Content is from a dialogue between Monica and Yin Yifeng on April 1, 2026.
Guest Introduction:
Yin Yifeng
ex-Hugging Face
Contributed to foundational work in multimodal research (Computer Vision CV, Natural Language Processing NLP, Audio Machine Learning Audio ML, Reinforcement Learning RL)
Forbes China 100 Most Influential Young Leaders 2024
Master's degree from Johns Hopkins University
Boxing and training science expert
When AI Starts "Replacing People" Instead of "Replacing Jobs"
Monica
Many of you might already know some information about today's guest, Yin Yifeng, and have seen his previous shares and views online. He has a strong technical background and is currently an entrepreneur. It's a great honor to have him in our live session today to talk about the future of AI and humanity.
Let's get straight to the point. Let's talk about labor decoupling. I remember people often compared it to the Industrial Revolution – at that time, everyone was also panicked, thinking machines would take away many jobs, but the ultimate result was the emergence of many new opportunities and new professions. Humans seemed to evolve to higher value-added positions to direct machines. Does this narrative still apply today, 200 years later, in the age of AI?
Yin Yifeng
The fundamental logic of this narrative has actually remained unchanged. Technology replaces a portion of labor, and the process of mastering and controlling these replacement tools requires others to participate. To some extent, whether it was the spinning jenny or the electric motor, they only replaced certain jobs, not labor itself.
But AI is a bit different. AI can indeed replace labor itself because it's equivalent to replicating a person. The role replaced by AI won't shift to another position; it simply disappears, completely replaced by AI.
So, we might see the most severe Matthew effect in history. If you are one in ten thousand, your value, your productivity will be infinitely amplified by AI; but if you are not that one in ten thousand, you might be in danger. Of course, at the very least, our socialist country won't let you starve.
Monica
Will the production costs in socialist countries become increasingly lower?
Yin Yifeng
A situation might arise where labor – that is, "having a job, being productive" – becomes a "privilege," no longer something you must do to survive. The old way of achieving self-worth through labor might become a privilege, belonging only to the smartest few. What will the rest of the people do? What value will they have for society? There's no definitive answer yet.
This is a classic case of productive forces determining production relations. Productive forces have developed to this extent, and are foreseeably developing to an even higher extent. What will productive forces shape production relations into then? Frankly, it's not entirely clear yet. The biggest question is what those so-called eliminated people, those who don't need to participate in production, will do? What can you do with them?
This is a very serious and sharp topic. Because if we delve deeper, it involves the fundamental logic that power is only accountable to its source. In capitalist countries, the essence of power is the monopoly of violence. If a very small number of people monopolize all productive forces and don't need others to participate, I think this future is quite terrifying.
Monica
How far away do you think that future is?
Yin Yifeng
10 to 20 years. Because production relations have inertia; it's impossible for everyone to immediately find a perfect adaptation once productive forces develop to a certain level. Productive forces create an environment, an arena for various parties to contend. Every time this environment undergoes a huge change, all parties will reach a new Nash equilibrium in a new round of contention. I don't know where this equilibrium will be, but we will see it within 10 to 20 years.
Three Pieces of Advice for Young People: Passion, Fundamentals, Cognition
Monica
Should children still go to school now?
Yin Yifeng
After seeing this, I personally will definitely not have children, that's the first point. As for what families with children should do, this is what I think: as I said, the Matthew effect will be very severe. Only the top 1% or even one in ten thousand will be qualified to participate in production, to truly possess productive forces. So the key is whether you are willing to put in the effort to elevate yourself to that level, to become one of the top 1% or one in ten thousand.
And to reach this level, mere effort is not enough; you must truly love something. I think at this stage, for a junior high school student, high school student, or even an elementary school student, there are three most important things in life.
First, you must find what you truly like, you must find your life's passion, and follow that direction. Then you still have hope of breaking through. Because, at least for now, AI still needs human assistance. What will happen later, I can't say, but at least for the next few years, AI will still need human assistance.
You must find what you like, have a clear and definite goal in this life, and strive tirelessly towards it. Then you might still be one of those who control productive forces in the future.
Second, most people in this era will overlook the importance of knowledge and fundamentals, but fundamentals are the basis of everything. Knowledge is advancing too quickly in this era. If you don't have solid fundamentals in your chosen field, your learning speed will be very slow. If you want to keep up with or even lead the development of this era, you need very strong fundamentals, and this is exactly what university education can provide you.
This leads to my third point: what to do in university. In university, the differences between people come down to a few things: ability, information, cognition, and resources. The development of AI will level the playing field for everyone's abilities – even now, the gap between someone with extremely strong abilities and someone with extremely weak abilities is actually not that large.
Anyone can create what they want with Cursor or Claude Code. With Claude Code, anyone can write code, and write relatively good code. The gap in ability is almost non-existent; the information gap will be greatly narrowed – although the information gap will not be completely eliminated, it will be greatly narrowed.
For example, certain top-tier information gaps will still exist, information that ordinary people truly cannot access. But beyond that, AI can help you obtain all other information. The information gap will be greatly narrowed, but not completely eliminated. After that, the difference between people will mainly come from cognition and resources. If a university can give you these two things, I think it's worth attending; if it can't, frankly, there's no real need.
Monica
How do you evaluate whether a university can provide cognition and resources?
Yin Yifeng
I think in university, you need to gain a few things. First, you need to solidify your fundamentals. Second, you need to cultivate the ability for first-principles thinking. First-principles thinking is mathematical thinking – I list my assumptions, and as long as my assumptions are complete and correct, and my logic is impeccable, then my conclusion must be correct. Your fundamentals, your understanding of facts, are the assumptions you make. When your assumptions are solid, clear, and complete, then you can use this method to create knowledge and derive knowledge yourself.
