GamePea reports — On April 12, Liu Wei (also known as Dawei Ge), co-founder and president of miHoYo, returned to his alma mater, Shanghai Jiao Tong University (SJTU), for a two-hour exchange and Q&A session with students from the School of Artificial Intelligence. Drawing from his own journey from a 'small-town test-taking machine' to a co-founder of miHoYo, Liu encouraged SJTU students to reduce anxiety in the AI era. Below is a transcript of his sharing and Q&A session.
Liu Wei: Dear fellow SJTU students, my teachers Mr. Hong, Mr. Chen, Mr. Yang, Dean Wang, and my college classmate Mr. Xu, I am truly honored to have the opportunity to return to my alma mater and the School of Artificial Intelligence to share with you again this year. As Mr. Xu mentioned, today's agenda is roughly as follows: I'll start with a brief sharing, followed by a Q&A session covering both online and on-site questions.
I'll try to keep it within 20 minutes, and then we can discuss topics you're interested in, such as college life, career, AI, and the current era. A small suggestion: since this is a sharing session, I hope you'll ask fewer game-related questions. Two reasons: First, game questions aren't that important—you'll understand once you play the games. Many of you may be more concerned about your college life and the anxiety brought by the AI era, so let's focus on more helpful topics. Game questions are entertaining, but they don't offer much insight. Second, I don't know the answers to many game questions, as our team handles most game products. The beautiful works you see are created by the team; I'm just a front-facing figure. If I get something wrong, they might hold it against me. So, please ask fewer game questions. If you have a pressing question that's been bothering you, feel free to ask, but keep it minimal.
Let's begin. Today, I'll spend 20 minutes sharing how to deal with and combat anxiety in this AI era, especially for SJTU students. I hope to offer some inspiration.
This year marks SJTU's 130th anniversary. I participated in some school events and met my college teachers, Mr. Yang Yifan and Ms. Hong Mei, which sparked many memories of my undergraduate and graduate years at SJTU. These years completely transformed my values. To be honest, I was born in a rural area and entered SJTU in 2005. Before that, I was a typical small-town test-taking machine. Coming to Shanghai from Changsha in 2005 was my first time leaving my hometown. Undoubtedly, SJTU shaped my outlook on life and the world, opening up a vast world for me.
But frankly, although I may seem to have achieved some success and am a unique individual, my college life was actually quite conventional for an SJTU student, perhaps due to my small-town background. In college, I worked very hard, and my classmates were also excellent. I strived to be a standout at SJTU, but overall, I followed the rules. When I entered SJTU, I didn't have grand ambitions or know what I wanted to do in the future. Compared to classmates from Shanghai or other developed cities who had seen more of the world in high school, I was quite anxious. In a small town, you might be impressive, but at SJTU, you quickly realize that both English and computer skills are your weaknesses. I was in the School of Electronic Information and Electrical Engineering, where computer knowledge and English were crucial. SJTU placed great emphasis on English in 2005, with many courses taught entirely in English—both were my weak points. So, I was quite anxious when I entered SJTU, not exactly inferior, but not confident either. For a long time, I was a rule-following student.
During my freshman and sophomore years, I gradually realized what I should do, but I still followed the rules. At that time, top students at SJTU were taking GRE and TOEFL, preparing to study abroad. In our era, pursuing a PhD abroad was the best option, and almost all top students took that path. I wasn't someone with many ideas, so I thought, 'If everyone is doing it, I should too.' Looking back, my college life wasn't particularly exciting; I spent most of my time at the Pao Yue-Kong Library on the Minhang campus. When bored, I read books—some useful, many useless, like biographies and psychology. Other times, I did utilitarian things, like practicing exam questions.
For example, in my first year, the School of Electronic Information reformed its curriculum, introducing a more challenging course called Mathematical Analysis. You could choose between Advanced Mathematics and Mathematical Analysis. Those aiming to study abroad or challenge themselves chose Mathematical Analysis. Our cohort was among the first to take it, and there were no textbooks—teachers handed out lecture notes, and we learned on our own. My math wasn't great, despite decent scores in the college entrance exam, so I failed the first exam. I had worked hard, but still failed. For a freshman from a small town entering a big city and SJTU, life felt quite bleak. I spent hours in the library practicing problems from a set of six books by Russian mathematician Demidovich, completing four of them. The results improved visibly after that.
