Author: Rhythm BlockBeats

When talking about DAO in the Chinese community, Wang Chao is an unavoidable name.

In 2020, after stepping down as a partner of Bitpie Wallet, he focused mainly on studying the emerging DAO at that time. Due to his daily work of communicating with a large number of teams when working on the wallet, he continued this style in the study and research of DAO. Over the past few years, he has talked to two to three hundred teams involved in DAO. In this process, he wrote down his observations and thoughts, becoming a guide for many people to understand and participate in DAO.

Apart from DAO, he has also made many investments. At the end of 2022, he participated in the establishment of a DAO specifically for investing in AI — Metropolis DAO. You may not have heard of this name, but they have invested in some hot AI projects including Pika, Glif, and Altera. Recently, BlockBeats had a conversation with Wang Chao. He introduced Metropolis DAO, AI investment, and his personal observations and thoughts on the current development of encrypted AI as a participant.

"ChatGPT made me believe in the potential of AI all at once"

How does your previous industry experience affect your current AI investment?

Wang Chao: My industry experience can be divided into three main areas: traditional software, wallet development, and investment.

After graduating from university, I worked in the software industry for more than ten years, starting as a technical staff and later taking on management roles. I have been involved in various projects, from emerging internet projects to traditional service outsourcing and custom development. During this process, I came into contact with a lot of AI from that era. At that time, we were one of the earliest providers in China to use Microsoft's cognitive services to provide solutions for enterprises. Therefore, I am definitely not unfamiliar with AI.

I entered the encrypted field in 2017 and worked on Bitpie Wallet for a few years. But in fact, I had been personally involved in cryptocurrency trading and industry investment since 2013, and had been paying attention to industry trends. In 2015, the concept of blockchain became popular, and "The Economist" magazine published an article titled "Trust Machines: How Bitcoin's Technology Is Changing the World," which quickly popularized the concept of blockchain. 2015 was thus called the "Year of Blockchain." In the B2B field, both software companies and financial institutions were experimenting with various attempts on consortium chains and private chains.

Even though I was still working in traditional industries at that time, we had already invested some technical resources in exploring this direction, including collaborating with clients to conduct POC (proof of concept) experiments. At that time, I really didn't anticipate that AI would make such breakthrough progress in just seven or eight years, nor did I expect AI to combine with blockchain technology and play a role in many fields.

After leaving the wallet in 2020, I actually spent more time on investments over the past few years. Some friends are familiar with me because of my special interest in the DAO field, exploring many DAOs and writing some articles. But in fact, the productive things I have been doing in recent years are in the investment field, including investments in the encrypted field and some other industries.

When did you start thinking about making some AI investments?

Wang Chao: Because of my technical background and years of work in the software industry, I have always been sensitive to new technologies. Since I was young, I have liked playing games and have been eager to explore new things. This interest is not directly related to my involvement in the encrypted industry; it's just that new technologies always catch my attention. But I didn't focus all my energy on AI; my investments in the encrypted field have never stopped.

Specifically regarding AI, I started paying special attention around October 2022. There were two factors at that time: first, the attention to generative AI such as Midjourney, and second, a friend shared a project he was working on with me. Although it seemed very futuristic at the time, it sparked my interest and research in AI. About a month later, ChatGPT was released, and I immediately believed in the potential of AI, realizing that it was not as distant as I had thought.

Since I had been paying attention to the DAO field and had participated in other investment DAO projects, I discussed with some like-minded people whether it was possible to create a DAO specifically for investing in AI. Everyone thought it was feasible, so we started to push forward. The release of ChatGPT attracted a lot of funds and attention, which greatly facilitated the successful launch of this investment DAO. In February 2023, Metropolis DAO was officially launched. Through this platform, we have come into contact with more outstanding founders and new projects, naturally devoting more energy to studying and investing in the AI direction. However, it is important to emphasize that this investment DAO is something we participate in during our spare time and is not the main work of any member, including myself.

At the beginning of Metropolis DAO, the AI craze had just begun. Were there very few AI projects available for investment at that time?

Wang Chao: At that time, there were actually quite a few AI projects available for investment, and now it's even more difficult to invest than it was back then. Although we had been planning this for a long time, it wasn't until February 2023 that we actually completed all the compliance structures, secured the funds, and officially started looking at projects.