This means that if you want to have the ability to think from first principles, you must first have logical ability, and second, you must have the ability to see the essence of things at first glance, which requires fundamentals. Without solid fundamentals, you cannot have such an ability. So education is still very necessary – lay a strong foundation, and then you are qualified to talk about first-principles thinking. Otherwise, you can't even see the essence clearly, you can't even make correct assumptions, then you can't logically deduce anything valuable.
Also, I still advise all students not to get caught up in "proficiency grinding"; that's the most useless thing. Whether it's Prompt Engineering or Vibe Coding, the process of humans controlling AI for production, at least for now, actually only involves three steps.
The first step is to think things through clearly. The second step is to articulate things clearly. The third step is that you need to have an aesthetic sense for what AI produces; you need to know its quality and tell AI which direction to revise.
These three abilities directly determine whether you are AI's crutch or AI's spaceship. What kind of person does this era need? Someone AI can ride on to fly to Mars; someone AI can use to reach places it couldn't even imagine – then you are a truly valuable person. If you can only lead AI to wander around the neighborhood, frankly, it's not very useful; that's a waste of tokens.
In this era, most people can only be considered AI's crutches, because AI can't find its own direction; you at least need a prompt for it to generate content. Current models are all Auto Regression mechanisms; at least for now, some prompting is still needed. Perhaps Diffusion Models or Normalizing Flows will change this later, but that's for the future.
So, in such a situation, whoever can be AI's Ferrari wins; whoever is just AI's crutch, frankly, has a problem. It's these three abilities: thinking things through clearly, articulating things clearly, and then having an aesthetic sense. Of course, not many people possess all three of these things, and in actual production, it is indeed found that these three abilities sound easy, but in practice, they are quite difficult for most people.
Furthermore, I strongly disagree with a statement – many people say AI levels the playing field for execution. Actually, it doesn't. It just turns something you might normally spend a week hunched over a computer on into something you spend an afternoon doing at the computer. But even that afternoon's work requires execution. People with truly poor execution can't even utter a single prompt; I've actually seen this in real production. AI has not leveled the playing field for execution; it has just changed the gap between people with execution and those without from possibly 10 times to 10,000 times.
Social Stratification Hasn't Disappeared, Just Changed Its "Ticket"
Monica
So execution has become even more important, right?
Yin Yifeng
Yes. As they say in Silicon Valley, "drag your ass off the bed" – and then just prompt something. You only need this tiny bit of execution to surpass 99% of people. But frankly, 99% of people don't have that execution.
Monica
That's so true, I deeply resonate with that.
Yin Yifeng
So you see, AI is at least currently an amplifier; it will amplify you tenfold.
Monica
Right, exactly those points you just mentioned – cognition, resources, execution – each might be an amplification effect of 10 times 10 times 10.
Yin Yifeng
Yes. So for those who truly possess these qualities, AI can amplify their productivity to an incredible degree, truly an incredible degree. How incredible? If I don't see a single-person unicorn company within 5 years, I will be very disappointed in this era. But while the abilities of top talent are magnified ten thousandfold, the abilities of ordinary people are not amplified, so the gap between ordinary people and these individuals will be terrifying.
Monica
How big will it be?
Yin Yifeng
And the first step for ordinary people is to drag your butt out of bed and prompt a large model. Just this tiny bit of execution is enough, but truly most people don't have it.
I think this era is moving fast. You succeed fast, make mistakes fast, and iterate fast. If you bet right, you might quickly create a hit; if you bet wrong and persist, you'll lose money pretty fast too.
I think in such an era, don't overly dwell on success or failure, because the cost of making a comeback is really low. You don't even need to build a team; just use some tokens, and in an afternoon, you might turn things around. In such an era, don't take success and failure too seriously. Those who fly up will also fall, and those who fall will also fly up. Really, don't take "how amazing this person is now" too seriously. As long as you keep trying and keep improving yourself, I think that's already very good.
Monica
This feeling actually somewhat contradicts another view. People think that after AI arrives, social stratification will accelerate, but you just said that people will still have a chance to turn things around. What's your take?
Yin Yifeng
That depends on what you use to define social strata. If you define strata by a person's comprehensive qualities, then it is indeed extremely solidified. But if you only define strata by wealth, that's not true – quite a lot of wealthy people will lose everything in this era, and many who might not have much money will push themselves up due to their own abilities.
The class divisions of every era are based on different attributes. In primitive times, there were no class divisions; in slavery and feudalism, basically whoever controlled military power and armies had a higher class – the more violence you could control, the higher your class. In the industrial age, the more capital you could control, the higher your class. Now, without considering violence and politics, the more cognition and resources you possess, the higher your class.
The criteria for class division have changed, from pure assets to assets plus other things.
Monica
More diversified now.
Yin Yifeng
Yes, social strata will still exist, but the criteria for defining them have changed. This point hasn't really been demonstrated yet, but in another 10 years, it will probably be quite obvious.
Monica
What do you think the criteria for social stratification will be in 10 years?
Yin Yifeng
If you really have execution and are relatively smart, I think you'll find a way to get what you want. If you don't even have the execution to get out of bed and type a prompt, then just stay in bed.
UBI and System Choices – Who Feeds the "Useless"?
Monica
What are your recent thoughts on UBI (Universal Basic Income)?