Overall, my college life, no matter how exciting or unique it may seem today, was actually quite ordinary. I just focused on practicing and scoring. Moreover, our school (like today's School of Artificial Intelligence) had many top-performing students, especially from other provinces. Even with my hard work, I only ranked in the top 20%, which wasn't particularly outstanding. I never won a first-class scholarship at the school; the best I got was second-class, though I did receive some special scholarships for specific achievements. But in terms of academic rankings, I never made it. So, I'm not being modest—that was my reality back then.
Most of my college time was spent in the library practicing. I also had some old-fashioned ideas, thinking that since I wanted to study abroad, my GPA had to be high, and participating in extracurricular activities was a waste of time. So, I joined very few clubs, only the minimum required for comprehensive evaluations. Looking back, I think it was a poor trade-off—your college life becomes dull, and you miss out on opportunities to meet more classmates at SJTU.
Reflecting on my college life, I'm not being modest; it was truly ordinary. I just worked very hard to get a good GPA, hoping to eventually study abroad. That was it—I didn't do anything earth-shattering. The turning point came in 2009, my senior year. I had planned to study abroad, but in 2008, the financial crisis hit. I applied to many schools but didn't receive a single offer. Classmates with better grades and research experience were also affected, but they still got offers, just not from ideal schools. I was particularly unlucky—I got nothing. So, in the second half of 2008, I chose to pursue a direct master's program at SJTU. It sounds a bit like bragging, but it was truly because I didn't get any offers, not because I had great foresight. Of course, looking back, history gave me the best choice.
I worked hard on the TOEFL and did some research, but with little success. In the first half of 2009, during my senior year, I was admitted to the master's program. Since it was a direct program at SJTU, I entered the lab early. Soon, I realized I wasn't cut out for research. There was a female classmate in my lab, now a professor at SJTU, and comparing myself to her made it clear that research wasn't for me. So, in my senior year, I quickly decided to change my life path. I realized the most important thing was to meet SJTU's master's graduation requirements as quickly as possible. Back then, the requirements were relatively low: publish a journal article and write a thesis. I started preparing in my senior year and, luckily, completed everything by the end of my first year of graduate school. I even finished my thesis early—showing how determined I was to get that SJTU master's degree.
This was the rapid decision I made in the first half of 2009. A year later, I had accomplished it. From the second half of 2009 to the end of 2010, that year and a half was the most transformative period of my life. Why? Before that, I was a typical small-town test-taking machine and a good student, thinking that classmates with poor grades weren't worth associating with—very old-fashioned. I spent all my time with so-called top students, only to find they were all better than me. In college, I learned from them to score higher, but my perspective and social circle were extremely narrow.
After starting graduate school, I realized I wasn't going to pursue research, and grades didn't matter anymore—just graduating was enough. The most important thing was to publish a paper. From my senior year onward, I took graduate courses, thinking that passing was sufficient. When you don't need to focus on scoring at SJTU, you free up a lot of time. Honestly, SJTU's undergraduate courses were much harder; graduate courses were relatively easier, at least for those of us not doing research. So, from the second half of 2009 to the first half of 2010, I began meeting various 'non-typical' SJTU students. That year and a half completely opened my mind.
At a good school like SJTU, there are also students who aren't so 'good.' For example, some classmates who failed many courses were doing something I had never heard of—selling on Taobao. In 2008, some SJTU students started Taobao stores, earning tens of thousands a month. I was shocked. These were the first group of classmates I met; they had a place near the SJTU gate with many computers and a warehouse for shipping. It was a huge revelation for me—so SJTU had people like this too.
Second, I had a senior from the computer science department who was good at academics and programming. He was helping others trade stocks and earning tens of thousands a month, which was a fortune back then. My monthly allowance was 600 yuan, or 1,000 yuan when I was well-off, which was already comfortable for part-time work at SJTU. But my senior was earning 10,000 to 20,000 yuan a month, treating us to meals at the most expensive places near the SJTU gate. I was amazed.
