99% of the assignments are written by AI: In the eyes of contemporary students from prestigious universities, what's left of college?

The first group of students in human history to grow up alongside generative AI are entering their graduation season. These graduates may find that whether they studied knowledge-based courses or practical skills in college, they are all subject to the impact of AI. To stand out, they must become those "who are better at using AI."

When knowledge can be retrieved by AI and skills can be replaced by tools, what unique things can universities offer students? In the face of the impact of AI, what kind of abilities are the "meta-abilities" that will not be reshuffled by the times?

The following are the highlights of this conversation:

01 This year's graduates: ChatGPT arrived when they entered college

Hong Jun: I've noticed that everyone is studying different majors. When you all entered college, it was the year when ChatGPT had just emerged. Since you started college, AI has already been a part of your life and studies. I'm very interested in Alfred because your major is law, which makes you a typical liberal arts student. Secondly, you also use AI to learn programming.

Alfred: Although I'm studying law, I'm actually not very good at my major. Basically, whenever I had free time in my freshman year, I would take computer-related courses as an elective. Around the end of 2022 and the beginning of 2023, we saw the emergence of ChatGPT. It's truly amazing. I envision a future where two or three years from now when I graduate, AI will definitely be much more powerful than it is now, while my level may remain about the same as it is today. What role will I play then? Will I still be needed?

There is a teacher at our school who has been a partner at a law firm in the United States for more than a decade, specializing in mergers and acquisitions. He is a typical white American grandpa, very kind, and he teaches in a very gentle way in class. When his course ended, I asked him, "Rosoff, do you think ChatGPT will have a significant impact on the future careers of junior lawyers and legal assistants?" Before I could finish my question, he immediately interrupted me and said, "Definitely yes." Then I realized that, well, with the mindset of "if you can't beat it, join it," I started to actively embrace AI.

At that time, I thought the smartest investment I made, at least up to now, was to spend more than $20 every month to use ChatGPT. It was actually quite expensive for a student at that time. I started spending every day with ChatGPT.

Hong Jun: What do you use ChatGPT for?

Alfred: I believe that if you can't beat it, you should join it. But to join, you first need to understand it. You can't just have an overview of the industry; you need to understand the basic theories of how AI is constructed and why it has become what we see today after a series of iterations. At that time, I had all kinds of questions and used ChatGPT to learn about AI. I asked it to help me plan my learning route and also told it about my knowledge background. I took many related courses as an elective, whether they were open courses from Stanford and MIT on Bilibili or YouTube. When I was studying those open courses, I would actually ask ChatGPT when I didn't understand something.

At that time, I felt that it also had a great impact on education. To be honest, many teachers are very good at research, but in terms of teaching, I found that they may not be very good at teaching you how to do something. However, ChatGPT is different. It's like a teacher who understands you very well, knows your current level, and can teach you in accordance with your aptitude and guide you patiently. I can clearly feel that it can accelerate my learning process in this field much faster than if I just choose a course at school.

Hong Jun: How much of your elective courses did you learn from teachers, and how much did you learn from online open courses and AI tools?

Alfred: 30% comes from textbooks and courses, and 70% comes from conversations with AI, as well as the most crucial project-based practice and review. This 30% is what you must know, such as what the syntax of Python is and what the most basic technical tips of programming are. It builds your basic understanding of a subject. However, the more practical 70% comes from multiple rounds of in-depth conversations with ChatGPT, as well as doing projects on your own, learning while doing, and then reviewing after finishing the projects.

Hong Jun: You just mentioned your elective courses. I'm curious about to what extent you use AI in your major, law?

Alfred: Oh my god, I'll always remember the fall semester of 2023. That was the first semester when I got a perfect 4.0 GPA. In fact, I was really bad at studying, and my mind was not on law at all. But in some professional core courses, such as corporate law, constitutional law, and common law, I actually relied on AI to help me review to a large extent. At that time, we were studying the case law of Anglo-American law. I gave all my case law course materials to it and told it to imagine that it was a teacher and how it would set questions. It really gave me two or three questions. At that time, GPTs had just come out, and you could encapsulate all the context in a folder and have a back-and-forth conversation with it in this folder. In this way, it really accelerated my efficiency in absorbing this subject at that time.