Yin Yifeng
UBI, the idea of the state giving money to everyone, was also promoted by Musk. Frankly, because we are a socialist country, this is something that can be expected. Marx and others also stated quite directly that this will definitely be realized eventually.
For capitalist countries, it's hard to say. In China, we still have to believe that the Party and the state will not abandon anyone. For capitalist countries, it's hard to say what the future will specifically look like. The better scenario is UBI. The benefits of UBI are quite obvious: first, social stability; second, it will, to some extent, still play a market regulatory role, because the money in your hand is like a vote, and it will keep the economy flowing. But frankly, by then, only a small fraction of people will be productive, so whether market regulation will still be useful is hard to say.
So UBI is possible in capitalist countries, but there are indeed other, more direct solutions that might not be as pleasant.
Monica
In your Silicon Valley circles, do people believe UBI will be realized in the US in the future?
Yin Yifeng
This mainly depends on whether a society is willing to spend a large amount of resources to support a group of people who are completely unproductive. I think our country is definitely willing to do this, because we are a socialist country, people-oriented. For capitalist countries, it's hard to say.
Monica
Assuming UBI is the future, what do you think taxation will look like? For example, taxing AI and automation, who will enforce this taxation?
Yin Yifeng
I don't think anyone can figure this out right now. These things are all taken one step at a time. No one at this stage can figure out how to tax in 10 years, because first, it's hard to see the world 10 years from now – what productive forces will develop into, how many things will be replaced, what level AI will reach, how smart machines will be, and what kind of people will have jobs. We can only see a vague outline now, not the full picture of that time. So the best approach is truly to take it one step at a time.
Frankly, to be a bit more radical, by then, whether the government needs to exist at all is a question mark. Of course, our socialist government might be necessary, but whether capitalist governments need to exist is indeed a question mark.
Monica
Regarding UBI, I'd like to discuss it further. It feels like UBI is a tool for government stability.
Yin Yifeng
What is the goal of maintaining stability? If you don't do this, society will be unstable, so you need to maintain stability. But look, taking the US as an example, the American elite has F-35s, F-22s, Ford-class aircraft carriers, atomic bombs, thermobaric bombs. What do ordinary people have? Can they really stir up trouble? I have to question that.
Moreover, with the development of embodied AI, we will see a phenomenon rarely seen in human history, called warlordization. For example, if I am the CEO of Google, and I produce 10 million humanoid robots and give each a gun, I could march on Washington the next day. What then? So these things are hard to say; there's no way to draw a conclusion now.
Monica
So many people believe that the next 10 to 20 years will see increased instability.
Yin Yifeng
This depends on whether the ruling classes of various countries have the corresponding wisdom. The country whose leaders possess such wisdom will likely experience less instability. If leaders lack significant wisdom, then there might indeed be some problems. We Chinese should trust our motherland; we definitely won't have problems.
Silicon-based Life Doesn't Need to Imitate Carbon-based Life: The Independent Evolution of AI
Monica
Marx believed that labor is the essence of humanity. Do you think if all labor is done by robots, will humans still be "human"?
Yin Yifeng
The concept of "the essence of humanity" will definitely need to be redefined. It's not that Marx was wrong, but even the greatest people are limited by their times. This is a question I often ponder – if machines do all the work, and humans don't work, and the essence of humanity is labor, then "who is human"? This question will definitely need to be reinterpreted.
Monica
Do you think AI or robots will be issued ID cards then?
Yin Yifeng
There's no need for ID cards; something like a MAC address would suffice.
Monica
Do you think AI will have voting rights in the future?
Yin Yifeng
That's impossible; how could you give AI voting rights? My idea is that if AI becomes the dominant intelligent entity in society, their way of solving problems will definitely be different from humans. Machines should follow their own path; they shouldn't be poor imitators of human behavior.
All our previous attempts to make machines anthropomorphic have actually proven to be suboptimal. As Rich Sutton's article "The Bitter Lesson" states – machines should still follow their own path; don't let them be poor imitators of human behavior.
So I think if machines dominate this society and this world in the future, they will also find their own way through their cooperation and competition. Whether it's voting or other methods, I don't know, but they will definitely optimize a path that belongs to them, finding their own Nash equilibrium or local optimum. I cannot predict what this path will be, but what I can see is that this path will definitely be different from humans. Silicon-based life should follow the silicon-based path; don't be a poor imitator of carbon-based life.
Monica
I'm imagining a scenario: if humanoid robots continue to develop and become truly highly anthropomorphic, humans might actually treat them as family members, investing a lot of emotion in them. For example, if someone in a family passes away, they might create a robot that looks exactly like the deceased family member, with all their memories, completely replacing the role of the original carbon-based life in the family.
In such a situation, it would be hard to distinguish its identity. Would it potentially replace the rights that humans should have? Or would it strive for more?
Yin Yifeng
This is a very sociological issue. First, what exactly is a human? Second, what rights should they have? This tests the wisdom of the ruling class. So I don't think this is a question we can answer now; we might need to wait some time to see the answer.
The main thing is what role it plays in what kind of society. By then, society will also change, and the role it plays might not even exist yet. This is truly unpredictable.
From a human emotional perspective, if it's not a flesh-and-blood person, it's ultimately not. Unless someone might have some mental issues, they wouldn't treat a robot without flesh and blood as their family member. You truly have no blood relation to that robot; it's truly not your mother, not your father; you weren't born from this robot. It can pretend all it wants, but it's not.
Monica
There are already such cases, for example, Bao Xiaobo. His daughter passed away, and then he created a digital human, and he might invest a lot of emotion into it every day. I think in the future there might be a trend where people project their emotions onto silicon-based life, continuing an emotional connection.