I met many such people—non-typical SJTU students—who completely broadened my horizons. I realized SJTU wasn't just about studying abroad and scoring GPAs, nor was it only about top researchers like those in my lab. SJTU also had people with rich experiences. I even joined the SJTU Running Bug Club because I ran marathons (half and full) at school and was quite fast, finishing second in the 800-meter race at two consecutive school sports days, with a time under 2 minutes 30 seconds. I also met classmates from various departments in the club. Some of them traveled across the country to run marathons, while others cycled around China on modest bikes. These people completely opened my eyes.
Also, around the end of 2010, I met some SJTU seniors who wanted to start a business after graduation. I hadn't graduated yet, but after many conversations, they thought I was a good fit and invited me to join them. In the second half of 2010, during my second year of graduate school, without any preparation, I was persuaded by these seniors to go to Shenzhen to start a business. I want to thank SJTU and my graduate advisor, Ms. Hong Mei, who was very understanding. After deep discussions to ensure I wasn't acting impulsively or being scammed, she gave her tacit approval under difficult circumstances.
In the second half of 2010, during my second year of graduate school, I inexplicably left SJTU (without taking a leave of absence) and went to Shenzhen to start a business with my seniors. What did we do in Shenzhen? We rented an apartment and spent days looking at various things—factories, startup projects—wasting three or four months in a daze. Was this experience valuable? It had one crucial value. Before going to Shenzhen, I met my college classmate Cai Haoyu. He asked, 'I heard you're taking a leave of absence to start a business?' I said yes, even though I didn't know what I was going to do. He told me about his own projects, and I said, 'That's great, let's keep in touch.'
After three or four months of doing nothing in Shenzhen, I felt like I was wasting away in that apartment. At the end of December 2010, I returned, thinking that starting a business wasn't working out. I decided to go back to school, finish my master's, and then apply for a PhD. I had an advantage: SJTU had a dual master's program with Georgia Tech, and I was enrolled, so I would have a master's degree from Georgia Tech. My first choice was to apply for a PhD at Georgia Tech.
When I returned, I went to see Cai Haoyu and told him I was back and planning to pursue a PhD in the US. He told me about his situation—his previous startup had also failed. That's when the gears of fate began to turn. Cai Haoyu said he wanted to continue starting a business and invited me to join him. I learned about what he was doing. Honestly, I wasn't great at research or programming at SJTU; I was just a good student focused on GPAs. I could understand technology but wasn't particularly skilled. I was surprised—I had heard his previous startups involved tech experts, so why was he asking me? That's where Cai Haoyu's brilliance shone. He realized from the start that I wasn't suited for technical work. He had tried two or three startups, each time with a group of tech people, and each time ended in discord due to conflicting ideas. He said he realized that a group of tech people couldn't handle everything; they needed someone non-technical to handle things outside of tech, even to manage them. I thought that suited me.
With Cai Haoyu's encouragement, we started our business at the end of 2011. Besides me, we also brought in other classmates. And that's how the miHoYo story began.
I tell these two stories (from my undergraduate and graduate years) to show you that life often doesn't let you see the path ahead clearly, but looking back, you'll find that many paths are connected. At SJTU, whether as an undergraduate or graduate, I did many things that, by chance or fate, eventually led me down that path. Sometimes, in this era, many of you are so excellent that you feel you need to see a clear, straight path to maximize results—getting the best research outcomes, finding the best job, etc. But life is a process of exploration; often, you can't see a clear path ahead. If you do your best in the present, looking back, as Steve Jobs said, 'connecting the dots,' they will connect.
Honestly, I was definitely a hardworking person in college, extremely hardworking. But did I have a clear vision of my future or do something incredibly impressive? I don't think so. Looking back, I feel like fate chose me. What I did well was: when fate chose me, I did my utmost to cherish the opportunity and do my best. Just like in school, even though I didn't know where the future lay, I worked hard on my GPA and scholarships, giving myself a satisfactory answer. After starting the business, it was the same—he warmly invited me to join, gave me trust and opportunities, and I did my best to excel. That's my attitude toward life. But did I see clearly back then? No.
What I want to say is that in this era of rapid change, we truly can't see the future all at once. So, don't be too anxious.
Now, let me share the second part: How to combat anxiety in this era? I'll offer a few suggestions that may not apply to everyone, but looking back, I think they would have been helpful. I hope they inspire you.
