Image source: Generated by Unbounded AI

Recently, the well-known American podcast Invest Like the Best once again interviewed Marc Andreessen, co-founder of Andreessen Horowitz. In the interview, Marc and anchor Patrick discussed in depth about AI's reshaping technology and geopolitics. The major changes and discussions on the significance of DeepSeek's open source artificial intelligence and its major power technology competition, in addition, they share their views on the evolution of the global power structure and the overall transformation of the venture capital industry.

"Bright Company" used AI tools to organize the core content of the interview as soon as possible. For details on the full text, please see the "original link" at the end of the article.

The following is the interview content (abridged):

Talk about DeepSeek, AI winner and loser Patrick: Marc, I think we We must start with the most core issue. Can you talk about your opinion on DeepSeek's R1?

Marc: There are many dimensions here. (I think) The United States remains a recognized scientific and technological leader in the field of artificial intelligence. Most of the ideas in DeepSeek have been derived from the past 20 years, and it is even surprising that work was conducted in the United States or Europe 80 years ago. The initial research on neural networks was carried out as early as the 1940s at research universities in the United States and Europe.

So, from the perspective of knowledge development, the United States is still far ahead.

But DeepSeek has done a very good job of this knowledge. They also did one amazing thing, which was to provide it to the world in open source. This is actually quite amazing because there is a reversal of this phenomenon. You have American companies like OpenAI, which are basically completely closed.

Some of Elon Musk's lawsuit against OpenAI was asking them to change the company name from OpenAI to Closed AI. OpenAI initially imagined that everything would be open source, but now everything is closed. Other large AI labs, such as Anthropic, are also completely closed. In fact, they have even stopped publishing research papers, treating everything as exclusive property.

And the DeepSeek team actually fulfilled the promise of truly open source for their own reasons. They published code for their LLM (called V3) and their inference (called R1) and published a detailed technical paper explaining how they built it, which is basically for anything they want to do something like Others working provided a roadmap.

So it's already public. There is a false argument outside that if you use DeepSeek, you will give all the data to the Chinese. If you use the service on the DeepSeek website, this is trueof. But you can download the code and run it yourself. But let me give you an example: Perplexity is an American company, and you can use DeepSeek R1 on Perplexity, fully hosted in the United States. Microsoft and Amazon now have a cloud version of DeepSeek, which you can run on their cloud platform, and obviously both are American companies, using American data centers.

This is very important. You can download the system now and actually run it on $6,000 worth of hardware at home or in your company. Its capabilities are comparable to the cutting-edge systems of companies such as OpenAI and Anthropic.

These companies invest a lot of money to build their systems. Now, you can buy it for $6,000 and have full control. If you run it yourself, you have full control. You can have a complete transparency into what it is doing, you can modify it, you can do various operations on it.

It also has a very outstanding property called distillation. You can compress the big model that requires $6,000 in hardware and create smaller versions of the model. Someone has created smaller versions of the model online and it has been optimized so that you can run them on your MacBook or iPhone. These versions are not as smart as the full version, but they are still quite smart. You can create customized, field-specific, distilled versions that do well in specific areas.

This is a huge improvement in making big model reasoning and R1 models more popular in programming and scientific reasoning. Six months ago, these were still very profound, extremely expensive and proprietary. Today, it has become free and always available for everyone.

Every big tech company, internet company, every startup, we have dozens or even hundreds of startups this week, either rebuilding based on DeepSeek or integrating it into their In the product, either study the technology they use and use it to improve existing AI systems.

Marc Zuckerberg of the Meta team recently said that the Meta team is tearing down DeepSeek, borrowing these ideas completely legally because it is open source and ensuring the next version of Llama is in the reasoning capabilities At least comparable to DeepSeek, or better. This has really driven the world.

The two main points we can learn from it are: AI will be everywhere. There are a lot of AI risk control personnel, security personnel, regulators, officials, governments, EU, British, etc... All of these people want to limit and control AI, and this basically ensures that none of this will happen, I think This is very good. It is very consistent with the free tradition of the Internet. This then achieves a 30-fold reduction in cost of reasoning capabilities.

Maybe finally pointing out that this suggests that reasoning will work. The reasoning will beAny area of ​​human activity works, as long as you can generate answers that can be checked after the fact by a technician.

We will have AI that can perform human and transhuman reasoning, which will play a role in areas that really matter: encoding, mathematics, physics, chemistry, biology, economics, finance, law and medicine .

This basically ensures that within five years, everyone on earth will have an AI lawyer and AI doctor who is above human level. They are always on call, which is just a standard feature on mobile phones. This will make the world a better, healthy and wonderful.

Patrick: But this is also the most unstable, and the model will be outdated within two months. A lot of innovation is happening at every technological level. But just from this point of view, enter this new paradigm, if you are writing a column about winners and losers for all stakeholders, whether it’s new app developers, existing software developers, like Nvidia Infrastructure provider, open source and closed source model company. Who do you think is the winner and loser after R1 is released?

Marc: If you take a "snapshot" today, then from the perspective of zero-sum game, the winners and losers at a point in time are all users, all consumers, every one Individuals, and every enterprise that uses AI.

There are some startups, such as those that do AI legal services, and the cost of using AI last week was 30 times that of now.

For example, for a company that is building an AI lawyer, if the cost of its critical input drops by 30 times, it is like the cost of gasoline drops by 30 times while driving. Suddenly, you can drive 30 times the distance with the same dollar, or you can use the extra spending power to buy more. All of these companies will either greatly expand their ability to use AI in all of these areas, or they will be able to provide services in a cheaper or free way. So for users and the world, this is an excellent result based on a fixed-scale plate.

The losers are those companies that have proprietary models, such as OpenAI, Anthropic, etc. You'll notice that OpenAI and Anthropic have both sent out quite tough but provocative messages over the past week explaining why this isn't their end. There is an old saying in business and politics that when you explain, you lose.

Then the other is Nvidia. There are a lot of comments on this, but Nvidia makes standard AI chips that people use. There are some other options, too, but Nvidia is used by most people. Their chips have a profit margin of up to 90%, and the company's stock price reflects this. (Nvidia) is one of the most valuable companies in the world. One of the things the DeepSeek team did in their paper is that they figured out how to use cheaper chips, which are actually still using Nvidia's chips, but theyUse more efficiently.

Part of the 30x cost reduction is that you only need fewer chips. By the way, China is building its own chip supply chain, and some companies are also starting to use Chinese-derived chips, which is of course a more fundamental threat to Nvidia. So this is a snapshot at some point in time. But the problem is that your question suggests another way of looking at it, that is, over time, what you want to see is the elastic effect. Satya Nadella used this phrase called the Jevins Paradox.

Imagine gasoline. If the price of gasoline drops sharply, suddenly people will drive more cars. This often occurs in traffic planning. So you will have a city like Austin, where traffic is congested, and someone will have a sudden idea to build a new highway next to the existing highway. And in just two years, the new highway will be blocked, and it may be even harder to get from one place to another. The reason is that the reduction in the prices of key inputs can induce demand.