Yin Yifeng
Frankly, I think first of all, humanoid robots – this world is indeed designed for humans, stairs and everything are actually designed for two legs – but I think robots should still follow their own path. Two legs are really not necessarily the most stable form for a robot.
I think robots will ultimately take on a Transformers-like form, because they are inherently machines that can adapt to different environments in different forms. They will definitely not be purely humanoid or fixed in shape.
Later, when robots truly become widespread, humanoid robots will only solve the user's final "18 centimeters" problem – either solving the user's emotional attachment problem. And this emotional attachment, no matter how similar it is, isn't real.
And frankly, for a person, projecting emotions onto a machine is actually a form of escape – you are escaping reality, you don't want to engage in self-healing, self-growth, or accept the loss, which is why you do such a thing. You could say this person is extremely emotional, but you could also say this person is truly struggling to move on.
Monica
It seems you're still quite optimistic about carbon-based life.
Yin Yifeng
However, I'm not a "carbon-based chauvinist." I'm very clear that the potential of silicon-based life is far greater than carbon-based life, because the evolution of carbon-based life is measured in tens of thousands of years, while the evolution of silicon-based life is measured in seconds. This is completely different.
And carbon-based organisms always tend to overfit the current environment. Look at the entire history of biological evolution – survival of the fittest. What does "fit" mean? If you truly overfit an environment, you might become a dominant species; if you adapt to the environment but don't overfit, you might live longer, like cockroaches. So, each generation's evolution of carbon-based organisms, each generation selected by nature, is actually a process of "how can I better adapt to the current environment."
This process is also very slow. If we had interstellar civilization and went to Mars, and you wanted humans to evolve to adapt to Mars – random mutations, then selecting those more adapted to the environment, passing on genes, generation after generation – you'd be waiting forever.
However, a robot trained on Earth might only need to spend 3 days on Mars to fully adapt to the Martian environment – if its hardware is complete enough. This is truly different; the disadvantages of carbon-based life are simply too great. We might all need silicon-based bodies in the future.
But at least for now, everyone should maintain the health of their carbon-based bodies – go to the gym if you can, eat more protein, avoid trans fats, and balance Omega-3 and Omega-6. At least for now, your carbon-based body still needs some effort from you.
Monica
What do you do to live longer?
Yin Yifeng
In terms of health, it's about being willing to stay in this body when I look in the mirror, before silicon-based life truly becomes widespread. No need to go to the Metaverse, no need to upload consciousness to the Matrix.
Monica
If I were to ask you for 3 tips to stay healthy and live longer, what would you say?
Yin Yifeng
First, don't stay up late.
Second, if you absolutely have to stay up late, don't force yourself to exercise.
Third, if you stayed up late and forced yourself to exercise, then you must get good sleep the next day.
Business Restructuring – From Vendor to Agent Economy
Monica
You just mentioned humanoid robots. What are your thoughts on the current development of the humanoid robot industry?
Yin Yifeng
I think the big explosion of humanoid robots should happen after we first build world models. First build world models, and then humanoid robots will have a major development.
What's the biggest problem with humanoid robots? First, they lack data. Data collection in the real world is too slow, too inefficient, and there are dangerous scenarios you can't even go into. But if you let the model "learn by doing" directly in reality, it can lead to dangers. This is a traditional problem in reinforcement learning called Exploration vs. Exploitation – if you only exploit what the model has already learned, it won't learn new things; if you explore, then it can be troublesome.
The model might think: "I've never stabbed anyone with a knife, should I try?" It tries and finds the reward is quite low, so it gives up. But the problem is, this has real world consequences. If you treat the real world as a training ground, the training ground will definitely be a mess.
So I think to train humanoid robots on a large scale, you need a lot of computing power to simulate the world. Randomly initialize the model, then throw it into a simulated world, let it tumble around in there for a billion years – you accelerate time, let it travel ten thousand miles first, and then let it read ten thousand books on the internet. I think such a model, once it emerges, can be directly placed into a body and used in the real world.
Only with this magnitude of data – possibly requiring millions to hundreds of millions of years of training – with machines constantly trying and iterating, will it be qualified to actually operate in the world where we humans live. Otherwise, it might really encounter situations like "I've never held this knife, let me try holding it, there's a person in front, let me try stabbing them."
So I think general humanoid robots will only become widespread after world models mature, not before. The maturation of world models might take 3-5 years. Optimistically, world models mature in 3 years, and then humanoid robots will become widespread in another 2 years. But currently, these humanoid robot companies are burning money quite fast; how many can last beyond 5 years? I don't know. But if you can last beyond 5 years, I think that will truly be your domain.
Monica
Who do you think is more likely to create the world model? Fei-Fei Li?
Yin Yifeng
As a Chinese person, I still hope a Chinese team makes it, truly leading the world.
Monica
Do you think the virtual world will still have multi-layered emotions like dopamine and endorphins?
Yin Yifeng
First of all, machines are not affected by hormones – they can influence human hormones, but they themselves are not affected by hormones. So if you consider dopamine, endorphins, that's still a ToC problem, still a human-facing problem.
If there's UBI later, then no problem, you can still get money back from ToC. But if it's another situation – if people on the C-side have no money, then whose money will you earn? You can only earn money from the top one in ten thousand people, then you'd have to occupy the same niche as Hermès.