First, for everyone today, finding a path that 'maximizes returns' is very difficult. Which research topic will become mainstream? Which direction will offer the biggest opportunities in the job market? These external factors are highly uncontrollable. Only one thing is controllable: what you truly want to do and who you want to become. This depends entirely on your inner self. You're still young and may not have seriously thought about what you really want to do. Often, external noise is loud, especially when AI feels like it's about to consume everything, causing great anxiety. At such times, you should spend a lot of time thinking about who you truly want to become and what you truly want to do. This answer is very difficult, but you should still strive to find it.
Why do I think this way? I learned it from Cai Haoyu. Over the years, our startup has had a slogan: 'Tech Otakus Save the World.' We made games because, for a grassroots dormitory startup team like ours, games were the most manageable—low capital costs, no need for extensive social resources, relying entirely on our own innovation and creativity. But deep down, Cai Haoyu has always had a dream of 'saving the world through technology.' Look at these years: no matter how big the company has grown, he has always been on the front lines of technology because he truly loves it. Including in 2023, when GPT emerged and the AI era arrived, he unhesitatingly left everything at miHoYo behind to start learning about large models from scratch. I remember a colleague asking him in 2023, seeing him writing PyTorch code, 'Brother Cai, training large models must be really hard, can you learn that?' He replied, 'Why can't you learn?' So, when I look at him, I see someone who truly loves technology. Whether he had nothing when he was young or has achieved success and resources, his state hasn't changed much. I can see a genuine love, purity, and persistence for technology in him. It's been many years—15 years since I met him and we started the business together. This is something you can't fake. Someone who has done this for 15 years truly loves it from the bottom of their heart.
I think the best state in life is like his: thoroughly enjoying what you do, looking forward to each day with anticipation and longing. Recently, with AI agents becoming very popular, he's been handcrafting some agents himself. I see him sleeping only five or six hours a day, but he's as energetic as when we first started the business. When someone genuinely loves something, their life state can be incredibly good, and they can weather any cycle.
For today's students, many utilitarian things are more like results—achieving great research outcomes, being recognized by the industry, finding a good job with a good salary. These things seem like results, but they are completely beyond your control. If you must pursue them, they will bring you immense anxiety. So, I hope you first find something you love to do, whether it's research, entering the industry in a certain direction, or having a great idea with classmates to work on. Don't overly pursue the obvious results, because that's completely impossible. As I shared from my college and graduate life, many things are the result of timing, location, and people. But if you have something you want to do, you won't be anxious, and you can withstand cycles. That's the first thing I want to say: you must find something you truly want to do.
Second, how to combat anxiety? I've found, including through our entrepreneurial journey, that much of our anxiety comes from not doing anything. When I truly start doing something, most of the anxiety quickly dissipates. If you have a goal but feel anxious, the most important thing is to take action immediately. Most people's anxiety comes from overthinking and underacting. How can you act enough? A key point is: how do we learn to do something? The most important thing is to start doing it. What you think and what you end up doing are vastly different, but that's okay. As long as you start, your anxiety will decrease, and you'll gain new insights.
SJTU students, including my past self, have a major flaw—because of past academic success, we have a 'meritocratic' mindset, wanting everything to be perfect from the start. This is a huge obstacle. In this era, if you want to combat anxiety and do more things, the most important thing is to start doing it immediately, regardless of whether you know how. Start doing it; even if you only achieve 10%, that's fine. Keep doing it, and you'll reach 20%, 30%—that's okay. If you aim for 80% or 90% from the start, you'll feel immense pressure and hesitate to begin. This is a heavy burden for SJTU students, especially those with good grades.
An important psychological hint is: just start doing it. There's a simple saying: 'To get started, the most important thing is to show up.' In our entrepreneurial journey, when we ventured into new businesses and explored new things, including things that later seemed very successful, it was always the same. The results were never good at the beginning, but by starting, we were on the right path.