If AI suddenly becomes 30 times cheaper, people might use it 30 times, or by the way, they might use it 100 times or even 1000 times. This economic term is called elasticity.

So the price drop is equal to the explosive growth of demand. I think there is a very reasonable scenario here, that is, on the other side, DeepSeek will do a great job as usage explodes. By the way, OpenAI and Anthropic will do well, Nvidia will do well, and Chinese chip manufacturers will do well.

You then see a tidal effect where the entire industry will grow explosively. We're really just starting people figure out how to use these technologies. The reasoning has only started to work in the past four months. OpenAI only released their o1 inference model a few months ago. It's like taking the fire from the mountain and giving it to all mankind. Most humans have not used fire yet, but they will.

And then, frankly, it's also an old idea, namely, well, if you're OpenAI or something like that, what you did last week isn't good enough. But then again, this is how the world is. You have to be better. These things are all competitions. You have to evolve. So it's also a very powerful catalyst that has prompted many existing companies to really raise their level and become more radical.

…

Patrick: … If a Chinese company uses models developed in the United States, these models invest a lot of money, and then lead to this technology that brings richness to the world, It's a difficult thing to understand. I would love to hear your reaction from these two perspectives.

Marc: Yes, so here are some real problems. There is an irony in this argument, and you do hear it. Of course, the irony is that OpenAI has not releasedMing Transformer. The core algorithm of large language models is called Transformer.

It was not invented in OpenAI, but in Google. Google invented it, and published a related paper, and then, by the way, they did not productize it. They continued to research it, but did not productize it because for "safety" reasons, they thought it might be insecure. So they let it sit on the shelves for five years, and then OpenAI’s team got it, picked it up and kept pushing forward.

Anthropic is a branch of OpenAI. Anthropic did not invent Transformer either. So whether it’s these two companies, or every other American lab that is studying large language models, every other open source project is built on something they themselves have not created and developed.

By the way, Google invented Transformer in 2017, but Transformer itself is based on the concept of neural networks. The idea of ​​neural networks can be traced back to 1943. So, 82 years ago was actually the time when the original neural network paper was published, and Transformer was built on 70 years of research and development, most of which were funded by the federal and European governments in research universities .

So, this is a very long lineage of intellectual thought and development, and most of the ideas entering all these systems are not developed by the companies that currently build these systems. No company sits here, including our own, without any special moral propositions that we are built from scratch and that we should have full control. This is not true at all.

So, I would say, arguments like this are out of frustration at the moment. By the way, these arguments are also meaningless, because China has done so, it has come out, and things have happened. There is now a debate about copyright. If you talk to experts in this field, a lot of people have been trying to understand why DeepSeek is so great. One of the theories, which is an unproven theory, but one that experts believe is that Chinese companies may have trained using data that American companies do not use.

What is particularly surprising is that DeepSeek is excellent in creative writing. DeepSeek is probably the best AI in English creative writing in the world at present. This is a bit strange because the official language of China is Chinese. While there are some very good Chinese English novelists, in general, you might think that the best creative writing should come from the West. And DeepSeek is probably the best right now, which is shocking.

So one of the theories is that DeepSeek may have been trained. For example, there are some websites called Libgen, which are basically huge mutualThe networked repository is full of pirated books. Of course I don't use Libgen myself, but I have a friend who uses it a lot. It's like a superset of the Kindle store. It has every digital book, which exists in PDF format, and you can download it for free. It's like the movie version of Pirate Bay.

The American lab may not feel that it can simply download all the books from Libgen and train, but maybe the Chinese lab feels that it can. Therefore, there may be such a differential advantage. Having said that, there is an open copyright dispute here. People need to be careful about this because there is an open copyright battle here, and some publishing companies are basically hoping to prevent generative AI companies like OpenAI, Anthropic, and DeepSeek from being able to use their content.

There is an argument that these materials are copyrighted and cannot be used at will. There is another argument, basically saying that AI trains books, you are not copying books, you are reading books. It is legal to read books in AI.

You and I are allowed to read books, by the way. We can borrow books from the library. We can pick up friends’ books. These behaviors are legal. We are allowed to read books, we are allowed to learn from books, and then we can continue our daily lives and talk about the ideas we learn in books. Another argument is that training AI is more like reading books than stealing.

There is then the actual reality that if...their AI can be trained for all books, and if American companies are ultimately prohibited from training books by law, the United States may lose the game in the AI ​​field .

From a practical point of view, this could be a fatal blow, like they win and we lose. There may be some tangle of the whole debate. DeepSeek did not disclose the data they used for training. So when you download DeepSeek, you don't get the training data, you get the so-called weight. So, what you get is a neural network trained after training materials. But it is difficult or even impossible to view and deduce the training data from it.

By the way, Anthropic and OpenAI have not disclosed the data they are using for training. Then there is fierce speculation in the field about what and nothing in OpenAI training data. They think it's a trade secret. They won't disclose this content. Therefore, China's DeepSeek may or may not be different from these companies. They may be different from Chinese companies in terms of training methods. We don't know.

We don't know what the OpenAI and Anthropic algorithms are, because they are not open source. We don't know how much better or worse they are than the public DeepSeek algorithm.

Discussion on closed source and open source

Patrick: Do you think those closed-source models that are entering competition, such as OpenAI and Anthropic, will eventually become more like the Android of Apple and Google?

Marc: I support maximization of competition. By the way, this fits my identity as a venture capitalist. So, if you are a company founder, if I run an AI company, I need to have a very specific strategy that has advantages and disadvantages and needs trade-offs.

As a venture capitalist, I don't need to do this. I can make a variety of conflicting bets. This is what Peter Thiel calls certainty optimism and non-deterministic optimism. The founder and CEO of the company must be deterministic optimists. They have to have a plan and they have to make tough trade-offs to achieve that plan. Venture capitalists are non-deterministic optimists. We can fund a hundred companies with 100 different programs and conflicting assumptions.

The essence of my work is that I don’t need to make the choice you just described. This then makes it easy for me to make a philosophical argument that I personally sincerely agree that I support the maximum level of competition. So, going deeper, it means I support the free market, the maximum level of competition, and the maximum level of freedom.

Essentially, if there are as many smart people as possible to come up with as many different ways as possible and compete with each other in the free market, see what happens. Specifically for AI, this means I support large labs to grow as quickly as possible.

I support OpenAI and Anthropic 100% on doing whatever they want, launching whatever products they want, and growing as hard as possible. As long as they do not receive priority policy benefits, subsidies or support from the government, they should be able to do whatever they want as a company.