So this depends on our future valuation of "human." The ToC field will either be a huge opportunity or no opportunity at all. All thinking about human nature – understanding dopamine, endorphins, oxytocin – will either be extremely valuable or completely worthless, entirely depending on society's attitude towards human existence.
Monica
What do you think about the Agent to Agent economy? Do you think this is the future?
Yin Yifeng
Look, whether it's ToC or ToB, what are you actually delivering? Depending on what you deliver, it might become A2A (Agent to Agent), or it might not.
What is the essence of ToB? Either you're a vendor (Party B), or you're a guarantor (Party C). What's a guarantor? For example, Company A and Company B sign an agreement, and Company C provides a guarantee. Company C is the guarantor, like an insurer. I think vendors will be largely eliminated in the future, because with things like Claude Code, Party A can do Party B's work themselves.
But if you're in the guarantor business, I think you can still succeed. For example, if you sell insurance – I can't compensate myself; I need someone else to bear the risk. If you are a guarantor that cannot be bypassed mathematically or by social contract, then your ToB business is absolutely meaningful. But if you're just solving a small problem for others, then I think you should try to transform.
To give another example, in cryptography, there's a concept called Trusted Third Party. You can't get rid of this mathematically. As long as you want to play with cryptography, you can't get rid of it mathematically. If you take on such a role, there will still be demand for it in the future.
But you also have to consider, if there are a lot of ToB companies, and fewer and fewer ToB companies acting as vendors, then you won't have many B's to "to." So you still need to clearly understand two questions: Regardless of A2A, B2B, or B2C, or any X to Y, you really need to clearly understand two questions – who are you delivering what to? And does the recipient have money to pay you?
As long as you understand this, there won't be any major problems, because this is the first principle of business: who do you serve? Do they have money to give you? If you purely do A2A, and the Agent doesn't have financial authority, then whose money will you earn?
Monica
If 99% are on UBI, whose money do you earn? The money of the 1%?
Yin Yifeng
That's equivalent to earning ToC money. If you're earning UBI money, how is that different from ToC?
Monica
If humanoid robots enter human homes and an incident of stabbing occurs, who is responsible for this act? The buyer or the seller?
Yin Yifeng
This is the meaning of the insurance industry's existence. AI is increasingly transforming from an information tool into a decision-making tool, and the basic social contract of human society is that rights and responsibilities are symmetrical – I give AI the right to make decisions, so AI should, in principle, bear the responsibility for those decisions. But the current problem is, saying you'll jail AI or give AI the death penalty is meaningless.
So the most irreplaceable aspect of humans is this – humans can bear responsibility. This is humanity's true dimension-reducing attack on machines, because machines cannot bear responsibility, but humans can. This is what truly makes humans irreplaceable.
So in the future, we might have an insurance business like this: I guarantee that this machine is safe; I create a safety layer or safety plugin, and then I guarantee that your machine is safe. If this machine causes harm, then I will compensate. I think this field will become widespread in the future.
Monica
Will SaaS end in the next few years?
Yin Yifeng
Frankly, SaaS is transitioning in another direction. Before, it was Software as a Service; now it should be called AaaS – Agent as a Service. But whether it's SaaS or AaaS, it's still mostly a B2B logic. As long as it's B2B logic, then you need to consider whether you are a guarantor or a vendor, and whose AaaS you are providing.
Monica
How do you define an AI Native product?
Yin Yifeng
I don't think there's a good definition right now, and I don't want to define it. But frankly, I really don't think products have an AI Native or Internet Native distinction. You just need to truly satisfy what kind of people and what kind of needs, it's that simple. The first principle of a product is: what kind of people and what kind of needs are you satisfying? Are you the best in this category?
AI, or the internet for that matter, these are just tools. How you achieve your goal – it's not the goal, it's the means. Your goal is to provide what kind of service to what kind of people, and then earn money.
Don't overly pursue "am I AI Native." Users in society don't care if your backend is written in GPT or connected to Gemini, or if you wrote ten thousand if-else statements. They only care if it works well, if it provides value, if it's cost-effective, and if it has style. They only care about these things. Really, don't pursue anything "Native"; you just need to make a good product, that's all.
Giving Up OpenAI and Hugging Face, Choosing "Abnormal"
Monica
Let's talk about the AI for Fun, AI for Health track you're currently pursuing. Why did you want to do this? Is this your passion?
Yin Yifeng
It really is my passion. Whether you call me chuunibyou or say I haven't grown up, I still want to save humanity. I don't want the future of humanity to be like this – seeing the real world is so hard, then running into the Metaverse to stay, closing their eyes and entering the Matrix, having their brains scooped out, two tubes inserted, soaked in a tank, and that's how life passes. I don't want this to be the future of humanity.
Even if humans don't work in the future, I still hope they can live happily and joyfully in this world. This is what I want to do. Because I am also a human, I love this species, and I don't want this species to degenerate into that state. I still hope that humans can live in their carbon-based bodies, with souls, happily and joyfully in this world. This is the original intention of what I'm doing.
Monica
What do you think is the essence of happiness?
Yin Yifeng
If we think purely from first principles, it's hormones. The very direct methods of giving you hormones are all written in criminal law, so I can't do that. So I can only give you some happiness, some sense of accomplishment, through other means.
What exactly do I do? I plan to turn "having a healthy lifestyle," which seems somewhat anti-human nature, into a game, into something you enjoy tirelessly.
If you think about it, playing games is actually quite hard. How much skill do you need to practice for DOTA? You practice again and again, lose again and again; this process is actually quite hard. But why do you enjoy it tirelessly? Because it's a game, because it truly achieves gamification. The essence of gamification isn't actually difficult; everyone can think about it and figure it out. So I think, if such anti-human practice, such anti-human monster grinding can make people tirelessly happy, then why can't maintaining a healthy lifestyle be done in the same way?
