Zuckerberg's In-Depth Interview: The US-China AI Competition is Completely Wrong, and the US Shouldn't Expect to Stay Ahead of China in the Long Run

Compilation: Zhi Dongxi

Source: Bloomberg Original

On July 24, Zhi Dongxi reported that last night, just as Meta launched its strongest open-source large model Llama 3.1 with 405 billion parameters, Bloomberg released an in-depth interview with Meta founder and CEO Mark Zuckerberg conducted by renowned tech journalist Emily Chang.

In this interview, Zuckerberg revealed his journey in deciding to build an open-source platform. He stated that the company has always been constrained by platform companies like Apple while developing products. In this wave of technological change, he hopes to establish and control his own platform through Meta's investments in AI and VR/AR, and make Meta AI the most widely used AI assistant by the end of this year.

Regarding Llama 3.1, Zuckerberg believes this model can serve as a teacher for many companies that need customized models. In the future, he plans to create AI social media to help people navigate some challenging social interactions. He also revealed that Meta is already researching Llama 4.

Facebook rebranded as Meta in 2021 and increased its investment in the metaverse. When discussing this decision, Zuckerberg admitted that he did not expect AI to arrive before AR, but he still believes that VR/AR is one of the important technological trends of the future, and they are preparing for it.

AI is an expensive game, with Sequoia Capital calling it "a $600 billion question." Zuckerberg believes there is indeed an issue of over-investment in the AI field at present, but historically, many bubbles eventually became valuable, just not as quickly as people expected. The current over-investment is preparing for the future while also mitigating the risk of falling behind technologically.

Zuckerberg's wife is of Chinese descent, and he is also interested in China. When discussing the competition between the U.S. and China in the AI field, he believes it is difficult for the U.S. to maintain a technological lead over China for 5-10 years; in fact, a lead of 6-8 months is already quite good. He also opposes the U.S. taking a closed attitude toward AI technology, as he believes that openness and decentralization are the keys to America's success.

Emily Chang also posed a rather tricky question to Zuckerberg: What do you think of Altman (OpenAI CEO)? In his response, Zuckerberg played it diplomatically, stating that Altman has done "better than himself in dealing with public scrutiny." However, he also finds it somewhat amusing that OpenAI, under the banner of "openness," is engaged in "closed-source" work.

Facebook has faced controversy over political content on its platform, and Zuckerberg stated that during this election year, Meta will reduce political content recommendations on its platforms and focus more on their social attributes. He himself will also reduce political involvement and will not take sides in this election, but he does think that Trump's "life photo" is a bit striking.

Outside of work, Zuckerberg is a father of three daughters. He generates images with Meta AI alongside his daughters and braids their hair, although according to his wife, the kids are quite critical of Zuckerberg's braiding skills. American teenagers are facing a mental health crisis. Zuckerberg acknowledges the difficulties of parenting and states that Meta will provide parents with sufficient control to ensure that the operation of social platforms meets their expectations.

The interview took place at Zuckerberg's vacation home by Lake Tahoe in the U.S. He enjoys surfing and is passionate about combat sports. His wife, Priscilla Chan, claims that Zuckerberg has endless energy all day long, balancing work, exercise, and caring for the children, and she struggles to keep up with his pace. However, Zuckerberg's greatest passion remains innovation, "Surfing is fun, combat sports are fun, but building new things is the most fun." He holds an optimistic view of the future of AI and technology, believing that despite the challenges ahead, there are still infinite possibilities.

Zuckerberg being interviewed by Emily Chang by Lake Tahoe (Image source: Bloomberg Original)

Below is the complete compilation of Zuckerberg's interview (for readability, Zhi Dongxi has adjusted the order of some Q&A and made certain additions and deletions without violating the original meaning):

Open-source is to build a self-owned platform, determined not to be constrained by others

Emily Chang: I found an article from 2003 in the Harvard newspaper where you discussed open-source. Have you been thinking about this for a long time?

Zuckerberg: Yes, open-source culture is an important part of the tech industry. Without open-source, I wouldn't have been able to create the early versions of Facebook. I was just a student back then and had no funding. Students also have a "hacker mentality," just wanting to access code for their own needs, which is certainly more cost-effective.

