
On the whole, the rapid development of deep learning + large model technology is promoting artificial intelligence technology into the stage of industrial mass production, which also means that it will have a huge impact on all aspects of our lives. In the future, all kinds of products and services used in our daily lives will also take AI as the core, and human beings must coexist and collaborate with AI.
Zhou Jingren, Chief Technology Officer of Alibaba Cloud Intelligence:
Future everyday applications will be "armed" with knowledge
What we are experiencing today is a disruptive technological revolution. In the Internet era, all kinds of information are already within reach, but the evolution of large models and technological changes are actually based on the accumulation of human wisdom, and there are more transformation capabilities, such as logic and the slow emergence of innovation capabilities. With such an intelligent capability, our current business will not be based on information acquisition as in the traditional way, but on the understanding generated by information.
We usually communicate well with people, even if they don't know what you're talking about, and they won't break down. Before large models, people would be very careful when interacting with conversational robots, because if they were not careful, they would touch the boundaries of the model. The current artificial intelligence big model is to develop in the direction of human intelligence in this sense, and how to make the crystallization of human knowledge continuously enhanced and innovated is the most important.
We may be at a starting point now, and we will find that all relevant daily applications will be "armed" with knowledge, that is, they can use knowledge to effectively solve the problems of the industry, and in the process, the big model itself will become the most central element. This day will come, and in order to do so, it will also be necessary to effectively and closely integrate infrastructure including computing power.
Overall, humanity has entered a knowledge-driven era, and while there will be ethical challenges, the technical and business opportunities are worth looking forward to.
Beijing Business Daily reporter Yang Yuehan