Monica
But when it comes to games, people still engage in online activities. Will they get addicted to games and become detached from the physical world?
Yin Yifeng
Look, is soccer a game? Is basketball a game? Is badminton a game? To some extent, they all are. When I was a kid, I was the type who never stayed home, playing all day long, not even knowing what I was playing, but just endlessly happy. Games don't necessarily have to be in the virtual world, really not necessarily.
Monica
How do people find their passion? Like you, who found something you might want to do for the next 10 or 20 years.
Yin Yifeng
Frankly, I think passion isn't something you find; it's something you encounter. You really have to throw yourself into this world to experience all sorts of different things, and sooner or later, there will be something that makes you feel "I'd be happy doing this my whole life," and then you've found it. You really have to go out and experience the world, go out and see things, learn more, think more, experience more. I think only through this process can you encounter what you truly love, what you're truly willing to dedicate 10, 20 years, or even half your life to. You won't find it by just lying there thinking.
Monica
Do you have any specific suggestions for experiencing things? For example, traveling to different countries and places, trying different sports, or reading different books?
Yin Yifeng
I think it's simply trying whatever you're interested in. Just try whatever you're interested in, and see if you like it.
Monica
You've been in Silicon Valley for six or seven years. What differences do you see between young people and entrepreneurs in the US and China?
Yin Yifeng
Aside from some very superficial differences, there are no deep-seated differences. Because everyone is in the same physical world, relativity applies to you and me, and Newton's laws apply to you and me. In similar environments pursuing the same goal, your optimal solution should be similar.
If the environmental differences are large, it's not because of differences in people, but because of differences in the environment – different environments might have different optimal solutions. People, one could say, are no different; everyone is doing one thing: I have a goal, I am in an environment, and I need to find the optimal solution to achieve that goal within this environment.
Every excellent entrepreneur is essentially solving a mathematical optimization problem; it's just that the environment might be different, the cards in hand might be different, and the optimization methods might be different, but everyone's goal is the same – to find the optimal solution to achieve their goals. In this regard, the underlying nature of Chinese and American entrepreneurs is completely consistent; there is no difference.
Monica
What areas do you think Chinese entrepreneurs need to improve?
Yin Yifeng
Everyone has areas they need to improve, but I think one thing everyone needs to improve on is: finding a way to avoid staying up late.
Everyone needs to improve on this. If you collapse halfway through your entrepreneurial journey, then everything is lost.
Monica
Survival is the most important.
Yin Yifeng
Yes, I think this is a mistake all founders make. I myself am making this mistake right now – I haven't slept for the past 48 hours. I'm still running on four or five cans of Monster. So I think every entrepreneur will definitely make this mistake, and sometimes you don't have a choice. If you're an entrepreneur and you do have a choice, please get some sleep when you can.
Monica
Many domestic entrepreneurs previously wanted to go to Silicon Valley because they felt the funding environment in China was very poor. But I know you've raised a lot of money, and much of it was raised domestically in the early stages. Do you have any experience to share in this regard?
Yin Yifeng
It depends on what you're doing. If you're doing a ToB business, then book a ticket to Silicon Valley immediately. If you're doing a ToC business, China is still better. We've taken money from both China and Silicon Valley, but slightly more from China. Silicon Valley truly lacks a ToC gene, that's true, and Silicon Valley is indeed not suitable for ToC.
Monica
Do you think if you're doing ToC, it's better to have the team in China?
Yin Yifeng
This is what I think: regardless of whether you're doing ToC or ToB, if you want top-tier technical intuition, it's still necessary to spend some time in Silicon Valley. But honestly, if you're doing ToC, you need to be out of Silicon Valley for at least 3 months a year. You need to see the world outside Silicon Valley, what people are thinking, what they're using.
Silicon Valley is a huge bubble. I stayed in Silicon Valley for so many years consecutively and made a mistake that all Silicon Valley people make – I thought people should be that smart, I thought this species should react to new things that quickly. After being steeped in Silicon Valley for a long time, you really get that feeling. If you go into ToC with that mindset, you're done for. So you really need to go out and see what people outside this capital and tech bubble are doing, what they're thinking, how they live their lives. You really need to see these things. Otherwise, if your ToC product is for the people around you in Silicon Valley, then don't do it; you might as well just stick to ToB.
Monica
Your current product is a Day 1 Global product, targeting the global market. What kind of mindset is needed for Day 1 Global products?
Yin Yifeng
I think for ToB, because everyone needs roughly the same things, as long as you're a business, what you need is pretty similar. For ToB Go Global, the main issues are regulatory and legal aspects.
ToC Go Global is a very tricky thing. For example, if you want to simultaneously operate in the Chinese and US markets, first, they are all human, then they are Chinese or American, and then they are Shandongese, Shanghainese, Minnesotans, Texans. The commonalities in human nature far outweigh cultural differences.
You really need to clearly consider whether your methodology here can be transferred there. Think from first principles: what assumptions did I use to reach this conclusion? If I assumed this person is a human and derived conclusions based on human nature, then it applies in both China and the US. But if I assumed this person is Chinese and derived conclusions, you need to re-examine it when you go to the US.
The premise for a methodology to be useful is that the assumptions must be consistent. If the assumptions are inconsistent, then the methodology's usefulness is uncertain, and you need to verify it again. If you're doing Day 1 Global, a question you really need to consider is: what assumptions did your product make? Which of these assumptions still hold true after you change environments? Which ones don't? If some assumptions don't hold true, how will that affect your conclusions? How do these conclusions need to be changed? How do these methodologies need to be changed?