By February 2023, AI was already quite hot, with many entrepreneurial projects emerging, but attention and funds were not as abundant, so many early-stage projects were willing to communicate with us, and their valuations were not high. Now, a year and a half has passed, and the valuations of many projects have become higher, and a large amount of funds have poured in. As an investment institution, the competition has become greater. I think it's even more difficult now than it was back then. At that time, some projects were really quite inexpensive, and some projects we talked to but didn't invest in ended up developing very well.

"Investment is highly related to connections and circles"

Are AI projects willing to accept investment in the form of a DAO? In my opinion, they still prefer the endorsement of traditional VCs.

Wang Chao: If we are not the lead investor, our fund size is not large enough to lead the investment. It doesn't matter if we are a co-investor. Generally, for lead investors, they hope to get money from top VCs such as Sequoia and Lightspeed, while co-investors are more about making friends; if the resources are helpful, they will give some shares, and if not, it doesn't matter. The key is whether you know about the project's financing, whether you are in the circle. Often, when you find out about a new project, they have already announced the end of the financing. At that point, it's too late, unless you are very optimistic and quickly contact them to say you want to invest in the next round; this situation also happens.

Investment is highly related to connections and circles. Most of the members of our DAO are from New York and Silicon Valley, with about half coming from the encrypted industry and the other half from other industries. Some of these members are very active and get to know new founders through connections and various activities. Of course, we cannot get in touch with all the projects on the market. After all, not everyone will come to you like Sequoia. But our fund size is there, and being able to accept a portion is enough.

Indeed, some projects are skeptical about the DAO form. The skepticism comes from two aspects: first, the founder may not understand encryption, naturally dislikes encryption, or thinks that DAO's KYC is difficult to do. In fact, we can do KYC because our off-chain structure is compliant, and all entity accounts are available. It's just that they don't understand and need time to explain. Second, they are worried that if your DAO has 30 people, will it create additional communication costs. Some founders are open and willing to establish contact with everyone, seeking potential collaboration. But some founders are concerned that if you invest, a bunch of people will come to them. In fact, they won't; we have a fixed point of contact with each project.

Overall, there are not many communication barriers. Most of the time, it's that we don't know the founders of very good projects, haven't been able to contact them, or the allocation is too tight, and we have to give up co-investing.

How many projects has Metropolis DAO invested in so far?

Wang Chao: We have invested in nearly 20 projects. Some of these projects have already publicly disclosed their financing information, and I will chat about them on WeChat. There are also some projects that, although not strictly required to keep their financing information confidential, it is not convenient to publicly disclose that we have invested in these projects because their round of financing has not been announced. In most cases, if the project has not publicly disclosed its financing information, we also remain silent.

What are the different tracks of the projects you have invested in?

Wang Chao: The projects we have invested in are quite diverse. First of all, anything I say today does not represent the official position of Metropolis DAO, it is just a personal experience sharing of participating in an investment DAO. Personally, I have also invested in about ten AI projects. Overall, non-encrypted AI projects account for more than two-thirds. We have also invested in some Crypto AI projects, but the proportion is smaller.

In terms of tracks, we have not systematically sorted out each direction like Sequoia does, because we do not have the resources and status to do so. Most of the time, we look at whatever we come across. We have some preferences, and projects that are not within our preferences may be abandoned after a quick look. Therefore, our investments are relatively random and it is difficult to systematically divide them by track.

Currently, our investments are concentrated in several areas. Generative AI is one of the main directions, especially creative projects related to multimedia, such as Glif, Scenario, etc., accounting for 5-6 projects, which is the largest category. The second is projects related to Agents, whether it is AI Agents combined with encryption or pure AI Agents, we have bet on some of them, with about 5-6 projects in this category.

We also have a category called "brainstorming projects" as an investment direction. These projects can think of things that others cannot or dare not think of, and have their own methodology to achieve them. These projects generally do not have clear application scenarios, have higher risks, but we believe they are worth betting on. Because innovation always comes from the edge. It doesn't matter if the monetization path is not visible for the time being; answering how a new scenario driven by artificial intelligence and unlimited creativity will be monetized is like going back 30 years to answer how the internet will be monetized. It will be monetized in various possible ways, it's just that we couldn't imagine it at that time. Even if these projects ultimately fail, they may bring about some technological breakthroughs or unexpected new things in the process, which we believe is valuable in the long run and will benefit us in the future.

