The new generation of young scientific research talents has started in Shanghai, and the artificial intelligence industry is at the forefront of the times. Artificial intelligence | the world | the first generation

The 2023 World Artificial Intelligence Conference opened in Shanghai yesterday. This is the sixth consecutive session of the World Artificial Intelligence Conference. Unlike the past, this year's artificial intelligence has reached a turning point in the industry, and this conference has also attracted attention from all parties. At the opening ceremony, AI experts such as Musk, Tang Xiaoou, and Hu Houkun gathered, and the academic and industry engaged in a heated discussion on the development of the artificial intelligence industry.

On Technology

When it comes to ChatGPT, I have to mention Microsoft. At the opening ceremony, Hou Yang, Senior Vice President of Microsoft Global and Chairman and CEO of Microsoft Greater China, gave a speech expressing people's current views on AI. "With the continuous display of enormous potential in generative artificial intelligence, every application of every company will be driven by artificial intelligence in the future, and every company needs to have the ability to master digital technology."

During his speech, Huawei's rotating chairman Hu Houkun also stated, "The emergence of ChatGPT has pushed artificial intelligence to a new level. It can be said that artificial intelligence, especially general artificial intelligence, has become one of the hottest topics in current society."

General Artificial Intelligence, abbreviated as AGI, is a key topic of this year's WAIC, referring to an artificial intelligence system that thinks, learns, and performs multiple tasks like humans. It is usually used to distinguish it from "proprietary field artificial intelligence" and can be understood as one of the ultimate goals of artificial intelligence development.

After ChatGPT became popular, many people believed that large-scale natural language models would be a path to AGI, and large models also became popular. Hu Houkun, who travels around the exhibition hall every year, also said that this year's exhibition is not only focused on the research of large models, but also on the application of large models in different fields.

However, there are also different voices within the AI community regarding whether large models can truly lead to AGI. Yang Likun, a Turing Award winner in 2018 and the chief artificial intelligence scientist of the Meta AI Basic Artificial Intelligence Research Institute team, believes that ChatGPT like products cannot truly understand the real world and cannot reach the level of human intelligence.

Yang Likun explained on site that large models are only trained based on text and not video, so they do not understand the physical world, cannot plan, reason, and cannot achieve their goals. "The big model may pass the lawyer qualification exam, which tests the ability of information retrieval, but they do not have a sufficiently intelligent system." Yang Likun believes that AI needs a world model. The world model is like the way humans understand the world, understanding everything about the world through physical intuition and other means, which is a huge challenge for artificial intelligence research in the next 10 years.

On Regulation

If people still hold a positive attitude towards the arrival of the AI era, then there are complex views on whether AI needs regulation.

Tesla CEO Elon Musk first raised questions about artificial intelligence during his speech. "General artificial intelligence has stronger capabilities than humans, and it may have a positive future, but there is also a probability of a negative future." Musk believes that a key issue in the development of future artificial intelligence is the "ratio of robots to humans.". There will always be a stage where the number of robots will exceed that of humans, which may have some negative impacts. Therefore, people need to consider how to regulate such deep artificial intelligence.

Yang Likun expressed complete disagreement with strict regulation of artificial intelligence. He agrees with the importance of ensuring the security of artificial intelligence, but he believes that humans can design controllable and operable systems to ensure that artificial intelligence will not be honest and submissive, although this is a very challenging engineering challenge. "In the long run, the only way to make artificial intelligence platforms safe, good, and practical is to make them open source. In the future, each of us will communicate with the digital world through artificial intelligence assistants. If artificial intelligence is only controlled by a few people, it is not a good thing," Yang Likun added.

During the fireside dialogue, Microsoft's former executive vice president and foreign academician of the National Academy of Engineering, Shen Xiangyang, and IEEE Chairman and CEO, Saif Raman, also discussed the security issues of artificial intelligence.

Shen Xiangyang believes that the differences in AI control may stem from different perspectives. "From an academic perspective, it is important to conduct open research. As someone from the industry who develops AI products, I strongly agree that there should be some regulation and guarantees, as artificial intelligence is becoming increasingly powerful."

Saif Raman stated that industry organizations like IEEE may be an important force in preventing AI security issues. He said that currently, artificial intelligence projects require manpower and computing power, and most of them are not conducted in a secret state. So industry organizations like IEEE can provide a platform for scientists, engineers, and developers to discuss their work together, in order to ensure the security of AI development control.

On Talents

In the round table discussion of the opening ceremony, in addition to the Turing Prize winner, CAS Member, and Yao Jizhi, Dean of the Institute of Cross Information Research of Tsinghua University, the average age of the other three scholars was less than 33 years old. Their youthful appearance made the audience sigh.

Yao Qizhi introduced that there is a young Chinese scholar named Gao Yang from the Cross Information Research Institute of Tsinghua University. He achieved a breakthrough in reinforcement learning algorithms more than a year ago, not only completing application progress, but also making theoretical contributions, which has attracted widespread international attention.

After ChatGPT, the next important goal of artificial intelligence research is to enable robots with multiple perception abilities to autonomously learn new skills in different environments. And the contribution of this young Chinese scholar is to increase the learning speed of robots hundreds of times. "According to the typical robot reinforcement learning method, learning new technologies often takes several months, which is too slow." Yao Qizhi said, thanks to the research of this young scholar, reinforcement learning can now be completed in just a few hours.

In the past six or seven years, there has been a debate among senior executives in the field of artificial intelligence over whether or not to take the path of reinforcement learning. Yao Qizhi believes that the breakthroughs of young Chinese scholars have tilted the balance, showing people the path of reinforcement learning and understanding that there is still a long way to go in the future.

Professor Tang Xiaoou from the Chinese University of Hong Kong also shared the experiences of his three students on site. The experiences of three students represent different stages of the development of the artificial intelligence industry: Wang Xiaogang planted many original seeds in the early years of the rise of deep learning, He Kaiming laid a very solid and deep foundation for deep learning, and Lin Dahua made everyone's efforts flourish through development and large-scale models. Just two weeks ago, the autonomous driving model studied by Tang Xiaoou's team stood out among 9155 articles and won the CVPR Best Paper Award in 2023. According to Google Academic statistics, this is the first best paper from the International Computer Vision Three Summit, completed entirely by Chinese scholars, in the more than 40 years since the reform and opening up.

Different sharing points to the same young scientific research talents in China. "In deep learning, a group of original researchers have planted big trees that have begun to bloom and bear fruit. The new generation of students has successfully started in Shanghai," said Tang Xiaoou.