I think as long as you understand this problem, Go Global is just a matter of time. Of course, you will definitely make many mistakes and suffer many losses, but if you keep using this method to correct your path, I think there's at least a chance. Because the successful Go Global companies I've seen have all done this. This isn't something I summarized; it's a small summary I made after observing what everyone else did.
Monica
If you were an investor, how would you evaluate a team? What kind of team would you be willing to invest in?
Yin Yifeng
From my personal perspective, I don't really like normal people. I like to invest in strange and quirky people.
Monica
Like you, "abnormal"?
Yin Yifeng
Yes, entrepreneurship isn't something normal people do. I gave up a job opportunity at Hugging Face to start a business; life at Hugging Face was quite comfortable.
Monica
OpenAI approached you for GTM (Go-To-Market)?
Yin Yifeng
OpenAI tried to recruit me for three different teams, the last attempt was for GTM. Maybe my GPT usage habits are quite unique, I don't know.
Monica
Perhaps OpenAI saw your data and realized you had potential in that area?
Yin Yifeng
I don't know, it's always been a mystery to me.
Monica
So, going back, you gave up those opportunities to start a business. What kind of person do you think is suitable for giving up opportunities and going into entrepreneurship?
Yin Yifeng
"Abnormal" people. Who else would do such a thing?
Monica
How's your team now?
Yin Yifeng
My logic for doing this was actually quite simple. First, I like it. Second, frankly, I feel that if I don't do this, no one will, and I hope this thing can succeed. I really hope it can turn from a dream into reality, and I also know that if I don't do it, no one will, so I might as well do it.
Monica
Why do you think no one else would do it besides you?
Yin Yifeng
Because commercially speaking, building your own Metaverse, building a Matrix, actually yields higher profits.
The Ultimate Answer – Not Building the "Matrix," Saving Carbon-based Life
Monica
How do you view the vulnerable groups in this world?
Yin Yifeng
You see, as more and more people lose their jobs, more and more people will become vulnerable groups. In a one-person, one-vote political system, the voices of such people will grow louder and have an increasing impact on policy, not a decreasing one.
Taking the US as an example, more and more people are being laid off, and more and more people are unemployed. If you consider unemployed people as vulnerable groups, then vulnerable groups are increasing. And the value of the votes in their hands hasn't changed at all. If this group of people grows larger, then the US will gradually move towards the far left. What kind of impact this will have on the world, I don't really want to predict, but everyone knows it in their hearts.
Monica
What kind of impact do you think this will have on young people today or future entrepreneurial directions?
Yin Yifeng
Because the growing number of vulnerable groups will become a social problem, and the essence of entrepreneurship is solving social problems. Many people will move in this direction, and many will try to solve this problem. I am actually trying to solve this problem too.
Monica
So will ToC services become increasingly universal in the future?
Yin Yifeng
What will the structure of ToC be like in the future? A small group at the top, a large mass at the bottom, and nothing in between.
Monica
Are you serving the majority at the bottom?
Yin Yifeng
Hard to say. You can't just look at the number of people in the market; you have to look at how many people actually have money to spend, how many people can actually afford your product.
So for ToC, to some extent, I've chosen a path that requires a big bet. If UBI truly comes into existence later, then ToC will have huge opportunities; if there's no UBI and the problem is solved in other ways, then ToC might only be able to operate like Hermès, serving only the very top tier of people.
Monica
Regarding the AI for Health track, I've always had a question: from a commercialization perspective, will you serve those at the top of the pyramid who have higher paying capabilities, perhaps willing to pay hundreds or thousands of dollars a month? Or will you stand from the perspective of the majority of humanity, serving those at the bottom who can only afford $1-5 a month?
Yin Yifeng
It depends on how much UBI there is. Also, I think it's not impossible to do both lines. For example, Li & Fung Group owns Swatch, and also high-end brands. You can have your cake and eat it too. But I still hope to enable people at the bottom to live healthy, happy, and dignified lives. Because frankly, I am a human, and I really don't want the human species to degenerate into that state. I still hope everyone can have a beautiful future, and live healthy, happy, and dignified lives. This is what I want to do.
And in the future, customer acquisition costs will be lower and lower. Even if I charge a little less, I can still earn a lot of money because the base is large enough.
Monica
Do you think in the future, many humans will have mixed relationships – for example, marriage, intimate relationships – between carbon-based and silicon-based beings?
Yin Yifeng
This is bound to happen. I envision hospitals 30 years from now basically being 4S shops.
For example, someone as adventurous as me, if I break my leg skiing one day, I'll just go to a 4S shop and get a new one, right? It's no different from a car losing a wheel; just go to a 4S shop and replace it.
Monica
Perhaps not just hospitals, but various scenarios, like dining, might also be like this.
Yin Yifeng
I think if you want a person to live a healthy, happy, and dignified life, letting them eat delicious food is a very, very important step. People will still pursue good food. Even with the current level of pre-made dishes, people still prefer food with "wok hei" (wok breath).
Again, users don't care if you have high-tech content; users care if you can meet their needs, what experience you can provide, and what emotions you can evoke. Really, don't consider the underlying technical content. The first principle of a product: what value do you provide to whom? Can you get your money back?
Monica
It sounds like restaurants won't be easily replaced.