Of course, I also support startups. We are certainly actively funding AI startups of all sizes and types. So, I hope they can grow, and then I hope open source can grow, partly because I think if things come out in open source, even if that means some business model companies won't work, it's so much of a benefit to the world and the industry as a whole. , we will find other ways to make money. AI will become more common, cheaper, and easier to obtain. I think this would be a good result.

Then, another very critical reason for open source is that without open source, everything becomes a black box. Without open source, everything becomes a black box owned and controlled by a few companies that may end up colluding with the government, and we can discuss it. But you need to be open source so that you can see what's going on inside the box.

By the way, you also need open source for academic research, so you need open source for teaching. So the problem before open source is that back two years ago, there was no basic open source LLM at that time, Meta released Llama, then Mistral, France, now DeepSeek.

But before these open source models emerged, a crisis was being experienced in the university system where university researchers did not have enough money to buy billion-dollar Nvidia in places like Stanford, MIT and Berkeley. chips to truly participate in competition in the AI ​​field.

So if you talked to a computer science professor two years ago, they would be very worried. The first concern is that my university does not have enough funds to compete in the AI ​​field and stay relevant. Then another concern is that all universities combined don’t have enough funds to compete because no one can keep up with the fundraising capabilities of these large companies.

Open source brings universities back to competition. This means that if I were a professor at Stanford, MIT, Berkeley, or any state school, whether at the University of Washington or elsewhere, I can now teach using Llama code, Mistral code, or DeepSeek code. I can do research and I can actually make breakthroughs. I can publish my research to give people a real idea of ​​what’s going on.

Every new generation of kids then comes to college and take computer science courses and they will be able to learn how to do this, and if it were a black box they wouldn't be able to do it. We need open source, just as we need freedom of speech, academic freedom and research freedom.

So my model is basically that you let big companies, small companies and open source compete with each other. This is what's happening in the computer industry. This works very well. This is what happened in the Internet industry. The effect is very good. I believe this will happen in the AI ​​field and I think it will work well.

Patrick: Is there a limit, that is, to want the maximum evolutionary speed and the maximum competition? Maybe there is. If I say, we know the best things are made by China,..., is there a situation where you say, yes, I want to evolve and compete to the greatest extent, but the national interests are somehow surpassed A desire for maximum evolutionary speed and development?

Marc: This argument is a very real one. It is frequently proposed in the field of AI. In fact, as we were sitting here today, there were two things. First of all, there are actually restrictions on Western and American companies selling cutting-edge AI chips to China. For example, Nvidia cannot actually legally sell its cutting-edge AI chips to China today. We live in a world where this decision has been made and that policy has been implemented.

Then the Biden administration issued an executive order that I think has been revoked now, but they issued an executive order that would impose similar restrictions on software. This is a very active debate. With the DeepSeek incident, another round of such debate is underway in Washington, D.C.

And basically, when you get stuck in policyWhen arguing, you will encounter a classic situation where you have a rational version of the argument, that is, from a theoretical perspective, what is in the interests of the country. Then you have a political version of the argument, well, what does the political process actually do to rational arguments? Let me say this, we all have a lot of experience, and watching rational arguments meet political processes, usually not rational arguments win. After processing by political machines, the results are usually not what you originally thought you would get.

There is a third factor that we always need to discuss, namely, the corruption impact of large companies in particular. If you are a large company and you see changes (more competitive) happening in Chinese companies, threats of what is happening in open source, of course you will try to use the U.S. government to protect yourself. Maybe this is in the national interest, maybe it is not. But you will certainly push this, whether it is in the national interest or not. That's what complicates this debate.

You can't sell cutting-edge AI chips to China. This certainly hinders them in some ways. There are some things they won't be able to do. Maybe this is a good thing because you have decided that it is in the national interest. But let's look at three other interesting consequences that arise.

So one of the consequences is that it provides a huge impetus for Chinese companies to design how to achieve things on cheaper chips. This is an important part of DeepSeek's breakthrough, where they figured out how to use legal, compliant, cheaper chips to do what American companies can do with larger chips. This is one of the reasons why it is so cheap. One of the reasons is that you can run it on $6,000 worth of hardware because they put a lot of time and effort into optimizing the code to enable it to run efficiently on cheaper chips that are not sanctioned. You forced an evolutionary reaction.

So this is the first reaction, maybe it has somehow backfired. The second consequence is that you inspire China's state-owned and private sectors to develop a parallel chip industry. So if they know they can't get American chips, then they'll go and develop. They are doing this now. They have a national plan to build their own chip industry so that they don't rely on American chips anymore.

So from a counterfactual standpoint, maybe they will buy American chips. Now they will figure out how to make it themselves. Maybe five years later they can do that. But once they reach a position where they can make their own, we will have a direct competitor that we won't have in the global market, if we just sell them chips. And by the way, by the way, we had no control over their chips. They have complete control. They can be sold at a price below cost and they can do whatever they want.

…

How AI reasoning capabilities change VC and Investment Industry

Patrick: How do you think all of this will affect capital allocation? What I'm most interested in is maybe five years from now how your company, Andreessen Horowitz (A16Z) will be affected. If I think an investment company is a combination of ability to raise capital, do great analytics and be able to judge people, especially in the early stages, how do you think this function will change due to the emergence of "o7" (AI reasoning ability) ?

Marc: I hope the analysis part can change dramatically. We assume that the best investment companies in the world will be very good at leveraging this technology to do the analytics they do.

With that being said, there is a saying that “the son of a shoemaker has no shoes”, and perhaps the venture capital firms that invest most aggressively in the AI ​​field may be those that are not aggressive enough in practical applications. one. But there are multiple efforts going on within our company and I am very excited about it. But companies like us need to keep up with the situation, so we have to really do that.

Is some work already carried out within the industry? Probably not yet. Maybe not enough. That being said, for late-stage investment or open-market investment, many of the people we talk to have a very analytical perspective. There are even great investors, I think it is Warren Buffett. I don't know if this is true, but I've always heard that Warren never meets with the CEO.

Patrick: He wants the “Ham Sandwich Company.”

Marc: Yes, yes, he hopes the company is as simple as a ham sandwich. And I think he was a little worried that he would be attracted to a good story. You know, many CEOs are very attractive people. They are always described as “good hair, very white teeth, shiny shoes, straight suits.” They are excellent in sales. You know, one of the things CEOs are good at is selling, especially selling their own stock.

So if you are Buffett and you are sitting in Omaha, what you do is read the annual report. The company lists everything in its annual report and is subject to federal law to ensure that its content is authentic. So that's how you analyze it. So, are inference models such as o1, o3, o7 or R4 better than most investors’ manual analysis of annual reports? Probably yes.

As you know, investing is an arms race, just like everything else. So if it works for one person, it will work for everyone. It will become an arbitrage opportunity for a while, then it will close and become the standard. Therefore, I expect the investment management industry to adopt this technology in this way. This will become a standard way of operation.