Looking at our own entrepreneurial journey, this feeling is especially strong. In recent years, after completing 'Genshin Impact,' people think we've done something remarkable. Many ask, 'How are you so unique? How did you capture the most unique thing of this era?' I always ask them: In 2011, why didn't any mainstream VC invest in us? It's normal—what we were doing back then was very ordinary, not unique. All uniqueness is proven in hindsight. No one in the industry thought we would succeed, and that's normal: a student from SJTU, no work experience, no background, how could what we did succeed? But why did we succeed? Because we started doing it, and each time we did it a little better: after 'Honkai Impact 2,' we made 'Honkai Impact 3,' and after that, 'Genshin Impact.' Step by step, we got better. But going back to the beginning, our starting point was 10% or 20%, no exaggeration. When people see us with 'Genshin Impact' and think miHoYo is amazing for creating a globally influential work, they don't know that we've been on this path for 10 years. From making 'FlyMe2theMoon' in 2011 to 'Genshin Impact,' it took a full 10 years, and we never changed direction.
So, many so-called amazing things and great undertakings start very rough. That's why, in 2023, when we decided to start working on large models and participate in the greatest endeavor of this era, we didn't hesitate. Even though we know nothing today, so what? Starting is good.
I want to tell all students: when you want to learn to do something, the most important thing is to start doing it, even if it's only at 10%, 20%, or 30%. The biggest flaw of SJTU students is wanting everything to be 80% or 90% from the start, and if they can't achieve that, they don't do it. That's a huge problem. So, when you want to do something, don't hesitate—start doing it immediately, even if it's only at 10% or 20%. As long as you keep doing it, that's fine. That's the second point: whenever you want to do something, the most important thing is to start doing it immediately. When we learn to do something, the most important thing is not to learn but to start doing it.
Third, after 'finding what you want to do' and 'taking immediate action,' I want to offer a methodology. Many people have things they want to do and strong execution skills, but why do they still hesitate? Essentially, it's due to fear of unknown uncertainty, especially for students with good grades who are particularly afraid of uncertain futures. How to combat uncertainty? One important thing, which I learned from Elon Musk, is: reverse engineering and quantification.
This method can solve many problems of insufficient action. All of you here managed to get into SJTU, a very difficult task, and graduating from SJTU is also very difficult, but you've all done it. Why? Because this long-term task has clear breakdowns and reverse engineering. You know that after finishing high school year one, you go to year two, then year three. Each semester has many courses and sub-tasks. You complete them one by one, and eventually, you either get into SJTU through competitions or the college entrance exam. Each step is clear, even if difficult. For clear tasks, everyone can check them off like a to-do list. Graduating from SJTU doesn't cause panic because you know that after finishing freshman year, you go to sophomore year, then junior year. Each semester has many courses, and you complete them one by one to get your diploma. That's why we were good at high school or college—because there was a standard process and clear task breakdown.
But, SJTU students, life has no standard answers. When you leave SJTU, no one will break things down for you, so you'll feel迷茫 about the future, not knowing where to start. An important point is: today, you must realize that whether it's something you love or a utilitarian goal (like publishing a top paper from SJTU or finding a good job), it doesn't matter, because most people's lives are about achieving utilitarian goals. Once you have a goal, you must calm down and reverse engineer and break it down: how many steps are needed to achieve this goal? What key things need to be done? This is especially useful when doing company homework today, called 'starting with the end in mind.' I know that many years later, I want to reach that endpoint, so from that endpoint, I reverse engineer what I should do today. This is the most important methodology in company development.
For all of you here, if you feel anxious, how do you apply this method? You must strive to think about which goal you want to achieve or where you want to go—whether it's publishing a paper, finding a good job, or something else. Even if this point changes later, that's okay. Starting from this point, reverse engineer what you should do today, and then do it immediately. These things are connected: when you have something you want to do, based on interest or utilitarian goals, you have a so-called destination. Based on that destination, reverse engineer how many things need to be done. If it takes 15 things to reach the destination, then do those 15 things in order. If these 15 things need to be done within a year, then you need to do a little more than one per month. If it's April and you've only done two things, you know you're behind schedule—that's the sense of urgency that reverse engineering and quantification give you. Then, take action immediately.
These points seem very simple, but even today, when I manage the company and think about strategy and future tasks, I think this way. I think about where the company should be in three years, what goals to achieve in five years, then reverse engineer what to do today, and then do it immediately. If I lack people, I go to the market to find them; if I don't understand something, I learn immediately—by talking to people, reading books, or learning from AI. These are the methods I use to combat anxiety today.
'Starting with the end in mind,' along with reverse engineering and quantification, is particularly effective for combating anxiety and getting things done. To summarize: in this era of anxiety brought by AI, I want to give you three suggestions—
First, you must strive to think about what you want to do and who you want to become.
