The ChatGPT era has come! In the next 5-10 years, AI will be able to perform better than humans in most human jobs. As far as education is concerned, how to deal with the challenges brought about by the rapid development of technology? How can college students be more competitive than machines in the age of AI?
According to Wen Wei Po, recently, at the "Interdisciplinary Development Forum on Artificial Intelligence Empowering Disciplines" hosted by Tongji University, many academicians and experts sounded the alarm for university education.
ChatGPT scored 90 points in the university final exam
What is the gap between domestic and foreign countries in terms of large models? Zheng Nanning, academician of the Chinese Academy of Engineering, former president of Xi'an Jiaotong University, and director of the Institute of Artificial Intelligence and Robotics, cited an example. He once took the test in the final exam of digital signal processing in the fall of 2022. As a candidate, ChatGPT scored 90 points out of 100 points in the test paper. Zheng Nanning said frankly that this is already the result of strict scoring. If this is a human candidate, he can give 94 points for the same answer sheet. In the face of the same answer sheet, a domestic model only scored 7 points.
"From a global perspective, it is true that the wave of natural language processing is coming. Domestic and foreign "politics, business, production, learning, and research" are all paying attention to and investing in this field, and many places are launching their own large-scale models , This is indeed an outlet.” However, Gao Wen, academician of the Chinese Academy of Engineering and director of Pengcheng Laboratory, said bluntly that behind the big model is general artificial intelligence.
Taking ChatGPT as an example, its rapid development is inseparable from the support of super computing power. For example, in the training of large models, the supercomputing device specially built by Microsoft used 10,000 A100 GPU cards, and it took one month to complete the training of GPT-3 with more than 170 billion parameters. Gao Wen said that if the Pengcheng Cloud Brain II independently developed by the laboratory is used, it will take 4,000 A100 GPU cards and take two and a half months.
"For us, the biggest challenge is the lack of computing power." He said that there are currently two types of startups for large models: "hot startup" and "cold startup". The so-called "hot start" is to "feed" the data to increase based on the results already given by GPT; "cold start" is to completely rely on yourself to clean the data and train a little bit.
"At present, there are a lot of 'hot start' in China, but very few can really do 'cold start'. This is an important challenge we are currently facing." Gao Wen said with emotion.
For example, in terms of the development of artificial intelligence, there is still a big gap between my country's basic theory and original algorithms, and the foundation is still weak in high-end chips, key components, and high-precision sensors. There are not enough platforms.
The reporter learned that at present, Pengcheng Laboratory is launching the Pengcheng Brain Big Model Project. The goal is to create the first fully autonomous, controllable, safe and controllable, open source and open natural language training large model base in China, with a parameter level of 200 billion and a performance benchmarked against ChatGPT. In August this year, Pengcheng Brain Big Model will complete the first phase of training.
Gao Wen mentioned that Pengcheng Lab will open cooperation based on the large model base of Pengcheng Brain Natural Language Training, and jointly build a domestically-made self-controllable and training large-scale ecological health model, creating opportunities for young people.
"Intelligence +" promotes professional upgrading, and university reform is on the way
The reason why it is called to create opportunities for young people is also related to the lack of high-end talents in the industry. Taking high-end artificial intelligence talents as an example, some experts said that in 2018, the number of high-end artificial intelligence talents in China was only equivalent to 20% of that in the United States. Now, five years later, although the gap is narrowing, the gap in high-end talents still exists.
Pan Yunhe, academician of the Chinese Academy of Engineering and academician of the International Eurasian Academy of Sciences, shared his observation on the spot: Judging from the list of previous winners of the "Mayor's Cup Design Award" in Hangzhou, the winners of the "Creativity Award" in 2015 were mainly universities, while By 2021, most of the winners will already be technology-based companies.
As a reviewer, he lamented that although colleges and universities have been submitting manuscripts, the content of the submissions has not kept up with the development of the industry. "This has to arouse the thinking of the education community: if our design majors continue to train people in the original way, we will probably not be able to find jobs in the future; the industry has already transformed into intelligence, and the people trained by colleges and universities are still in progress. Traditional industrial product design.”
To achieve leadership in the AI era, high-end talents are essential. At the scene, the experts all said that this round of artificial intelligence provides a good opportunity for the cross-development and empowerment of disciplines. They called for more colleges and universities to work together to train more talents in the industry.
In order to make up for the shortcomings of talents, since 2017, more than 30 "intelligence +" majors in my country have been included in the catalog of undergraduate majors in colleges and universities. The Ministry of Education has been committed to breaking the barriers of disciplines and majors, deepening the interdisciplinary integration of disciplines, innovating discipline organization models, reforming the talent training mechanism, and cultivating outstanding young innovative talents.
The reporter learned that in Shanghai, 16 colleges and universities and 2 scientific research institutes have set up artificial intelligence-related disciplines, among which Tongji University took the lead in the construction of intelligent science and technology. , was approved to build an interdisciplinary doctoral program in intelligent science and technology, and took the lead in building an independent intelligent unmanned system, which has achieved remarkable results in the field of artificial intelligence.
It is reported that in the new round of "Double First-Class" construction, Tongji University promotes the construction of "8+2" discipline groups, and promotes the development of the school's overall disciplines through interdisciplinary integration and coordinated development. Among them, the discipline group led by the discipline of Intelligent Science and Technology cross-links the other 9 discipline groups, playing a core empowering role, and giving birth to new discipline research directions and new fields of scientific and technological breakthroughs. At the same time, Tongji University is committed to exploring a new model of interdisciplinary talent training, and has established a system that integrates multidisciplinary interdisciplinary advantages and combines science, education, production and education with new artificial intelligence talent training.