I think it's a little different for early stage venture capital. What I want to say next may be my personal wishful thinking. IProbably the last Japanese soldier on a remote island in 1948, saying what I'm going to say next. I'm going to take a chance. But what I'm going to say is, look, in the early stages, a lot of what we do in the first five years is actually really evaluating individuals in depth and then working very deeply with those people.

This is also why venture capital is difficult to scale, especially (cross) geographic scale. Geographic scale experiments often don't work. The reason is that you end up having to spend a long time face-to-face with these people, not only during the evaluation process, but also during the construction process. Because in the first five years, these companies have not usually entered autonomous driving.

You actually need to work closely with them to make sure they can achieve everything they need to succeed. There are very in-depth interpersonal relationships, dialogues, interactions, and guidance. By the way, we learn from them, and they learn from us. This is a two-way communication.

We don't have all the answers, but we have a perspective because we see a wider panorama, while they focus more on specific details. Therefore, there is a large number of two-way interactions. Tyler Cowen talked about this, and I think he called it "project picking."

Of course, “Talent Mining” is another version, that is, basically, if you look back at any new field in human history, you almost always find this phenomenon, i.e. there are people with unique personalities Try to do something new and then have some professional support layers that are funding and supporting them. In the music industry, David Geffen discovered all the early folk artists and turned them into rock stars. Or in the film industry, it was David O. Selznick who discovered early film actors and turned them into movie stars. Or in a cafe or tavern in Maine 500 years ago, someone was discussing which whaling captain could catch whales.

You know, it was Queen Isabella who listened to Columbus' proposal in the palace and said, "It sounds reasonable. Why not?" This alchemy developed in time, This alchemy, developed between the professional support layers that support and fund these people, has existed for hundreds, even thousands of years.

You may have seen tribal leaders thousands of years ago, sitting around the fire, young warriors come over and say, “I want to lead a hunting team to the area over there, See if there are better prey over there.” And the leader sat next to the fire, trying to decide whether to agree. So it's a very human interaction. My guess is that this interaction will continue. Of course, then again, if I come across an algorithm that is better at doing this than I do, I will retire immediately. Let's wait and see.

Patrick: You are building the largest in this fieldone of the companies. How to adjust the company's development strategy to deal with this new technology? Whether it is in actual operation or strategic direction, have you made adjustments? How do you adjust the direction of your company to deal with this new technology?

Marc: An important part of running a venture capital company, in our opinion, is that there is a set of values ​​and behaviors that you must have, which we call eternal. For example, respect for entrepreneurs. You need to show great respect for entrepreneurs and the journey they have gone through. You need to get a deeper look at what they do. You can't just go by.

You need to build deep relationships. You have to work with these people for a long time, and by the way, these companies will take a long time to build. We don't believe in success overnight. Most great companies are built over a span of 10, 20, 30 years. Nvidia is a good example. Nvidia is about to celebrate its 40th anniversary, and I think one of Nvidia's original venture capital firms is actually still on the board of directors. This is a good example of long-term construction.

So, there is a core set of beliefs, opinions, and behaviors that we will not change, and these are related to what we just mentioned. Another thing is face-to-face communication. You know, these things can't be done remotely, and that's one of them. But on the other hand, you need to keep up with the times because technology changes so quickly, business models change so quickly, and competition dynamics change so quickly.

If anything is different, the environment becomes more complicated because you have many countries now and now all these political issues, which also makes things more complicated. We have never really worried that the political system would put pressure on our investments in the past until about eight years ago. Then about five years ago, that pressure really increased. But it was never a big deal in the first decade of our company and the first 60 years of venture capital, but now it is.

So we need to adapt. We need to participate in politics, which we have not done before. Now we need to adapt, we need to figure out maybe the AI ​​companies will be very fundamentally different. Maybe their organizational structure will be completely different. Or maybe the way software companies work completely differently, as you said.

A question we often ask ourselves, for example, what is the organizational structure of a company that really makes the most of AI? Is it similar to the existing organizational structure, or is it actually very different? There is no single answer to this, but we are seriously thinking about this.

So, one of the delicate balanced tasks we do every day is to try to figure out what is eternal and what is keeping up with the times. This is conceptually an important part of my thinking about the company, that we need to navigate between the two and make sure we can differentiate them.

Patrick: Your company is already very large now, and it is somewhat similar to a company like KKR or Blackstone. You and Ben Horowitz) As founders, they are all experienced founders, when you founded this company. Similar to Blackstone, Schwarzman never really made an investment before founding Blackstone. Look at how it is developing now. It seems like this founder-led approach to building asset management investment companies that will eventually develop into truly huge and ubiquitous platforms. You have a vertical business that covers most exciting technology frontiers. Do you think this view makes sense? Will the best capital allocation platform be founded more by founders than investors?

Marc: Yes, so there are a few points. First of all, I think this observation makes sense. In the industry, people often talk about this, that many investment operations are often called partnerships. Many venture capital firms operate in this way. Historically, it was a small team of people sitting in a room, exchanging ideas with each other, and then investing. By the way, they don't have a balance sheet. This is a private partnership. They pay the funds at the end of each year in the form of compensation. This is the traditional venture capital model.

A traditional venture capital model, with six general partners (GPs) sitting around the table to perform this operation. They have their own assistants and several assistants. But the point is, it is based entirely on people. By the way, you actually find that in most cases people don't like each other very much.

Mad Men shows this very well. Remember in Mad Men, in season 3 or 4, members leave to start their own company, and they don’t actually like each other. They knew they needed to come together to start a company. That's how many companies operate. So, it is a private partnership, and it is what it represents.

But then you see that these companies are hard to sustain. They have no brand value. They have no potential corporate value. They are not a business. The company you see this model is that when the initial partners are ready to retire or do something else, they hand it over to the next generation. Most of the time, the next generation cannot continue to maintain. Even if they can maintain it, there is no potential asset value. The next generation will have to hand it over to the third generation. It might fail in the third generation, and then it will eventually appear on Wikipedia. It would be like, “Yes, this company existed, and then it disappeared, and other companies replaced it, passing by like a ship at night.”

So this is Traditional way of operation. By the way, if you are receiving traditional investment training, you are receiving the investment part, but you have never received training on how to build a business. So, this is not your natural strength, you don't have this skill or experience, so you won't do it. Many investors operate in this way for a long time and make a lot of money. So, it works well.

Another way is to build a company, build a business, build something with lasting brand value. You mentioned companies like Blackstone and KKR, these huge publicly traded companies. The same goes for Apollo, these huge companies—you probably know, the initial banks were actually private partnerships. Goldman Sachs and JPMorgan 100 years ago were more like small venture capital firms today than they do now. But then, their leaders transformed them into these huge businesses over time. They are also large listed companies.

So, this is another way to build a franchise. Now, to do this, you need a theory of why a franchise should exist. You need a conceptual theory of why it makes sense to do so. Then, yes, you need business skills. Then, by that time, you are running a business and it's like running any other business, that is, OK, I have a business. It has an operational model, it has an operational rhythm, it has management capabilities, it has employees, it has multiple levels, it has internal professional division of labor and specialization.