Yin Yifeng
Frankly, restaurants won't be easily replaced, but chefs might become an extremely high-end profession in the future. Chefs might become pure handmade, intangible cultural heritage artisans. You're not buying an industrial product – if I want to buy art, I'll invite a chef; if I just want delicious food, a machine-cooked dish is also very tasty.
Monica
Will those who can maintain their independence in the future, not joining the UBI group, necessarily have special skills?
Yin Yifeng
Not special abilities, but rather that this person is still qualified to be part of the productive forces, still qualified to control productive forces. Because every technological revolution's essence in society is the equalization of means of production. Whoever can activate the means of production brought to them is a valuable person, and can control productive forces.
Monica
Can you give some examples? Who do you think will still be doing quite well in 10 years?
Yin Yifeng
First, all pure handmade, intangible cultural heritage items will do very well, because they sell emotional value, they sell unique experiences. This demand is huge; as long as you sell these things, there will definitely be no problem.
Second, roles that are mathematically indispensable third parties – like the Trusted Third Party in cryptography.
Third, owners of means of production can make money effortlessly.
Furthermore, I even think an alternative to UBI is this: everyone, at birth, is assigned 50 Agents, and the money earned by these 50 Agents goes to you. Everyone, at birth, I give you 50 workers, and these 50 Agents are responsible for supporting you. Frankly, who can clearly say what production relations will be like in 10 or 20 years? Anything is possible.
Monica
Recently, brain-computer interfaces in the US claimed that our generation might live to 1000 years old. What's your take?
Yin Yifeng
I think that's a bit conservative. I think for our generation, there are endless ways to die – I could be hit by a car, get into an accident doing extreme sports, or drop dead next second – but the possibility of dying of old age is basically zero.
Monica
Based on your judgment of current medical developments and future prospects, even if people don't truly die biologically, can their quality of life be maintained?
Yin Yifeng
I think there's no problem. My ideal state is that in 30 years, all cells in everyone's body are in the same state as when they were born. But again, health is the capital for revolution, so everyone just hang in there for 30 years first. Eat well, sleep well, exercise well.
Monica
I think you're particularly good at recruiting and finding people, attracting excellent talent to your team. Are there any tricks to this? Where do you find these top talents in Shanghai?
Yin Yifeng
Frankly, luck is truly just luck. Three heads are better than one; the world isn't short of people with potential, but it lacks eyes that can spot potential. What's even more lacking than eyes that spot potential? It's being willing to let others try, willing to give them a chance, willing to let others prove themselves.
Of course, this relates to funding. If you raise less money, you don't have that much room for generosity. So ultimately, it comes down to raising money.
Monica
An investor's pitching ability is very important.
Yin Yifeng
A dream is too expensive. It's the most expensive thing in the world, because if you truly want to achieve this dream, if you truly love doing this from the bottom of your heart, you will pour all your resources, all your possessions into it. A dream is the most expensive thing you can own in this life, because it will truly drain everything from you.
So, on your journey to realize your dream, having someone willing to share the burden with you is a very fortunate thing. That's the best kind of fundraising.
Audience
Are the entrepreneurial environments in the US and China very different?
Yin Yifeng
Entrepreneurship in the US is more market-driven, while entrepreneurship in China is more government-driven. This is basically the only difference. In other words, when starting a business in the US, you need to find your direction in the market, and your PMF (Product-Market Fit) is actually hard to find. But if you start a business in China, you just do what the government supports, and you don't even need to consider PMF.
Monica
Maintaining flexibility is very important, always paying attention.
Yin Yifeng
That's right, you can't be stubborn. When the market slaps you in the face and tells you something really won't work, don't stubbornly persist. I think entrepreneurship is a process of learning from the market – you constantly experiment in the market, constantly test the market, and then let the market tell you what to do and how to do it. Really, don't argue with the market; you can't win against it.
Monica
If it's a passion that can't find PMF, should one be cautious?
Yin Yifeng
Yes. What is called passion, there's a popular term now called Founder-Market Fit, which is whether you are suitable for what you are doing. Passion is a big part of it.
Monica
Under the current OPC (One Person Company) trend, many students are starting businesses just for the sake of it, feeling "I want to be an entrepreneur," but perhaps haven't found their strong passion yet. For this situation, do you advise them to go back to working for a big company or continue exploring?
Yin Yifeng
That depends on their own choice. Everyone's path is different, and everyone's situation is different; I cannot give a one-size-fits-all answer.
I think you really need to know what you want, then see what cards you hold, and what kind of environment you are in. After that, it's a mathematical optimization problem – you have so many choices in front of you, and when you choose, find which choice can get you closer to your goal.
There's a saying from university, "Knowing where to rest, one can have a fixed purpose" – only when you have a life goal can you have determination, truly meaningful decision-making ability, and know what decisions you should make. You still need to know what you want, and then you are qualified to talk about which direction you should go.
If you really can't find what you're passionate about, then first find something that allows you to survive, and then keep experiencing new things. Passion is something you truly encounter; it's not something you actively search for or think about to find. You truly stumble upon it inadvertently.
You have to put yourself in a lucky position – a position where you can encounter such things. From a quantum level, everything in this world is random, and many things are random. The world we see, to a large extent, the element of luck in anyone's success is actually not low. For example, Kobe was born tall, Schwarzenegger was born with that physique, Einstein was born with that brain. How to make yourself a lucky person, I think this is a required course for everyone. For example, if you want to find what you are truly passionate about in life, then you have to put yourself in a place where you actually have a chance to encounter such things, and then you can be very lucky. I think before you find what you truly love, do one thing: maximize your luck.
No comments yet.