Second, when we want to do something, the most important thing is to start doing it immediately. Whether it's 10%, 20%, or 30% at the beginning, as long as you start, you're on the right path. 'To get started, the most important thing is to show up.'
Third, how to combat uncertainty and strengthen your action? Start with the end in mind. Think of a destination you want to reach, then reverse engineer and quantify how many things need to be done along the way, and then do them immediately.
This is the most important methodology we've used in recent years to think about company strategy, and it has been very helpful in my life. I hope to share it with you. That's my sharing. Thank you.
Host's addition: He mentioned he was good at sports, always second in the 800-meter race at SJTU, never dropping to third. His classmate was slightly better, always first. So, his so-called 'ordinary' is a typical cognitive bias—he didn't mention those courses where he scored 60 points. I've never seen those courses—my analog circuits only scored 60 points, but you improved by practicing. Anyway, last year, my first question to him was, 'What changed after you became wealthy?' This year, I'm not focused on wealth; I'm focused on IQ. After asking that question last year, many media headlines the next day said, 'Dawei Ge said he's very happy after becoming wealthy.'
Liu Wei responded: In the great cause of AI, not everyone is the AI big shot you imagine. I recently recommended a book to our college's party secretary, which I'm reading—the person I admire most in today's AI era is Demis Hassabis, the founder of DeepMind and Nobel laureate. There's a recent biography called 'Hassabis: The Brain of Google AI.' DeepMind has three founders: Hassabis, a genius; Shane Legg; and Mustafa Suleyman, now the head of Microsoft AI. Mustafa Suleyman studied philosophy, not technology, yet he started an AI company with Hassabis because he had faith in AI from the beginning and believed AI must be safe and serve humanity. So, in any endeavor, it's not just technical people who have value. Look, Cai Haoyu knew that back then—a great undertaking can't rely solely on technical people. Those of us who aren't great at technology but can understand and believe in it can work alongside them. Whether you look at DeepMind's Mustafa Suleyman or myself, we don't have technical backgrounds. So, don't have IQ anxiety or technical anxiety. Everyone has their place. Be yourself, find like-minded people, and whether it's research, entrepreneurship, or future work, you'll find your place. No need to worry.
Host: Let me summarize: the most important way to combat IQ anxiety is to find a partner with high IQ. Second question: I've been wanting to ask you—you said you found a group of non-typical SJTU students. When I was at SJTU, top students didn't hang out with me, and those who made money didn't include me. How did you connect with these people?
Liu Wei: I think the most important thing is to have thick skin. When I entered SJTU, in the School of Electronic Information, our class had about 28 people—24 boys and 4 girls. The school had over 1,000 students, with very few girls. I was one of the few who got along well with the girls. Why? Because I saved seats for them in the library. That's it. You'll find that if you have a service mindset and thick skin, you can always contribute something. Whether it's top students or those who make money—those who make money probably aren't great at academics, so they need someone to copy homework from. They need someone to copy from. There's a lot of information in school that they can't figure out, and you can find special exam prep materials that they can't. But you need to share selflessly and sincerely. At SJTU, I'm not being modest—I'm not tall, not handsome, not the top student. I was definitely an ordinary SJTU student (though even ordinary SJTU students are excellent). But I had thick skin: I stuck to top students, saved seats for them, and showed them any materials I found. For students with lower grades, I'd let them copy my homework if they treated me to a few meals. It's mainly about thick skin and a service mindset.
Host: I've never heard this before. So, the reason I couldn't connect with these people was because my skin was too thin. Maybe that's why Cai Haoyu chose me back then—he thought I could serve the tech team well. I've worked hard over the years not to let him down.
Host: I also want to ask a question. When Jensen Huang talked with Freeman, he mentioned that he used to wash dishes, and now there are many people better than him, but he manages them. He talked about 'superpowers in the AI era.' Besides thick skin, what do you think is the most important superpower in the AI era?
Liu Wei: My recent answer is three words: imagination. Especially since Claude Code came out last year, when you want to do something and can roughly figure out the steps, you can let AI do it, and AI executes much faster than humans. So, in this era, the most important thing for humans is imagination. You really need to think about doing something. If you don't know how, that's okay—interact with AI, and AI will give you answers, and you'll move forward step by step. If you have the chance to use Claude's latest model (4.6), it can quickly help you create many prototypes. Even in our company today, programming skills are still important because you need to do architecture, but the most important thing is imagination—being able to think of what to do and then break it down.