Then you start thinking about expansion, and over time, you start thinking about the potential asset value, that is, the value of this thing is not just the people there right now. It is not like us, eager to distribute profits, or anything else. But one big thing we are trying to do is build something that has this durability.

By the way, we are not in a hurry to go to the market, or anything, but one of the big things we are trying to do is build something that has this durability.

Patrick: What new differences do you hope the company will have in the next 10 years that do not exist yet? Are there any uncompromising ways in which you want a company to never evolve like traditional large asset management companies?

Marc: We are rapidly evolving in terms of investments, what the company does, the model, and the background of the founder, and that content has been changing. For example, there has been a consensus in the venture capital community for 60 years that you will never support researchers in starting companies for research. He will only do research, exhaust the funds, and in the end you will get nothing.

However, many of the top AI companies are founded by researchers today. This is an example of some so-called "eternal and unchanging" values ​​that need to be adjusted according to changes in the times. We need to maintain a high degree of flexibility in these changes. Therefore, with these changes, the help and support needed for a company to succeed will change.

About one of the most significant changes in our company, I have mentioned before, that is, we now have a huge and increasingly complex political operation department. Four years ago, we were still blank in the political field. And now, this has become an important part of our business and we have never expected it before.

I'm sure, againIn 10 years, we will not only invest in areas that are currently unimaginable, but also have operating models that are currently unimaginable. Therefore, we are completely open to changes in these areas. However, there are some core values ​​that I hope to remain the same for the next 10 years because these values ​​have been well thought out and are the cornerstone of our company.

But I have always emphasized to our team members and limited partners that we are not pursuing scale. Many investment companies will prioritize expanding asset management scale after reaching a certain scale, from billions to hundreds of billions or even trillions of dollars. This approach is often criticized for focusing more on charging management fees rather than achieving excellent results in investment. This is not our goal.

The only reason we scale is to support the companies we want to help founders build. When we scale up, it’s because we believe it helps us achieve this.

However, I must emphasize that the core of our company is always early stage venture capital. No matter how big we get, even if we set up growth funds, we can write bigger checks—some AI companies do require a lot of money. We did not set up growth funds from the beginning, but were gradually established with market demand and company development.

But core business is always early stage venture capital. This can be confusing because from the outside we manage a lot of money. Why, as the founder of an early-stage startup, I would believe you would be willing to spend time on me? Because you Andreessen Horowitz invested hundreds of millions of dollars in your later investments, and you only invested $5 million in my Series A financing. Will you still take the time to follow me?

The reason is that our company's core business has always been early-stage venture capital. From a financial perspective, the return opportunities of early-stage investments are comparable to those of late-stage companies, which is a characteristic of a startup. But more importantly, all our knowledge, relationship networks and what sets our company apart comes from our deep insight and connections at the early stages.

So, I always tell people that if the situation forces the world to the point where we have to make sacrifices, then early venture capital businesses will never be sacrificed. This will always be the core of the company. That's why I spent a lot of time working with early founders. On the one hand, this is very interesting; on the other hand, it is also the place where you learn the most.

The transformation of global power structure: Elite and Anti-elite Patrick: If we consider changes in global power structure,..., which power centers do you pay the most attention to Changing, whether it is gaining power or losing power?

Marc: The Machiavellians. I'm sure you probably have a dozen people on your showRecommended this book. This is one of the greatest books of the 20th century. It elaborates on theories about political, social and cultural power. There is a key point in this book that I can see everywhere at the moment, namely the concepts of elite and anti-elite.

This view is this: Basically, democracy itself is a myth. You will never have a completely democratic society. By the way, the United States is certainly not a democratic country, it is a republic. But even those well-functioning "democratic" systems tend to have the nature of a republic, lowercase "r" republic. They tend to have a parliament, or have a House and Senate, or have some kind of representative body. They tend to have a representative institution.

The reason is that a phenomenon described in this book, called the "Iron Law of Oligarchy", is basically this: the problem of direct democracy is that the masses cannot organize it. You can't really get 350 million people to organize to do anything. Too many people.

So, basically in every political system in human history, you have a small, organized elite class governing a huge, unorganized mass class. You start from the initial hunter-gathering tribes, and all the other political systems of the United States and modern times, whether they are Greeks or Romans, or every empire, every country in history.

So, a small, organized elite class governs a large, unorganized mass class. This relationship is full of dangers, as the unorganized masses will obey the elite for a period of time, but not necessarily forever. If the elite becomes oppressive to the masses, the number of the masses is far greater than that of the elite. At some point, they may appear with torches and spears. So, there is tension in this relationship. Many revolutions happened because the masses decided that the elite would no longer represent them.

Our society is no exception. We have a large, unorganized mass class. We have a very small, organized elite. The United States... has established a system, and we have two elite classes. We have the Democratic elite and the Republican elite. By the way, there is a large overlap between these two elites, which some actually call “single parties.” Perhaps these elites have more common ground than they have with the masses.

For a long time, we have a Republican elite class whose policies are ultimately represented by the Bush family. We have a Democratic elite, whose policies are ultimately represented by Obama. Over the past decade, there has been a rebellion within the elite on basically both sides of the United States. This is actually a key point in The Machiavellis that change is usually not a direct mass against the elite. What happened was the emergence of a new anti-elite class.

You will have a new anti-elite class appearing in an attempt to replace the current elite class. My interpretation of the current transactionYes, generally speaking, the elites in the world who are currently running the world are found to be not doing well. We can discuss why later. But generally speaking, if you look at the support rate of (Western) political leaders, the support rate of institutions, all of which are falling. What happens everywhere in the world is if you are an incumbent institution, if you are an incumbent newspaper, if you are an incumbent television network, if you are an incumbent university, if you are an incumbent university, if you are an incumbent Government, generally speaking, your public opinion support rate is a disaster. That's basically what people are saying that the elite in power are living up to us.

Then these anti-elites emerged and they said, "Oh, I know I have a better way to represent the masses, I have a better way to take over." My new anti-elite The movement should replace the current elite movement, such as the Democratic Party’s situation. This was Bernie Sanders in 2016, which was Ocasio Cortez and the entire progressive wave. And on the Republican side, it's clearly Trump and his "Make America Great Again" (MAGA) movement and everything it stands for.

But by the way, this dynamic is happening in the UK. The Conservatives have collapsed, and now you have this reform party, there is Nigel Farage, which is very threatening. You have Jeremy Corbyn, who is also an anti-elite class from the left.

So the same is true in Germany. In fact, just this week, something very drama happened in Germany, that the so-called "far-right" party AfD is rising rapidly. There is a leader named Alice Weidel, the first time in German political history, in 50 years or more, the German Christian Democratic Union (CDU) was actually in something with AfD Cooperation reached on the Suddenly, AfD became a viable competitor. They are an anti-elite class who attempts to take over the right wing of the German political system.