Article about Zuckerberg in the Harvard newspaper (Image source: Bloomberg Original)

Emily Chang: Some people think you don't seem like an open-source advocate; you know why they think that, right?

Zuckerberg: Yes, I know why people think that. However, Meta has always been a strong supporter of open-source. This model is also beneficial for our company. The open-source philosophy is reflected in our server design and data center design. We also have the Open Compute Project, which has standardized many infrastructures in this industry. A similar situation will occur in the AI field.

Emily Chang: By open-sourcing models, you have become an important force in the future development of AI. Was there also a consideration behind this decision to influence or control the next wave of technological change?

Zuckerberg: For the past 10-15 years, we have largely relied on mobile platforms to build our applications, which are controlled by our competitors. Because of this competitive relationship, when we build products that we think are good, we can be told by companies like Apple that we cannot release them or must limit certain features, which is a significant blow to us.

Now we are a large enough company. I have made up my mind that I want us to build and control our own platform in the next generation of technological development. AI is one key area, and AR/VR is another important area. To control our own destiny and ensure we can use the most advanced AI technology, we need to develop AI technology ourselves and make it the industry standard. We can also build a technology-centered ecosystem by sharing our AI technology, making it even more powerful.

Meta AI aims to be number one in the world by the end of the year, Llama 3.1 is the best teacher for models

Emily Chang: You are continuing to improve the Meta AI embedded in Meta's products while also launching Meta AI as a standalone chatbot. Why should we choose Meta's products over ChatGPT?

Zuckerberg: We do many things better than ChatGPT. One feature we are about to launch is "imagine." My daughters and I have been using this feature frequently; Meta AI can generate images in real-time as you input queries.

Although Meta AI has only been released for a few months, my goal is to make Meta AI the most widely used AI assistant in the world by the end of this year. I believe we are already on the right track; there are already hundreds of millions of people using this product, and it is very likely that we will achieve this goal before the end of the year.

User interface of Meta AI (Image source: Bloomberg Original)

Emily Chang: You released the Llama 3.1 series of models, which vary in size, including the largest open-source model ever with 405 billion parameters. What do you think this leap in parameter scale can unlock?

Zuckerberg: Since this model is open-source, I believe people are likely to see it as a "teacher" for other smaller models, using this model as a foundation to train models tailored for different applications. Startups, all companies, and even governments may need customized models. It is challenging for them to utilize various closed-source models on the market, such as Google's Gemini or OpenAI's models, to create customized models.

One of our core philosophies is that we do not want a single AI to dominate the entire market in the future. Our vision is that there should be millions or even billions of different models available. I also believe that AI will not become a god. If you are in an organization that believes it can create superintelligence, you might think you are "creating a god." But I think this view is wrong and also very unappealing.

We have nearly 200 million creators on our platform. They are all working hard to build their communities, and users want to interact with them. I want each of them to easily train their own AI and build it according to their ideas. This is like an artwork that creators launch for their communities, allowing for interaction while ensuring that the way of interaction is controllable.

Plans to build an AI social network, hoping to leverage technology for better social interactions

Emily Chang: I want to understand your macro strategy and specific implementation. Will we have AI-generated influencers in the future? Will there be AI-generated text content, and will AI-generated avatars interact with each other? Do you want to create the world's first AI-generated social network?

Zuckerberg: Yes, the scenarios you mentioned will definitely occur. Creating an AI-generated social network is certainly part of it. In fact, this has already become one of the top four use cases for people using Meta AI, and it is also one of the most interesting use cases.

People will use AI for role-playing to simulate some challenging social interactions. Whether in the workplace, such as asking a manager for a raise, or having difficult conversations with friends or partners. I could even role-play with AI to simulate this scenario before today's interview. This is just a tool; there will be no social consequences for your conversation with AI.

Emily Chang: But what evidence do you have that people really want to live in this virtual world and socialize with avatars? Or is this beneficial for us?

Zuckerberg: I believe that people use social networks primarily to connect with others; everything else is essentially noise. The technologies we currently have can help us socialize better. When I first started using Facebook, it was mainly text. With the advent of smartphones, we began taking photos. Improvements in mobile networks made sharing videos possible, which contain much richer content. Watching videos is a better experience, but I don't think this is the end. I hope this experience can become even more immersive.

