Pursuing a PhD in Computer Science!, 53 year old Zhou Hongyi brings good news! Admitted to Tsinghua Company | Zhou Hongyi | Tsinghua University

While high school graduates were busy choosing and filling out majors, they didn't expect to wait for the Tsinghua admission letter posted by Zhou Hongyi, the founder, chairman, and CEO of 360 Company.

On the evening of June 27th, Zhou Hongyi posted an admission letter to Tsinghua University on his personal Weibo account, stating that "I have finally passed the exam and hope that 360 Intelligence Brain can help me graduate smoothly." The letter showed that Zhou Hongyi was admitted to the Electronic Information major of the Department of Computer Science and Technology at Tsinghua University and will pursue a part-time, targeted employment doctoral degree.

This also means that 53 year old Zhou Hongyi is going to school again!

Is pursuing a PhD or related to 360 Brain Intelligence?

According to public information, Zhou Hongyi was born on October 4, 1970 in Zhumadian City, Henan Province, with ancestral roots in Hubei. In 1992, he graduated with a bachelor's degree from the Computer Science Department of the School of Telecommunications at Xi'an Jiaotong University. Due to his excellent academic performance, he was recommended to pursue a master's degree in System Engineering at the School of Management at Xi'an Jiaotong University. He graduated with a master's degree in 1995.

This means that even though Zhou Hongyi is 53 years old, he still chooses to continue his studies and pursue a PhD. So, what is the department and major he is studying this time?

According to the official website of Tsinghua University, the Department of Computer Science and Technology was established in 1958. After more than 60 years of unremitting efforts, it has developed into a computer science department with strong teaching and research comprehensive strength and great influence in the field of computer science in China. It plays an important role in the development of China's computer industry and even in the construction of the national economy. In 2022, Tsinghua University ranked among the top computer science disciplines in multiple globally influential universities, with the UK QS ranking 15th globally, The Times ranking 16th globally, and the US News ranking continuing to rank 1st globally

It can be said that Zhou Hongyi has entered the most prestigious computer series in China, which is also in line with the positioning of 360 software manufacturers. In addition to posting the admission notice from Tsinghua University, Zhou Hongyi also emphasized 360 Intelligent Brain. The phrase "I hope 360 Brain can help me graduate smoothly" may also mean that Zhou Hongyi's research topic for this doctoral degree may be related to 360 Brain or large models.

It is reported that on June 13th this year, at the 360 Intelligent Brain Model Application Launch Conference, 360 officially released its new AI product "360 Intelligent Brain Model". Zhou Hongyi introduced that the development and implementation of 360 models have eight natural advantages, covering the entire training process. At the same time, the generation and creation, multi round dialogue, coding ability, reading comprehension, logic and reasoning, multimodality, and other ten core abilities of 360 Intelligent Brain can cover all application scenarios of large models. Since March 29, 2023, 360 Brain has iterated from version 1.0 to version 4.0, continuously implementing application results and refreshing the big model "China Speed".

In addition, 360 Brain 4.0 has officially been integrated into 360's full range products, including search, browser, security guard, etc. 360 founder Zhou Hongyi stated that 360 has passed the evaluation of the Ministry of Industry and Information Technology's Institute of Information and Communications Technology, becoming the first trustworthy AIGC model in China.

In fact, when discussing how to fill out majors and preferences recently, he has also combined this issue with AI and big models. Zhou Hongyi said, "Someone asked me how to fill out my college entrance examination preferences. In my opinion, a good major is more important than attending a certain university. Others asked me which major will not be eliminated by artificial intelligence. I believe that the future development of AI will be people-oriented. Big models are the greatest tool in human history and also a tool for 'assisted driving'. They should become a good helper for people and will not completely replace their work."

Make every effort to "call" 360 Super Brain

It is understood that most of the content posted by Zhou Hongyi on Weibo recently is related to AI big models, and he personally is also fully committed to "calling" 360 Super Brain!

For example, on June 7th, Zhou Hongyi took advantage of the popularity of the college entrance examination and directly wrote a college entrance examination essay using his 360 degree intelligence.

On June 1st, he used his 360 degree brain to help adults find several reasons to celebrate Children's Day.

In fact, despite the booming concept of AI this year, the stock price of 360 has skyrocketed all the way, but there has been a certain pullback recently.

On June 26, at the main forum of Nishan Dialogue on Digital Civilization of the World Internet Conference, Zhou Hongyi once again talked about his views on the big model and mentioned for the first time that there are four shortcomings in the existing public big model:

Firstly, although the public ownership model is a generalist, it lacks industry depth. "We used to think that GPT knows everything, but if you are an industry expert, you will find that GPT does not have enough knowledge depth in vertical fields such as security and finance. Many companies have discovered this characteristic in their own training models, and if they want their abilities to be balanced, they will sacrifice depth," said Zhou Hongyi.

He believes that the future vertical big model is an important direction for development. The combination of universal models and proprietary knowledge data in various fields can transform big models from "all-in-one" to "government, industry, and enterprise", which is the true value. "The latest data shows that GPT4 is also composed of eight vertical models, which indirectly confirms this viewpoint."

Secondly, public ownership models can easily lead to internal data leakage within the enterprise. On the one hand, public models are not deployed locally, and there is inevitably a risk of data leakage when communicating information with external parties; On the other hand, public ownership models are also unable to achieve hierarchical management of internal organizational permissions. Therefore, the use of public ownership models by governments and enterprises inevitably poses security risks.

Thirdly, for enterprises, public ownership models cannot guarantee the true credibility of content. Zhou Hongyi stated that there is an illusion in the big model, which is commonly referred to as "nonsense" in a serious manner, and there are often problems of being arrogant and untrustworthy, which need to be corrected through internal search and internal knowledge base of the enterprise.

Fourthly, the public ownership model cannot achieve cost control. He gave an example that many enterprises only need the ability to write code using large models, and a vertical large model worth tens of billions can meet the demand. Using a large model worth hundreds of billions is a waste of cost. In terms of cost control, vertical large models will have significant advantages. On the basis of training a "general education" big model with public data, training a proprietary big model can achieve "twice the result with half the effort" and reduce costs and increase efficiency for enterprises.

In addition, as a company that started with "security" as its selling point, Zhou Hongyi also attaches great importance to the security issues of large models. He previously stated that large enterprises or governments must pay attention to security issues when using public GPTs. When large enterprises use GPT, they not only search, but also feed it a lot of data and inform it of many contextual prerequisites, which means that the privacy of a lot of information may be compromised. A few days ago, a similar incident occurred at Samsung where software engineers inputted company confidential code into ChatGPT, resulting in the leakage of company confidential information. So, it is necessary to privatize large models for large enterprises or governments. On the basis of incorporating public GPT knowledge and capabilities, a private GPT is trained and only used by the enterprise itself or customers.