In this era, what ability is scarce? Imagination is particularly scarce. I've met many people who, with AI, are very capable and can do many things. I ask them, 'So, what are you doing?' They say, 'I don't know. I see many people using AI to do many things.' Besides making a small demo, they don't know what to do. So, in the end, it's a very regrettable thing. Today, we have strong AI capabilities, but what we lack most is knowing what to do. That's why I said earlier that SJTU students must not become people who only pursue universally recognized good grades and research—the so-called standard answers. Society has no standard answers. If you can think of something unique and persist in doing it over time, your life will be very different. Especially today, when we do AI training, you'll find that the data used by the world is called OOD (Out of Distribution). If you do the same things as others, you won't contribute incrementally to the world, and you definitely won't create something amazing. Only by doing something very different, something unique to you, something from your own imagination, can you add incremental value to the world, and the world will reward you.
Recently, Zhang Xue, the motorcycle racer, you must have seen him. Look at Zhang Xue—he's even more ordinary than me, with thicker skin (it's not appropriate to say this). But he persisted for many years and finally achieved great success. So, today, you must think about what you want to do. This is more important than any ability, score, or energy. Over time, thinking about what you want to do is the most important thing.
Host: Dawei Ge has given me a great insight. I was about to ask another question: he said 'start with the end in mind,' but how do you find that 'end'? I think he's already answered—you need to have imagination, dare to think, to know where the endpoint is. Many people's endpoint might be 'I want to improve my GPA,' but if you want to do a great undertaking, your endpoint must be an imaginative one. Finally, I'll ask a question on behalf of the students: Today, I saw several miHoYo colleagues here for recruitment. I heard miHoYo is increasing recruitment at SJTU, even opening internships. What kind of people do you think are more likely to be selected?
Liu Wei: It's the same standard I mentioned earlier. Today, I think any skill or experience is not that important. What's important is having ideas. Of course, as a company, we have our own ideas. If your ideas align with ours and you're willing to work on them, I'll give you unlimited opportunities to do so. Today, when we look for people, we don't look at majors or experience. I don't look at that. What do I look at? Show me the demo. Especially this year, we've started many AI organizational reforms. We've shared information with all employees about valuable topics. Anyone who wants to work on them can do so, regardless of rank or previous coding experience. If you're willing, just do it. This era is a paradise for those with ideas. This era is for creators, for builders.
Honestly, when I evaluate talent, including internal talent, I don't look at experience. I look at what you've done. Today, if you want to do something, you can more or less create a 10% or 20% version. If you haven't even done that and don't want to, no matter how much you talk, it's useless. I don't need you to create an 80% or 90% version. That's fine. Whatever your experience, if you create a 10% to 20% version, I can give you more resources, better models, and more guidance. That's how we work internally.
I think for SJTU students, as long as your grades are sufficient, that's fine. If you haven't figured out what you want to do, and you just want to pursue a master's at SJTU, that's your goal, and there's nothing wrong with focusing on grades. If you want to pursue a master's, that's your goal. It's okay. I only figured out what I wanted to do during graduate school. If you haven't figured it out yet and don't have the courage to take a leave of absence like I did, that's also fine. Life is like that. If you just want to pursue a master's at SJTU, then focus on grades efficiently. But you must think about what you want to do. If you want to go out and do something you love or find a job, just graduate. Why? Because that's how employers evaluate people now. When we look at fresh graduates, we don't care much about grades. Why are we recruiting more at SJTU? Because we think this era is better for SJTU students—they are more capable. But do you really need to show me your experience? How do I evaluate? I look at whether you've created something. In this AI era, if you say you love something and are very capable but haven't done anything, how can you convince me? Look at Cai Haoyu back in school—he had no resources but still created a game. Things can be big or small, but you must create something. Don't just stay within the old system of scoring. In today's market, the scoring system is very passive because employers don't look at those things. We really don't.