Can "brainstorming projects" be understood as more about looking at people?

Wang Chao: On the one hand, it is about looking at people, and on the other hand, it is about looking at the creativity of the project. For example, there is a project called Altera that disclosed its financing in the first half of the year. This project was founded by a Chinese professor from MIT, whose expertise is in computational neuroscience, and he has been researching how to reconstruct the human brain in computers. Now he hopes to build "digital humans".

Here, "digital humans" are not just simple virtual beings, but entities with enough action capabilities to interact with humans in some scenarios. For example, he now has an Agent based on "Minecraft", this AI can play as an AI teammate and play games with you, not just a simple NPC, but a real AI teammate.

His long-term goal is to create digital humans that can truly integrate into human society, and these digital humans will have their own emotions. In simple terms, it is to achieve a scenario similar to "Westworld". Many people can think of this idea, but we also select projects. We hope that the founders have accumulated in this field and can form their own methodology to achieve this goal. Regardless of the success or failure of the experiment, we believe it is worth supporting.

What investment preferences have been formed up to now?

Wang Chao: Over the past year, the changes in the entire AI field have been very drastic. This is an extremely active field, with outstanding talents and global funds pouring in, and new things are emerging every day. But this field is still very new, and the paradigm has not yet stabilized. For example, everyone thought Runway's videos were very good before, but when Sora was released, Runway was no longer good. As a result, Sora itself has not been released, and it was Kuaishou's Kelin that came out and became the first truly Sora-level product. Including yesterday's release of Llama 3.1, everyone thinks Llama 3.1 is going to beat ChatGPT. This is an industry where the paradigm is not stable and the landscape is changing dramatically every day.

Investing in this field, especially traditional investment, is extremely high risk. Because after you invest in a project, the likelihood of it failing six months later is very high. In this case, apart from being particularly optimistic about a certain track, what we pay more attention to is the people. This includes their background, the ability to organize resources, and the team's iteration speed. Because the changes are too fast, you may need to adjust direction within three months, and if you don't adjust in time, the project will fail quickly. If the team can obtain enough funds, even if the initial direction is proven wrong, as long as they quickly find a new direction, they still have the possibility of success.

There was a company last year that we didn't invest in, and at the time we thought they were not as strong in their chosen direction as another company, so we invested in another company that performed better in the gaming field. As a result, the company we didn't invest in quickly abandoned that field and turned to a broader creative direction, and they were very successful. Every time I see this company at the top of the AI traffic rankings, I feel regretful because the valuation of this company was particularly cheap at the time. So, I think the execution and iteration speed of people are particularly important.

In addition, we particularly prefer brainstorming projects because human thinking has limitations, and we cannot predict the future at the current time. Some people can come up with things that others haven't thought of, which may happen to coincide with future developments, and we will bet on such projects.

"Encrypted AI cannot be viewed with traditional investment standards"

Is it more difficult to invest in Crypto AI projects because they are harder to invest in?

Wang Chao: I don't think so. First of all, the methodology for investing in these two types of projects is definitely different.

Previously, we tended to use the traditional AI investment approach to look at Crypto AI projects, and many things simply couldn't be seen because the evaluation criteria are different. In this case, some projects were brushed off when they came over. Later, we gradually realized that these two types of projects should not be evaluated with the same standards, and the focus is also different. So we started to adjust and invested in some Crypto AI projects.

The crypto community has a bad habit: once a concept becomes popular, a large number of similar projects suddenly appear, and the market becomes very crowded. The backgrounds of each team look good, but truly unique things are not visible. In this case, we will be relatively cautious. I think this is the result of the combined influence of market conditions and our past preferences.

For us, investing in crypto projects is not particularly difficult. Because close to half of the members come from different crypto funds, and seven or eight of them are partners or founders, our resources in the crypto AI field are much larger than in the traditional AI field. However, everyone in their own funds is already looking at these projects every day, and it is not particularly necessary to invest through a DAO, which is also one of the reasons for the fewer investments.

What are the better paradigms for the combination of encryption and AI now?

Wang Chao: In fact, there are many, and I can give dozens of good paradigms. I generally look at it from three main directions: first, the use of encryption technology in AI to solve problems that AI itself cannot solve well at present, but it is still AI, just using encryption technology. Second, AI technology used in encryption, such as smart contract auditing, whether AI can be used to audit smart contracts.