When I was studying corporate law, I fed the AI with the works of our teacher, who is also the author of the textbook. I named this GPTs "Shi Tiantao's substitute" and then started to have a conversation with it. During the conversation, I understood some of his academic views and key points on this subject. My one-on-one communication with AI helped me understand the most essential methodology and knowledge in this field more quickly. Although it's been a long time since the fall semester of 2023, I still think that some of the key knowledge in the two courses that I used AI to assist my learning in are still in the shallower part of my memory and can be retrieved. This means that I may have learned them quite well.

02 Re-examine the value of universities

Hong Jun: Do you have on-site discussions in exams, or can you use AI to complete your assignments?

Alfred: It's been a long time since I took an exam. However, at least in domestic universities, including Tsinghua University, some adjustments have been made to the assignment assessment methods. When I first started learning Python at Tsinghua, the teacher was very resistant to code written by AI. So students spent a lot of time making the AI-written code look a bit stupid when submitting their assignments. But now, what I've seen is a shift. More and more disciplines are actually actively encouraging the use of AI, and even using AI in some key assessment methods.

Hong Jun: Did this change happen in the computer science department or the law department?

Alfred: I haven't seen a clear change in the law department yet, but in the computer science department, they are not as strict in restricting you. At least it's not like before, where if you used AI, it would be considered academic fraud and you would fail the course, which was a very terrible consequence. I can clearly feel that in the past two years in China, teachers' attitudes towards AI have changed from being indifferent, a bit conservative, or even negative to being very positive and open.

Hong Jun: Then I'm curious. If everyone writes code with the help of AI and the teacher sets a question, and everyone can write beautiful code, do you think there will be any changes in the evaluation criteria for students' performance and their mastery of knowledge in computer science?

Alfred: At least at this moment, evaluating students in the old-fashioned way is obviously out of touch with the times. We urgently need a brand-new evaluation system. Training AI is actually similar to educating a good student. For example, the improvement of AI capabilities today is often driven by benchmarks. Although many AI models perform very well on benchmarks, they perform poorly in some more open real-world tasks. In human terms, it may be a student with very high exam scores, but when you put it in a completely open environment to solve some practical problems, it will encounter great difficulties. This also makes me think about what the value of a university is. If we don't think about this question, it will be difficult for us to establish our evaluation system and so-called value indicators.

Universities used to have two coexisting values. One value is to figure out who I am, where my passion lies, and where I'm going, which are very basic and important questions. At the same time, it also needs to carry the value for society, that is, "you need to become a valuable person." This "valuable" means your unique skills and specialized knowledge that can help you earn money.

In the universities I've experienced, more emphasis is placed on the latter. Even at Tsinghua University, most people don't think about who they are, what they really want, what they are good at, and what their preferences are. Many people may not have a clear answer even when they graduate from college.

When we focus too much on the latter, we will find that this system has been eroded. Because the valuable skills we thought we acquired in college, AI can actually do better than us, and it's even cheaper and available 24/7. This is especially fatal for those knowledge workers who used to receive, process, and output information. The corresponding industries include programmers, lawyers, finance, and so on.

I was very lucky in college. I took a course at Peking University, which is similar to Western liberal arts education, to train a person's critical thinking. After that, when I was choosing undergraduate courses, I deliberately avoided some memory-based courses. From the perspective of the law school, these are courses like "Practical application of securities law in the capital market" and "International trade arbitration," which are very practical, as well as some professional courses like international law and trade law. After studying these courses, you will definitely have a corresponding major and may even get a good salary.

Liberal arts education originated from the 'seven liberal arts' in ancient Greece (grammar, rhetoric, logic, arithmetic, geometry, music, and astronomy). The medieval illustration 'Philosophy and the Seven Liberal Arts' vividly shows this tradition - the Queen of Philosophy sits in the center, surrounded by the seven arts.