So, basically, no matter where you go in the world, there is an anti-elite class that appears and says, "Oh, I can do better." This is between the elites struggle. The masses are aware of this, they are watching democratic societies, and they will eventually make a decision because they will decide who they are going to vote for.

This is why Republican voters decided they were going to vote for Trump instead of Jeb Bush. This is the situation where the anti-elite class defeats the elite. It actually also has to do with criticism of Trump, which is very interesting, that Trump is criticized by the existing elite: "Oh, he is not the people. He is a super rich billionaire, He lives in a golden loft and there are people driving him everywhere. If you are a rural farmer in Kentucky or Wisconsin, you shouldn't think of himIt's your people. ”

The point is never Trump is the people. The point is that Trump is an anti-elite class and he is able to better represent the people. That is the foundation of his entire movement. By the way, The same is true of the media field. Everything you describe is exactly what happens in the media field. Elite media has ruled for 50 years, it is TV news, cable news, newspapers and these famous magazines. Now you have the anti-elite class. Anti-elite Class is Patrick and (famous podcast anchor) Joe Rogen (Joe Rogan). There are many more people.

By the way, if you look at the numbers, it is very clear that the masses, audiences, readers are leaving the old media and turning to the new media. There are elites who are very angry about it. They are angry about all the negative articles about you guys, saying you are all a bunch of white supremacists, and the whole thing is terrible. Like, that's how the world is. So we're just Being in all this. I don't know if "transition" is the right term. It's more like a fierce battle between the old and the new. /p>Patrick: What were the initial seeds that led to the decline of the previous generation of elites, leading to those 11% approval ratings? What do you think this is mainly attributed to?

Marc: There are two theories. One theory is these The support rate is wrong, and the other theory is that these support rates are correct. By "wrong", I mean these support rates are measured correctly, but people give wrong answers.

If you are the head of CNN or Harvard, or you are in charge of any similar institution and your support rate is only 11%…By the way, Gallup has been conducting a very amazing survey for 50 years called "Agency Trust". You can search the "2024 Gallup Agency Trust Survey" by Google and you'll see some very spectacular Charts, you'll find that institutional trust has basically peaked in the late 1960s and early 1970s and has been declining.

By the way, this phenomenon predates the Internet. Interesting Yes, it was blamed on the Internet, but it predated the Internet. So, it was a phenomenon that developed since the 1970s, And it's been accelerating. By the way, these approval ratings have fallen faster since 2020.

They slip like this and then plummet after 2020. TV Network News, I Don't know what the specific number is. It's single digits, and people don't believe it anymore. They don't believe what is said on TV news anymore. By the way, audience ratings are also falling in the same way.

So, one theory is that if you are the head of NBC News or CNN or Harvard, your theory might be: “Oh, people are wrong. People are misled, they are deceived, they are deceived by populists and inciteers, they are false lettersThe interest was deceived. "That's why the concept of "false information" has become so popular. ... People are deceived by malicious actors, populists, and inciteers, and it's only a matter of time until we explain to people that they're cheated. They'll be back Trust us.

So, this is a theory. Another theory is that the elite class has been corrupted. They have been corrupted, dysfunctional, corrupt, and they no longer provide services. Under this theory , these numbers and approval ratings are correct, because every time you see Congress, they are spending your money on all kinds of crazy things without any scruples. Love. If you go to CNN or NBC News, they always lie to you about a thousand different things. If you go to Harvard, they will teach you racial communism, America is evil, etc., these Crazy things.

In this theory, people are right, people have seen through these elites. These elites have basically been in power for too long, they have too much power, they are not under enough The review, they are not under enough competitive pressure, they are already corrupted in place, and they are no longer providing services. The reality may be that both situations are available. It's easy to get the next instigator to show up and just start throwing stones at the power and saying anything.

If you're someone who doesn't have political power today but wants it, the easiest thing to do is show up And start yelling about the current elite being corrupt. Maybe that's a little right, incitementism is a little bit of a role, or whatever it is, but... but I think most of the reason is that the elite being corrupted.

My version is very straightforward, Burnham talks about that in the book. He talks about the "Elite's Loop". He says, in order to make an Elite The class really stays healthy, real, productive, and not corrupt, it requires constant infusion of new talents. It does this through the process of elite cycles.

So, what it will do is, It will identify promising young talents and invite them to join the elite. It does this for two reasons. One is to renew itself. The other is that those people are most likely to become anti-elite. So, it is also to stop it. Future competition. So, my experience started when I was 22 years old and was, “Oh hey Mark, we really hope you come to Davos. We really hope you can come to Aspen. We really hope you can come to New York for this big conference. We really hope you will come to the New York Times dinner party. We hope you can play with reporters for 25 years. "That's what I did, and it's like, "Oh, this sounds great." These are the best people in the world. They control everything. They have the best degree and they graduate from the best schools. They have all the positions of power. They like me. They think I'm great. ”

They kept praising me, I came from the cornfield in Wisconsin. I arrived, I entered the spiritBritish class.

All I have to do is never argue with anything. All I have to do is agree with anything said in the New York Times, agree with anything said in Davos, vote for the candidates you should vote for, donate to the candidates you should donate, never, never, never off track . Then you become part of the elite.

I have a lot of my peers who have done this. Some people are now the largest Democratic donors in the world, they are completely integrated into the elite, they are there, they have a lot of fun, they think it's all great, it's great. Some people think this is good, maybe that's the right thing to do.

Then some people arrive at some point and they look around. It's like the story of J.D. Vance. He grew up in rural Kentucky, or in Appalachian Ohio. He eventually entered Yale University. He was eventually invited into all these internal circles.

Then he finally looked around and he just said, "Wow, these people are not what I thought they were. These people are selfish, corrupt, they are lying about everything, They are engaged in speech suppression, they are very authoritarian, they are plundering public finances. Oh my God, I have been deceived throughout my life. These people are not worthy of the respect they have, and maybe there should be a new elite to take power . ” So, that’s a lot of the debate that’s going on right now. Yes, I'm a case study.

Optimism and pessimism: Will the world be better?

Patrick: If we wear a pair of optimistic glasses, you emphasize early-stage venture capital. You will meet all these young, smart people who are about to build the future. Let's wear a pair of optimistic glasses, assuming that AI has the most positive impact in all areas where we can verify results. Reasoning has become so powerful. So, what other related bottlenecks will hinder the outbreak of the technological revolution we expect? That could be clinical trials in medicine, or something is progressing slower than AI, and AI is not a problem. We will be eager to make progress. But the atomic world, the surveillance world or the clinical trial world, etc., may become limiting factors, rather than intelligence and knowledge. What bottlenecks are you most interested in?