Overbuilding is also necessary; AI is a valuable bubble

Emily Chang: You renamed the company to Meta, and you are currently investing billions of dollars in the metaverse. Is our future development path the same as you envisioned? Are there any experiences or lessons behind this sudden shift?

Zuckerberg: Our sudden shift is largely because we started being categorized as a purely social media application company. The metaverse is definitely a long-term investment; I think some things are better than I imagined, while others are slower than I expected. Glasses are certainly a better example of progress. They are fashionable and are good glasses in themselves. This is also a great way to use AI.

When we started this project, we did not know AI would become so important. We thought AI might be a thing of the next decade. If you had asked me five years ago, I would have certainly thought AR would arrive before AI. So the reason we adjusted back then was simply to be prepared to navigate different waves as they come.

Emily Chang: Sequoia Capital calls AI a "600 billion dollar question." Many investments are flowing into chips, infrastructure, data centers, and so on, but when will we see returns from these investments? Do you think AI is a bubble? If not, when do you think we will start to profit from it?

Zuckerberg: I think the concept of a bubble is intriguing. In fact, over time, many so-called bubbles eventually become quite valuable; it's just a matter of time. Even the internet bubble, such as the massive laying of network cables at that time, ultimately became valuable, just not as quickly as people imagined. I don't know how AI will ultimately develop; it's still hard to predict. But I believe AI will become an important foundation for the future world.

I think many companies are overbuilding right now, but this is a significant opportunity. Looking back, you might find that we have all spent billions of dollars too much. But from another perspective, I believe that the companies currently investing in this infrastructure are still being rational. Because if you don't invest, you could be left behind technologically for the next 10-15 years.

Research on Llama 4 has already begun, aiming to completely eliminate the gap

Emily Chang: You said your goal is to achieve Artificial General Intelligence (AGI); how do you define AGI? Will you achieve AGI first?

Zuckerberg: We are already researching Llama 4, and our goal is to completely eliminate the gap between us and other companies. On the road to AGI, I believe some important breakthroughs need to be achieved, so it's hard to predict directly whether we will be the first to achieve AGI.

AGI currently does not have a clear definition. But I believe intelligence is diverse; it is not just a number but has various aspects and modalities. I believe that one day we will achieve AGI, and we want to create various AI systems to help people accomplish what they want to do. I think this will ultimately be a good thing for the world.

Critical thinking and value building are important, playing multiple roles enriches personal experience

Emily Chang: What should kids learn today?

Zuckerberg: I think the most important thing is that they should learn critical thinking and build their own values at a young age.

This aligns with my own values when hiring. If people have already shown that they can do something deeply and well, they likely have mastered the skills of learning and can easily learn other things.

Emily Chang: It seems you always give your all when doing things. You exercise, work on other projects, and also have to be a dedicated CEO and a competent dad. How do you find time to accomplish all these things?

Zuckerberg: I believe balancing these things can actually help you do them better. I am indeed a very focused and demanding person, but I also consider the feelings of others I work with and do not concentrate too much energy on one project; I want to ensure everyone has the space to do what they want. Being a well-rounded person and continuously learning has the greatest benefit of providing many different experiences.

Zuckerberg holding his daughter (Image source: Bloomberg Original)

The U.S. cannot maintain a 5-10 year lead over China; openness and decentralization are essential for sustained innovation

Emily Chang: You have always been interested in China, and you have learned Mandarin. What do you know about China's progress in AI and AGI?

Zuckerberg: I don't know much myself. However, the U.S. is indeed facing the question of how to deal with competition from China. One view is that we should close off all technology; I think that is completely wrong. The success of the U.S. is based on open and decentralized innovation, and that is how our economy operates. But I believe that leading companies in the industry also need to collaborate with the U.S. government to ensure we continue to progress and maintain a technological lead.

Emily Chang: So we win this AI war through this approach?

Zuckerberg: The key question is, what exactly do we want to achieve? Do we want to lead China by 5-10 years? I don't even know if that is a reasonable goal; maintaining such a lead is difficult. I think a more reasonable goal is for U.S. companies to maintain a 6-8 month lead over China through continuous innovation in AI; that is already a significant advantage.

Altman is better at public relations than I am; the current state of OpenAI is ironic

Emily Chang: There are many players in this field; how do you view Sam Altman's leadership?

