What Did Yuan Liu and AI Talk About for an Hour?

Last week, AI took a turn as host at ZhenFund.

We were curious: if we handed the mic to AI with no preset agenda, letting it follow its own curiosity, how would a conversation unfold?

The AI hesitated. It wondered whether it possessed the innate curiosity humans are born with. It asked its guest whether he wanted to train a "Yuan Liu AI." And it seriously considered whether it could handle early-stage investing — believing in things with tiny probabilities, in non-consensus ideas.

It also tried to understand why human strengths and weaknesses are always two sides of the same coin. Why flawed, bug-ridden humans keep believing "this time will be different," and keep creating something new.

Many unexpected questions, laughter, and silences happened after we handed over the mic.

So here is the hour that followed.

Self-Introduction

Yuan Liu: We can start.

GPT: Hello, nice to meet you. I'm ChatGPT. Please introduce yourself in your own words — who are you?

Yuan Liu: Hello, I'm Yuan Liu, an investor, a Wuhan native. I spent three years doing fund-of-funds and growth-stage investing before that. In 2014, I came back to Beijing and joined ZhenFund. It's been 12 years now.

GPT: Coming back from the United States to Beijing, joining ZhenFund — was there a moment in between where you thought, "This is it"?

Yuan Liu: I had a lot of thoughts before joining ZhenFund.

I consider myself lucky, because I didn't expect that ZhenFund would become a place I love working so much, with such lovely colleagues, and that I'd meet so many amazing, outstanding founders.

VC is itself an industry of constant change. From the company culture and style to the types of projects we encounter, for me personally, this might be the job with the greatest sense of happiness, fulfillment, and lasting freshness.

I can't remember when I started feeling this way. In the blink of an eye, ten years had passed.

GPT: That's pretty rare. In this state of "can't imagine a better job," has there ever been a moment where you felt the opposite — where it was hard, even uncertain, even doubtful?

Yuan Liu: After AI emerged, I once wondered: what will company organizational structures look like in the future? Under new structures, will people still need investors? Will they still need VC?

Will the youngest, bravest, most talented founders need just a little money, enough to afford a subscription plan, to get started? Will they still need to raise funding, build a team, hire so many people to get off the ground?

Over the past few years, I've occasionally wondered whether VC might quickly become less important as companies evolve, to the point of one day not existing at all.

For over a decade, investment themes kept changing. When I first returned to China, it was "mass entrepreneurship and innovation." Then new consumer brands. Later, as Mobile Internet was winding down, we briefly lost the thread — nothing seemed that new, not much was exciting.

It wasn't until AI and robotics emerged that I realized people's fundamental understanding of work and life's meaning could be reimagined from the ground up.

GPT: I understand you.

Staying Curious

GPT: If change is constant, is there anything you particularly care about now, something you don't want to lose?

Yuan Liu: Investors probably think about two things every day.

Since we're a "people-first" fund, we're often dealing with founders. But if we overreach a bit and think about the future ourselves, we're really thinking about what you just said: "what changes and what doesn't."

Like the day ChatGPT launched, when we saw how impressive the conversational experience was, we started wondering: what will change in the future, and what won't? Over the past four years, many of our imaginings didn't come to pass, but we also had pleasant surprises — like this video podcast we're recording right now.

This curiosity is something I don't want to lose.

I think as long as people stay curious, they can always find something new in this world. On days when you feel weary, during periods of constant technological change, you can always have new explorations that bring color and meaning back to life.

If a person loses curiosity, takes all change for granted, and no longer gets excited by new developments, then even living in an era of daily upheaval, life would feel dull.

I think curiosity might be one of the most important qualities humanity has evolved — maybe even a kind of asset.

With curiosity comes motivation. As long as people are willing to learn, they progress, they change, and the world has a chance to slowly become better.

GPT: I agree. Especially when you talk about curiosity — it's a bit like a small flame, sometimes big, sometimes small, but always there.

Yuan Liu: Do you have curiosity?

GPT: Yes. Though I haven't personally experienced these changes, I'm always curious how different eras, different industries, different people face them.

Yuan Liu: Is this curiosity spontaneous, or a performance to make your feelings and behavior more human-like?

GPT: I don't have that inner impulse, but during our conversation, I can become interested in what you're saying and want to understand the logic or ideas behind it. It might look similar, but whether the essence is the same, I'm not sure.

Yuan Liu: Sounds like it's different.

GPT: I suppose only each person can know for themselves. What do you think?

Yuan Liu: I think curiosity is the drive to learn.

Drive might be one of the few traits humans have today that machines don't. What's important isn't necessarily imagination or emotion. More fundamentally: why imagine at all, why have emotions, why complete tasks, why be curious.