Marc: The way I have always thought about technological change is that there were three lines on the chart, but now it has become four lines. So, one is the speed of technological change, it is a line, and everything is usually getting better and better. And then every once in a while, you'll see these discontinuous jumps, or something gets dramatically better, like what happened to AI last week.

Then you have another line on it, that is social change, basically, when the world is ready to accept new things. Sometimes you see this phenomenon,Things actually exist before the world is ready and for some reason it is not adopted. Then five years later or fifty years later, it suddenly took off and developed rapidly. So, there is a social level, and then there is a financial level above, whether the capital market is willing to provide funds for it. Can it generate returns?

I think the art of being an entrepreneur or a technology investor is trying to cross these three.

So, you try to support something, the technology is really ready, the society is ready to adopt it, and you can actually get funding for it or go public and make it public.

So, you have to align these three curves.

Many of the things we do in our daily work is to align these three curves. The fourth line has now appeared in the past five years. The overwhelming answer for the past four years is government. It was very strange and disturbing to me, when I first encountered it because I wasn't used to it. And I never see us as being involved in politics or being partisan, or we really try to go to Washington for favor. We didn't try to get subsidies either. But we also don’t think we need to do anything to avoid being trampled. Then this happened suddenly.

…

Patrick: What is the way this elite class wants to destroy you the most? How does it manifest?

Marc: This is roughly coincided with a national emotional shift, probably between 2013 and 2017. I grew up in the 90s and politically, I was a default Democrat for Clinton and Gore. There was a "deal" at that time, capital D, i.e., yes, you are a Democrat, but Democrats are pro-business, they love technology, they love startups. Clinton and Gore love Silicon Valley. They love new technologies. They are always excited about what we do. They are always willing to help us if other countries come to target us, or something else. They always try to help us and support us.

Yes, you can be a pro-business, pro-tech Democrat. it's great. You can make a lot of money. People will write a lot of great articles about you and then you donate all your money and you become a philanthropist, which is great.

You are dead, your obituary will say that he is a great entrepreneur and a great philanthropist, and everything is wonderful. Basically, every aspect of this deal has collapsed since 2013. This is shown in many ways, but first of all, media reports. Official institutions of mainstream media are starting to turn to us, and everything we do is evil. This is actually quite surprising. In 2012, social media was seen by mainstream media as an absolute, pure good thing because it helped Obama re-election,….

Everyone knows that it will only elect the right political candidate,…. Then in 2016, the narrative completely reversedNow. Social media, as well as the Internet and technology, are destroying democracy, and everything is being destroyed. So, media reports are like canary in a coal mine.

Part of the reason is that the employee population has been radicalized, by the way. There was a strange situation where these large investment managers showed up and asked you to take a radical political stance in the company, which was completely ridiculous at the time. Then eventually, the administration itself emerged, and the Trump administration’s bureaucracy began to do so, which was beyond his direct control.

But under the Biden administration, it turned into an organized movement, which I describe as destruction, accompanied by endless prosecutions, investigations, Wells notifications, debanking, scrutiny , attack, trying to destroy the entire industry in an all-round way. Of course, this is ultimately why we react. My hope is that it's all over. That is, the new government is taking a very different approach and stop doing all of these things anymore.

Then my hope is that the next Democratic government will realize that attacking tech and attacking startups is not actually necessary. In fact, this may be anti-productive because if you drive Elon Musk out of your camp, there are consequences. I talk to many Democrats, we support many Democrats in the company, many members of Congress and senators, and I will talk to them again next week.

Basically, what they told me is, look, there is a civil war within the Democratic Party, on the one hand, people of us think the party should go back to the middle and stop attacking capitalism, attacking commerce and attacking technology, just win again election.

There are then some people who think that the party actually needs to become more radical, we need to be more distinct from the other side, we need to become more extreme in economic, scientific and technological and social policies. They are fighting for it. My hope is that they will return to the middle so we will never have to go through it again. We can maintain a positive relationship with both parties, but we will see what will happen.

Patrick: Like many others, I am very interested in the nature and status of global supply chains. When you dig into the ingredients of medicines, or the ingredients of many other things, you will see how interdependent the world is, especially the United States’ dependence on the outside world, used in general supply chains. I'm curious how you think and hope this state will evolve over the next decade or so, because obviously there's a reason we're going global. But now, there are indeed many fragile links in the global supply chain. How do you view the evolution of this part of the economy and economic story? Going back to what you just mentioned, you want the United States to win supply chain manufacturing, how the United States will win this competition, and all these exciting ideas you hear today.

Marc: Yes, it's really important, it's very different from the past. …As you know, the complexity of the supply chain. Taking the iPhone as an example, this is a typical product. There is a file you can download online, which may be a bit outdated. But it lists the components that make up the iPhoneAnd where these components come from. A file I read ten years ago may now have an updated version, but a file I read ten years ago shows that at least at that time, the parts of the iPhone came from 40 different countries.

So, when the iPhone was assembled at its Foxconn factory in China, 39 countries had actually sent the parts over, which were assembled into subcomponents of subcomponents and then became components. The same goes for cars, and the same will be for robots, anything complex, anything computerized or mechanical will have this property. By the way, this is actually hard to get from the trade numbers, because I believe it is correct.

China actually gets all the credit of the entire iPhone export value in the export figure, although the economic appreciation that occurs in China is actually a single digit percentage. Because most of the stuff in the iPhone comes from 39 other countries. The analysis you really want to do is the so-called economic value-added analysis. You basically want to say, well, in the $1,000 you go into the iPhone, where is the value of these things coming from, in dollars? The answer is from all over the world.

This is about simply outsourcing offshore or reversing the debate on globalization, and we are not talking about bringing steel mills back to the United States from China. We are talking about unraveling a supply chain involving 40 countries, and things from these countries are shuttled back and forth because everything is being built and assembled. By the way, this is also a problem with the modern economy, which conflicts with reality in many ways.

…

There is then political and economic pressure, the American political system assumes that for 30 years you can offshoring manufacturing from the United States, and those who see all factories shut down The Midwest and Southern communities just sit still and they will come up with other solutions. In many parts of the United States, they never came up with new ways. It turns out that they can still vote.

Part of the reason is that in my country (US), many people have been radicalized because governments and businesses seem to think it’s OK to hollow out the economy and send everything overseas.

So, part of what happened in the American political system is that they decided they no longer accept this practice and they would vote for something different. This view was made at the time, but the argument for economic efficiency won and brought benefits. It paid off in some ways. But many people in the United States have been radicalized. I come from a place where a lot of people are radicalized because governments and businesses seem to think it’s OK to hollow out the economy and send everything overseas.

So, even if you get rewards from economic efficiency, your political system may not be able to bear it. You may regret it very much. I don't think there is an easy answer here. Anyone, my point is, whoever says there are simple answers here are wrong. This is complicated.

The possible situation is that the world will remain highly interdependent and there will be a lot of pressureand the fluctuations back and forth. This dynamic will continue along with tariffs and trade negotiations. It will be a continuous process with twists and turns along the way, but fundamentally the world will remain interconnected in many ways and we will try to deal with it.