©Dnalor_01/Wikimedia Commons/CC BY-SA 3.0

But when I was choosing courses, I deliberately avoided those I was not interested in. I put more energy into courses that focus on cultivating my critical thinking. For example, there is a course in the law school called "History of Western legal thought," as well as jurisprudence and constitutional law. Outside the law school, I was very lucky to choose and finish a course called "Modern Western philosophy" at the Xinya College of Tsinghua University. This entire course is actually about discussing a book called "Meditations on First Philosophy." I remember that when I actually read the original work, I was a bit dazed after finishing it. This means that you start to examine all things around you from a critical perspective. And at the first moment of contact, it deconstructs your personal beliefs. But this doesn't mean it leads directly to nihilism. Instead, it's more like when you start to deconstruct all things and think they are meaningless and can be doubted, you will at least find that there are some things that always exist. When I experienced that moment, I could more or less feel that this moment is actually very difficult to be replaced by AI. I also gradually got a sense of the key quality of "critical thinking" that we've always emphasized in quality education.

When I have this awareness to choose my courses, my reading materials, and how to interact with the world, I will find that I don't feel so lost and drifting anymore. I don't feel like I may have no place in society as AI capabilities continue to evolve.

So I think the current undergraduate evaluation system, at least the one I've experienced, needs to be reconstructed in the next era. Universities should take on the role of a place to explore students' inherent potential. In this evaluation system, whether you can coexist well with AI to solve problems is one of the very important indicators.

03 The core value of future universities: social interaction

Hong Jun: You just mentioned that in the entire college process, tools are not the most important thing, and what practical things you learn are also not the most important. What do you think is the most important asset that Tsinghua University has given you?

Alfred: I think there are two points. The first is the excellent environment and the peers. When students from different majors and all over the country gather together, communicating with them is always a pleasure, and you can always learn a lot from them.

The second point is the teachers. I've met several very important teachers at Tsinghua. I often make appointments with them during their office hours to have one-on-one conversations about many things. When a person with knowledge in a certain discipline and some social experience communicates with you, they can let you see some things in advance and avoid stepping into unnecessary pitfalls.

I think the more crucial point is this "people connection." At this moment, when other skills start to fade and seem less irreplaceable, the ability to build trust and persuade others becomes more important. Because this can also help you maximize what you gain when you enter a new environment.

Image source: Tsinghua University official website

Hong Jun: This also reminds me of what Elon Musk said in an interview recently. He said that in the future, the knowledge in universities will become outdated, and academic degrees will lose their effectiveness. The core of future education is to cultivate qualities that AI cannot replicate, such as curiosity, creativity, and critical thinking. He even said that in the future, universities will be about social interaction. I think it makes a lot of sense.

Alfred: Yes, if I could go back to my freshman year with my current knowledge, I would definitely try my best to socialize. This kind of social interaction doesn't just mean going to bars and participating in various activities every day. Instead, I would look forward to having more one-on-one in-depth conversations.

Hong Jun: Speaking of social interaction, next I'm going to ask Kolento. I remember the first time we met was when you came from New York to the Bay Area for some visits. Soon, Kolento introduced many classmates to us, and we recorded a podcast together. I know that during this process, Kolento has been in long-term contact with many founders of startup companies in Silicon Valley, including investors. He has also been participating in our topic selection meetings and has provided us with many good ideas, useful tools, and information about industry developments. I also know that during this process, you have interned at many companies and are also working on your own Agent. So Kolento, in your current college situation, how much time do you spend focusing on your own courses, and how much time do you spend on your Agent research or socializing?

Kolento: Most of my time is not spent on socializing or completing courses. Instead, I spend more time doing what I want to do, which is actually outside of school. The idea that "universities are about social interaction" that Musk mentioned appeared in my college application essay. I told NYU that I think the most important purpose of going to college for me is social interaction.

Those who know me should know that I usually spend most of my class time reading papers or working on my own products, and only a small part of the time on learning the content in class. Because the content in class is pre-set, and its purpose is to