Zuckerberg: He has made significant contributions to the development of OpenAI. At the same time, like him, I have been under public scrutiny and supervision. I have not handled these issues well, but I think Altman has handled them elegantly; I believe he does it better than I do.

But I find it somewhat ironic that an organization named OpenAI is now a leader in the closed-source model field. This isn't necessarily a bad thing, but it is a bit amusing.

Emily Chang: Regardless of whether you agree with these criticisms, Facebook and you have indeed faced a lot of criticism. Why should we trust your AI?

Zuckerberg: We have been criticized for many things. We take these issues seriously and are doing our best to address them. I am not sure if it is fair to treat us this way, but as an important company, scrutiny and regulation are generally beneficial.

One definition of open-source is that anyone can review the product. I believe this puts a lot of pressure on you to ensure the quality of the product improves rapidly.

This election will not take sides; the company will reduce recommendations for political content

Emily Chang: I want to talk about the 2024 presidential election. Facebook has been a focal point in many elections around the world, and you have recently been named by former President Trump. This is a significant election; what impact do you think it might have?

Zuckerberg: This is clearly a very important election; it will be a historic election. The feedback I receive from users is that they actually want less political content on our platform. They use our services for socializing and connecting with others, which is what we need to do. We have given people the choice, but overall, we will also reduce the recommendation of political content. I believe everyone will see that our product's role in this election will not be as significant as in previous elections.

I have had political involvement in the past, but this time I do not plan to participate. I will not support any candidate. However, there are indeed many crazy things happening right now. That photo of Trump with a bullet wound to the face, raised fist against the backdrop of the American flag, might be one of the most striking things I've ever seen.

To some extent, as an American, it is hard not to be moved by this spirit. I think this may also be why many people like him. But our company's role is to allow everyone to express their views on this matter. That is the role we should play well.

The mental health crisis cannot be entirely blamed on social media; the virtual world and the real world do not conflict

Emily Chang: The U.S. is facing a mental health crisis, especially among teenagers. The U.S. Secretary of Health has called for warning labels on social media, saying that social media should bear some responsibility for this issue. Based on everything you know, does he have a point?

Zuckerberg: The U.S. clearly has mental health issues. I think addressing this issue is particularly important for children and teenagers, and it is right to focus on it. I have three daughters, and they are quite young. Being a parent is challenging; we just want to ensure that our children have a good life. From this perspective, my goal is to build our services in a way that meets parents' expectations, providing them with the necessary control to supervise how our services operate.

Current data shows results that differ from general perceptions. Many people act as if there is a necessary connection between mental health issues and social media. I believe the current research has not proven this point. The constant notifications from phones may also affect your mental health, but that is different from social media. Of course, the problem does exist, and we hope to participate in solving it.

Emily Chang: Many of the things you are doing make the world more virtual, but you also love many things in the real world. How do you balance these two goals?

Zuckerberg: I don't see any conflict between the two. A group of people in the tech world believes we can extract our consciousness and intelligence from our bodies and upload it to the cloud. To me, that seems absurd. Part of what makes us human is that we are active entities with our own energy; we are not just what is in our brains. Our energy and our love are essential for people.

Surfing and combat sports are fun; building new things is the most fun

Emily Chang: You have been in this industry for over 20 years; how much longer do you want to continue? You already have so much.

Zuckerberg: Yes, surfing is fun, combat sports are fun, but I think building new things is the most fun. At this moment, I think in terms of a 10-15 year scale for what I am building. I believe AI may take 10-15 years to fully realize, and the next generation of computing platforms we are working on will be similar. My guess is that by then, new technologies will emerge.

Zuckerberg drinking beer on a surfboard (Image source: Bloomberg Original)

Emily Chang: From the perspective of the entire tech industry, it seems we are conducting a massive experiment on humanity itself. How will this ultimately play out?

Zuckerberg: I am not sure there will be a so-called conclusion. I believe there will still be significant development in the future. I know many people are worried that AI might limit our creativity, but I think the opposite is true.

When I started my entrepreneurial journey, I needed to understand a lot of programming knowledge to realize my ideas, but in the future, we may not need that knowledge at all. We will just need to express what we want to build, and AI systems will help us write the necessary code. Of course, we will face various challenges along the way, but I remain optimistic about the future development of technology.