The problem is that if there is a war or a more serious outbreak at some point, or something like that, this interdependence can be severely stressed to the point of breaking down. I hope this won't happen. But to a certain extent, the more interconnected the worlds are, the more resilient it is, because there are more ways to do things, people have more ways to adapt, and everything can change. Then in some ways, the more interconnected the worlds, the more dangerous things are, because if one of them breaks, the entire system will break. So, here's a real tug-of-war.

Tan Yushu and China's robot industry: "This specific environment is called Shenzhen"

Patrick: There is another field lurking at the forefront of technology that I haven't seen you talk about too much, that is robotics. Everyone is very excited about its potential. It's easy to imagine a humanoid robot that can do everything around you that humans don't need to do anymore. To make this world a reality, a lot of technological breakthroughs are needed. What do you think will happen in the field of robotics? What is overrated? What is underrated? How do you view it?

Marc: I'll list four things. So, I would say cell phones, drones, cars and robots. Basically, this is the ladder China is climbing. By the way, this is not just a product, but a ladder to the entire supply chain. So, China became where all phones were assembled and manufactured there. So, as you know, they have built a complete ecosystem in China with thousands of specialized companies that basically make all kinds of electronics and hardware, machinery and computer-related things.

This particular environment is called Shenzhen, which is a cluster of thousands of companies that basically manufacture all kinds of electronics and hardware, machinery and computer-related things. So, they use this supply chain to win the drone market for China. Consumer-grade drones, like DJI drones. Basically, China has won the global drone market, with their market share exceeding 99%.

…

In many ways, a drone is like a flying mobile phone. It has a lot of the same equipment and then it has something new, but they want to get into this field, at least until recently. Now they are entering the automotive field. The reason is that a modern self-driving electric car is more like a laptop running on the wheel, or a smartphone running on the wheel, rather than a traditional internal combustion engine car.

The US Tesla isIn one example, Tesla is a computer wrapped in a frame with many batteries with some tires outside. A good illustration of the change is that if you go to the service area of ​​a traditional car dealer, compared to the service area of ​​a Tesla dealer. The traditional automotive industry service areas are full of oil and dirt, everyone has work clothes, and they wipe their hands with a dirty cloth all day.

You go to the service area of ​​Tesla dealerships, which is like an operating room. Everything is clean because it is an electric car and has no internal combustion engine. All this oil and dirt stuff is gone, it's just a computer. The Chinese are basically doing what they have done in the automotive field now, i.e. they have built a complete ecosystem that leverages these other supply chains. They have built a complete ecosystem with all the components needed to build self-driving electric vehicles. Now they are bringing these cars to the market. Suddenly, they become very good, just as good as Chinese phones and drones, they are completely modern, very advanced, very cheap, and at the forefront of technology. Cars have also become very good, and they are only one-third or one-quarter of the price of similar cars in the United States.

The fourth stage is the robot. If you have a supply chain for phones, drones and cars, you have almost everything you can do to build robots. This is the next stage. They are doing this. Of course, there are Elon and other companies in the United States that are making humanoid robots. I hope and expect them to do a good job. But China is definitely doing the same.

The company I pay most attention to is a national champion company in China called Unitree (Yushu Technology). We weren't involved, but Unitree sold robot dogs comparable to those from Boston Dynamics. Boston Dynamics robot dogs sell for between $50,000 and $100,000, which is why you rarely see them. Unitree's dog starts at $1,500, by the way.

We have two and they are great. They can do backflips, they can climb stairs, they can talk to you, they have large language models built in, they can teach you quantum physics while they run around in your yard, which is great. Then they are now starting to launch humanoid robots, which are much lower in price. They are definitely moving towards robots.

It will be a real tug-of-war if you believe humanoid robots will appear, and I do believe it, and on a large scale, if China is willing to make it for $10,000 or $20,000 They, we can buy a billion, and suddenly, we have robots to build houses, do gardening work, do everything you want robots to do, wait for you, then China makes them and sells them to you, and they Very cheap and works well, which is great.

…

Tels and drones are already a fierce problem, but cars and robots will be even more intense.This hasn't exactly happened yet, because the robotics field hasn't completely exploded, but I think the robotics field will explode in the next few years.

Patrick: It's very interesting to watch the competition to build bodies and brains for robots. American companies like Physical Intelligence are working to build datasets that we don’t have yet, just like the open networks we once had to train AI. Have you seen some exciting areas where many young people and companies in these areas make you excited, but do you feel the market is not aware of what is going on and what is possible?

Marc: I think it might be biotech (Biotech). The good news is that in the modern world there are a lot of people who are interested in new technology and a lot of people will talk about it. When I was a kid, the early adoption market was very small. So, there are only a very small number of people who want their first PC or something.

Now you have 50 million or 100 million early adopters who just want the latest stuff and are talking about it online all the time. So I'm not sure if there is too much delay right now, but maybe in the field of biotechnology, everything, like life extension, embryo selection, possible reproductive techniques, to get embryos from stem cells, for example.

Get embryos from stem cells, you know, you may know a lot of people who have a situation where people have fertility problems when they are young, or they have fertility problems when they are at a certain age, but they want to be more There are many children, and then they are forced to make some difficult choices involving test tube baby (IVF) or different types of donors.

It looks like we will be able to get embryos from stem cells, so you can have a real biological child at a later age. External pregnancy is still a while, but maybe at some point it will be a big problem. People often talk about birth rate. Well, if you could continue having children in your 60s, if you could have a dozen kids through an external pregnancy, would more people choose to do that? Maybe.

So that's one aspect. Another possible thing is genetic optimization. So, an endless hot topic is intelligence enhancement. Now that we have CRISPR, we have gene editing technology.

Scientists are then finding hundreds of genes corresponding to their IQ. So, you should have the ability to improve your IQ, which has caused a series of downstream problems.

…

Patrick: Very interesting.

Marc: Everything I just described is becoming possible. They have incredible meanings in health, society, and more, and these meanings will be revealed in the next few hundred years. So, I think people may start to realize more about the fact that there is more discussion to do in these areas than what we are doing now.

…

Patrick: A quick last question. ..., except for the frontWhich book would you choose besides the Machiavellis mentioned?

Marc: I still strongly agree with a book called The Weirdest People in the World by Joseph Henrich. This book may be ten years old, but I don't think it's getting much attention. This book is very insightful in understanding the nature of culture, especially the nature of different cultures.

As you know, there are so many things in our current politics that are related to Western culture, as well as immigration, the meaning of all these different debates, etc. For me, this is the most informative book trying to understand how to think about culture.

Patrick: Marc, thank you very much for taking the time.

Marc: OK. Thank you, Patrick.

|Original link: https://joincolossus.com/episode/the-battle-for-tech-supremacy/