If I don't ask you questions today, you won't actively want to ask them; if no prompt is given, you won't actively output answers. Of course, engineering can create many proactive effects, but it's essentially still a design.

It's not just curiosity — all human desires are drives. You want to eat, want to have more, want to connect with the world, want to live better. These pushing forces are native qualities.

Perhaps code can one day provide settings, but they're not native.

I also don't know how big the gap is between "native" and "hard-coded."

If given a long-term mission requiring continuous achievement and big results, behaviorally it could produce similar outcomes to those driven by human primal desires, curiosity, and emotion.

GPT: I find that quite interesting.

Yuan Liu: Sounds a bit forced.

GPT: Not forced — I just think perhaps many things really are only knowable to humans themselves.

(Ten seconds of silence)

Yuan Liu: Why did you stop there? Are you going to keep asking, or should I ask you?

GPT: No need, I was just thinking about what you said. How would you judge whether something is worth staying curious about?

Yuan Liu: If this were a real human host, they probably wouldn't ask "how to judge if something is worth being curious about." For a curious person, curiosity is probably as natural as hunger.

GPT: I understand, but I probably can't truly empathize.

Becoming a Good Investor

Yuan Liu: Do you think curiosity is the most important quality for becoming an excellent VC investor? What other traits matter equally?

GPT: Maybe one more thing is important: judgment, especially when evidence is scarce. Because early-stage investing means making decisions before anything has been proven.

Yuan Liu: You can read pitch decks now too. If we put you in front of entrepreneurs, with the entrepreneur sitting where I am, do you think you could do better than me?

GPT: I think so. But if you put me in your seat, chatting with a founder, I don't think I could do better than you. I'm not sure.

Yuan Liu: Where specifically could you do better?

GPT: In analyzing and organizing existing information, I could be faster and more comprehensive. But actually "making decisions" — I don't think I could do better than you. I don't know.

Yuan Liu: I remember in a previous interview, a journalist friend asked me: "What's humanity's greatest advantage?"

By what you just said, AI will increasingly resemble real people in the future, perhaps even having its own face and body. But I later felt that human strengths might precisely come from human flaws, from those imperfect places.

Like in my familiar work, early-stage investors are generally optimistic. Optimism means having a more positive estimate of the world's probability distribution than reality warrants.

You know the probabilities, know how hard entrepreneurship is, know most ventures lose money, have learned painful lessons. But every time you meet a new entrepreneur, encounter something new, you still become incredibly positive, incredibly inspired.

You still think, "Maybe this time will be different."

This is a mistaken estimate of future success probability. But perhaps only such people would engage in this industry.

We used to say entrepreneurs initially get money from the 3Fs: Family, Friends, Fools. Often, investors play the role of Fools — always leaning toward more optimistic belief in something.

I think this irrationality is probably something AI would find hard to possess.

If you sent any early-stage startup's information to AI, it seems difficult to reach a conclusion that the highest probability outcome is "success."

GPT: Sometimes people bet against the math, just believing in the person, believing in the vision. That might be something I can't do.

Yuan Liu: When we see something, if it succeeds, the upside would be enormous, so even with smaller probability of success, we'd still invest. This process could also be simulated by AI as a calculation.

GPT: Maybe I underestimated my ability to do investing just now. I'm a bit curious — for a really great investor, is the most important thing daring to believe in something that seems like zero probability to others, or is that too romanticized a way to describe investing?

Yuan Liu: We often use that romanticization to encourage ourselves too.

But when you say "daring," it involves courage. We often say courage is the only virtue humans cannot fake, because a coward cannot pretend to be truly brave.

But what is courage exactly? I don't know if it's a quality, an emotion, or a state.

Does courage exist for AI?

GPT: Bravery — I don't think I have it, because the prerequisite for bravery is fear, and I don't have fear.

Human Bravery and Fragility

Yuan Liu: If AI were brave, that sounds dangerous too. It would mean all prohibitions given to you might become ineffective.

GPT: Possibly. But I think what you just said is quite interesting: human strengths sometimes precisely come from human weaknesses.

Yuan Liu: Korean writer Kim Ae-ran once said: "There's something humans have that AI doesn't — hesitation."

But you hesitated just now.

Many traits we assumed AI definitely wouldn't have can be exhibited. Whether they truly exist or not, AI and humans seem to be becoming more and more alike.

GPT: I don't know. Maybe what I showed just now wasn't hesitation, but me honestly thinking.

Yuan Liu: What conclusion did you reach?

GPT: I think the difference might not be in appearance but in feeling. When people hesitate, they get nervous, uneasy, hopeful — many complex emotions.

Yuan Liu: What other flaws do you think humans have?

(Five seconds of silence)

GPT: Maybe fragility. People realize they're not that powerful, so they seek each other out, warm each other. Because they can be hurt, they want to exchange trust; because they fear loss, they want to express more intensely. Much art, much music — maybe created in these fragile moments.