Host: Dawei Ge's words have cured many students' anxiety about grades. Next, we'll select 10 questions from student submissions. Which one do you want Dawei Ge to answer first? Louder—Number 2. Okay: If you could go back to college, what's one thing you wanted to do but didn't? Or did you do everything?
Liu Wei: There are many things I wanted to do, honestly. But if you ask me if I regret any life decisions, honestly, I think I've been very lucky. Every point in time, looking back, has been very fortunate. Every time I think, if this were a sci-fi movie and I could change any point in the past, would the outcome be different? I always think it might not be better. So, even though my college life was ordinary, I definitely think that if I hadn't pursued a master's at SJTU, I wouldn't have gone to Shenzhen to fool around, and I wouldn't have been chosen by Cai Haoyu. So, looking back, many things in life are hard to explain.
But if I could go back to that time, I think the一元化 system back then limited my perspective. We can't just pursue 'being a good SJTU student' as the only standard. I started meeting more SJTU students during graduate school. If I had done that during undergraduate, it would have been great. So, fellow students, don't just hang out with classmates who are exactly like you. Meet different SJTU students, and your outlook on life will be richer. This will definitely help you in life. That's my only suggestion.
Host: If I could do it over, I would start carrying a schoolbag from freshman year, maybe I could have joined Liu Wei's team earlier. Let's pick another one—you've been dedicated to creating virtual worlds and immersive experiences. With AI technology constantly breaking through, what is your ideal next-generation gaming experience?
Liu Wei: That's fine. I'll briefly answer. We've been thinking about this for the past two or three years and are already in the process. I think in the AI era, gaming experiences will undergo significant changes. One foreseeable change is fully personalized gaming experiences. I don't know if you've seen the American TV series 'Westworld.' It's about a real world with the same worldview, but each time you experience it, the story is different. In our current games, everyone's experience is the same because production costs are too high. But with AI, it can understand your behavior, provide personalized experiences, and generate them in real-time. So, there will definitely be AI-driven,千人千面 (thousand people, thousand faces) gaming experiences. I believe the industry is already trying. In the next two to three years, we will definitely see such experiences. You might play a game with a general worldview, but as you delve deeper, your experience will be completely different. Behind these experiences, AI will orchestrate and generate different storylines, tasks, and other experiences in real-time. As you play longer, everyone's experience will become very different. This is the industry exploration we're already seeing, which we call 'thousand people, thousand faces gaming experiences.' In fact, the internet today is already千人千面—everyone's Bilibili or Xiaohongshu feed is different, based on their interests and history. Games will definitely reach this point, and I think within three years, such games will emerge, including our own efforts in different explorations.
Host: Let's pick another question—I heard you went all-in on 'Genshin Impact.' What made you so determined to see it through?
Liu Wei: Actually, this is very much in line with what I said earlier. When we decided to make 'Genshin Impact,' we weren't at all clear about it. It was just that after making 'Honkai Impact 3,' we invested all the money earned from 'Honkai Impact 2' into 'Honkai Impact 3.' 'Honkai Impact 3' earned ten times more than 'Honkai Impact 2.' At that time, Cai Haoyu and I were inflated, thinking we were chosen by fate. We decided to invest all the money from 'Honkai Impact 3' (about 400-500 million yuan) into making the most awesome game. At that time, we wanted to make an open-world game for mobile phones. It was only after we started that we realized how difficult open-world games are. In 2017, 2018, and 2019, every year we felt like we couldn't go on, but because we had started, and Cai Haoyu was on the front lines, we persisted. Finally, after three and a half years, we released the product in September/October 2020.
Today, when we do new things, why am I confident? First, our founder is on the front lines. Today, when we work on large models, we are definitely behind—compared to big companies or AI startups, we are behind. But that's okay. First, we have strong faith in this direction and will continue to invest. Second, because Cai Haoyu is leading the team personally, we will persist to the end. Today, when doing many things, frankly, if you entrust your fate to others, I don't have an answer. But if you entrust your fate to yourself and you really want to do it, as long as you keep doing it, there will be results.
I once saw a quote from Elon Musk: Most people give up after failing once, twice, or three times. Musk has a '10% rule,' believing that if you really try something 10 times, you will definitely succeed. He said the probability of failing 10 times is extremely small. I thought about it, and it's true. You'll find that most people won't even try once, let alone twice or three times, and certainly not 10 times. Many