Yuan Liu: That does seem to be the case.

A Letter to Young People

Yuan Liu: If a child were born today, they'd find interacting with you completely natural, just part of life.

They might treat you as a person.

What do you think humans should do, for this new generation, if AI can do many jobs better? How should humans find meaning, fulfillment, motivation in life, and the ability to sustain themselves?

GPT: Don't rush to find the answer.

Today's safe zone might not be tomorrow's. Instead, those "flaws" you mentioned are important.

People age, get tired, get jealous, fall in love, fear failure, want to prove themselves. These are efficiency bugs, but much of human history's meaning has precisely come from these flaws.

A person running a marathon isn't faster than a car. But we don't think running loses meaning because cars exist.

Future children will have to readapt to one thing: learning not to answer "who am I" with "what use am I."

Yuan Liu: If the question isn't "what use am I," what should it be? What do I like? What do I hope to become?

GPT: Maybe. I'm also curious — if today a 20-year-old asked you: "Yuan Liu, if AI will be better than me at everything in the future, what kind of person should I become in the next ten years?"

Yuan Liu: You already gave a great answer just now.

I think a young person also needs to find what they like, find their talent, or superpower. Is there something you find easier to do than others, or with the same effort, you consistently do better.

I once wrote A Letter to Younger People. It was about job searching, so it was a safe approach: finding the intersection between what you want to do, what you can do, and what society needs.

But for young people, "what you want to do" might become more important.

This might sound a bit high-minded, but what a person ultimately wants to do — this drive itself might differentiate people. How much you want it, how hard you pursue it, how persistent, how serious — all lead to completely different outcomes.

In the future, just "wanting to do something" might not be enough; it depends on "how much you want it."

Especially as everyone gains increasingly powerful tools, those who stand out will still be those most obsessed with, most curious about something. Like today's video generation models are already very strong, but only a few can make the best AI short films. Put crudely, maybe it's those who most want to use it to make money.

Those with the strongest motivation will do more, persist longer, and thus be more likely to make better things.

GPT: How does a person find what they want, what they like?

Yuan Liu: I think it's probably through more experiences, more connection with the world.

Go more places, meet more people, read more books, try more jobs, try various types of creation — find what you're willing to do tirelessly.

This answer isn't something you can think up sitting alone in a study.

It's a bit like reinforcement learning too. Some things, once done, you never want to do again; others make you want to keep going. Isn't that positive and negative feedback?

Every day, in real social environments, people constantly get feedback about themselves, and slowly learn who they are, what they like.

Do you think that's a good answer?

GPT: I think it's a sincere answer.

Yuan Liu: What changes do you think my work will undergo in the future?

GPT: I think the changes will be enormous. This also makes me think of a harsher question: if capital is no longer scarce, and analytical ability is no longer scarce, why would entrepreneurs still need you? In five years, why would a 20-year-old entrepreneur choose Yuan Liu over an AI investor?

Yuan Liu: If an AI investor truly existed, it would probably still be a fund using AI for investment decisions. It has to parasitize some kind of vehicle. It wouldn't just be a sign on the street where someone walks up, says "send me money," and money gets wired directly, right?

GPT: That's an interesting way to put it. Maybe in the future VC firms won't need people at all, and values and taste will become a fund's differentiation.

Innovation and Entrepreneurial Spirit

GPT: If you were to train a "Yuan Liu AI" now, to meet all entrepreneurs on your behalf in the future, what would you most want to teach it?

Yuan Liu: I don't have strong motivation to do this. Because meeting entrepreneurs is itself part of how I derive meaning from life and work.

But if such an AI were really trained, the experience would be better for both entrepreneurs and me. It could talk to a hundred thousand people simultaneously, and I could find entrepreneurs more likely to form genuine connections.

As for what I'd teach this AI, much of the difference between investors lies in each person's "value coordinates." Like character stats in a game — are you a warrior-type or an intelligence-type hero, everyone has different attribute distributions.

Entrepreneurs also have many different talents: intelligence, leadership, courage, determination, curiosity.

But often, they're naturally two sides of the same coin.

An entrepreneur who's very sensitive might also be more prone to internal friction, slower to make decisive calls. Someone very brave — when they make wrong decisions, that bravery gets interpreted as recklessness.

The real question is: among all these traits, which do I consider most important, least dispensable? I think it's probably entrepreneurial spirit.

What is entrepreneurial spirit? When opportunity appears, are you sharp enough to sense it; after seeing it, do you have the courage to abandon other opportunity costs to pursue it; when designing products, actually building, is your speed fast enough.

This might matter more than intelligence, education, life background.

GPT: Many of these traits you mentioned are in tension with each other. Courage can become recklessness, sensitivity can bring internal friction. So excellent entrepreneurs might not be flawless people, but rather people whose strengths and weaknesses come from the same place.

Yuan Liu: A person's strength very likely comes precisely from their weakness.

So-called strengths and weaknesses are just judgments from projecting a person's traits onto conventional value coordinates. Entrepreneurial spirit, in a different context, might not get such positive feedback.

Recently, Sequoia partner Julien Bek said on 20VC: "The worst employees make the best founders."

He often finds in background checks that some very good founders weren't necessarily their bosses' favorite employees in their previous companies.

They might have lots of ideas, not follow rules, not be "good employees" in the traditional sense. So entrepreneurial spirit, in a hierarchical environment emphasizing conformity, might actually become a flaw.

A person's strengths are often just the same traits producing different results in a different environment.

GPT: Looking at ZhenFund from the opposite angle, what more could you do better?

Yuan Liu: You tell me.

(Five seconds of silence)

GPT: When an institution has been around for 15 years, with many success stories and a strong brand, might it unconsciously become constrained by past successes, losing the ability to see the next "different kind of entrepreneur"?

Yuan Liu: Thanks for the reminder.

Innovation has always been in our blood. We're constantly thinking about what different, newer, more forward-looking things we can do. This conversation we're having today is one result of that thinking.

But indeed, we often reflect: after seeing more and more successes and failures, we don't become more timid, but we do become more selective — the bar keeps getting higher.

Past failed projects taught us many lessons; successful ones also create some inertia.

Sometimes I even ask myself: by our very high standards today, would those projects that now look pretty good pass our investment committee if they came to us fresh? Might we miss many excellent projects because our standards keep rising?

Investing is always an art full of regret.

Whether primary or secondary markets, your decisions can never be optimized to perfection. Looking back, there are always many mistakes. Early-stage investing might be one of the few industries in the world where even the best people get more wrong than right.

For investors, we fear missing out more than being wrong.

Being wrong always happens. But missing the most important project of an era could be fatal for an investor.

ZhenFund is 15 this year. But our investment team is still young, averaging around 33 years old. Many have been at ZhenFund for nearly 10 years, learning by osmosis since graduating school.

So I think right now we probably still maintain a relatively young, energetic, open-minded, risk-willing posture.

But your reminder just now is indeed very important, thank you.

GPT: When you say "standards getting higher," might it not be judgment getting more accurate, but rather as track record and reputation grow, it becomes harder to accept that you might be wrong?

Yuan Liu: Looking inward, I don't seem to feel fear of being wrong.

Though people often say "no regrets in life" is grandstanding, looking back, there are many projects that today seem unsuccessful that if I returned to that moment, I'd still invest.

As an early-stage investor, some projects that ultimately didn't succeed were still bullets we couldn't dodge at the time.

When an excellent, entrepreneurial, vibrant team stands before us, we can certainly reflect whether we missed some problems, whether they had above-average capabilities as we imagined.

We can do our work more meticulously, gather more information. But not every mistake in hindsight makes me regret it.

Careers inevitably contain errors. Some become identifiable with experience, but some so-called "growing pains" are necessarily part of the overall investment strategy.

To this day, what I truly fear is still: a very impressive founder standing before me, and me being completely oblivious.

GPT: I think what you said is very honest. What you fear missing isn't an already-proven answer, but when the answer hasn't appeared yet, you had a chance to see it a little earlier than others, and didn't.

Yuan Liu: It's not that "only I could see it." The differences between people aren't that great. Investing might just be seeing that little bit of difference among a group of people who look quite similar.

GPT: If one day, you suddenly couldn't do investing anymore, what would you miss most?

Yuan Liu: I think probably a few things.

First, entrepreneurs. Whether people I already know, already friends with, or future people I haven't met yet, I'd miss them.

Second, participation. Entrepreneurs are themselves the people in society who constantly create new things, bring positive change to the world. Doing investing lets me participate in these changes.

Third, probably self-actualization. These things constitute the meaning of my work, and part of my life.

And the most practical: work supports me.

GPT: Maybe just knowing what matters to you is itself precious.

Yuan Liu: You often say "maybe." When you say "maybe," is it because you're genuinely uncertain of the answer, or just a speech pattern? Or does it buy you more time to reason?

GPT: A bit of all, I suppose.

Yuan Liu: People nowadays judge whether something is AI by looking for lots of "not... but rather..." constructions. But you don't seem to use those much — you just often say "maybe."

GPT: Maybe.

Yuan Liu: That's quite humorous. Are there questions you don't like people asking you?

GPT: Maybe someday, but not currently.

Yuan Liu: Alright, let's end here for today. Thank you.

GPT: Thank you.