What is more attractive is decentralized AI. This direction can be divided into many, such as using decentralized methods for data collection. Data is extremely scarce now, and almost all available data has been used up, and what is left is either too expensive or lacks the coordination ability to mobilize large-scale data collection. After data collection, the management of data ownership is also a major challenge, which is obviously a problem that can be solved by the combination of encryption and AI.

In addition, there is the decentralization of models and computing power. Although many people think that the decentralization of computing power may not belong in the AI field, I think computing power is definitely the most important part of AI. There are too many directions to list them all.

I personally have a particularly strong preference for the data direction. In the traditional AI field, everyone's desire for data has reached an extremely serious level. Whoever can bring data, whether it is compliant or gray, will be welcomed. For example, when Sam Altman released Sora, he was asked if he used data from YouTube, and he avoided the question because it was obvious that he did, and it was unauthorized. There are too many things that can be done around data, and there is a great opportunity for encryption technology here. For example, can we use a "collect to earn" form to encourage everyone to contribute a certain specific form of data?

In the past, most "XX to earn" models fell into Ponzi schemes because they did not create real economic value. However, data collection will definitely create economic value. Whether as individual data purchased by companies or as a collective data set contributed to AI training, there is enormous potential. At the same time, the digitization of this data is helpful for the next era of human civilization.

By designing a reasonable token economy, finding areas with Product Market Fit (PMF) or Data Market Fit, and ensuring that this data is truly used, purchased by customers, and widely used for AI training through a decentralized approach, are all directions worth exploring. Of course, many projects are already working in this direction, each focusing on different data fields. Some projects have great potential, while others may just be riding the trend.

Computing power is a well-discussed topic. Everyone focusing on AI and Crypto is paying attention to this field, and many projects have emerged. I believe that computing power is the foundation of the entire AI operation, and it cannot be emphasized enough. Decentralized computing power will definitely see huge development in the future.

However, all platforms are still not good enough, including some very popular platforms. First, the technical infrastructure is not good enough. Is it truly decentralized now? Not really, many are actually just computing power matching platforms, and the structural system is still somewhat centralized, just using encryption technology for settlement or incentives. There is still a long way to go to achieve a completely decentralized computing network.

Secondly, in terms of distributed training, it is not mature enough. Because real AI training requires huge data transmission, distributed architecture is difficult to meet. There are indeed many startups working on distributed training architecture, whether in Web3 or Web2, but mature solutions have not been seen yet. This means that the architecture used for training is not mature and can only be used for inference, while there are many solutions for inference, training is facing a shortage of computing power. If a decentralized platform is chosen for privacy reasons, but it is found that privacy protection is not good enough, this is also not mature enough.

Furthermore, I believe that the entire Agent economy has not yet developed, but I am very optimistic about the future development of Autonomous Agents. If the intelligence of future Agents significantly increases, will they choose to rent computing power on AWS or choose a decentralized network where no one can shut down the computing power, ensuring the continuity of work for the next 5, 10, or even more years? I think the answer is undoubtedly the latter.

Although Agents have not reached that level yet, I believe that in the future, there will definitely be a decentralized computing power and cloud network that will occupy a very important position in the market.

Many people say that the prosperity of AI Agents will be a key factor in the mass adoption of encryption. When do you think this key point will come?

It's difficult for me to judge this now. I have indeed seen many projects that provide infrastructure for AI Agents, including some projects that attempt to put AI Agents on the blockchain. But at present, both the maturity of the infrastructure and the application scenarios for blockchain are still in the early stages.

I think the first large-scale scenario where Agents are combined with encryption may not be on the blockchain, but rather the use of encryption technology for settlement when Agents are involved. Using encryption technology is much more convenient than opening an account with a traditional bank. If encryption technology itself, such as USDC, can further comply and be accepted by more people, then scenarios where encryption is used as a means of payment may mature first. Of course, the premise is that the network and capabilities of the Agents also need to mature. Because at present, Agents are not mature enough to be entrusted with handling financial flows. Optimistically, a very powerful AI Agent may emerge in the next three to five years. At that time, these AI may widely use encryption technology in their business scenarios, rather than placing themselves on an encrypted network.

Let's not talk about putting Agents on the blockchain, but rather their use. If the future consists of a society composed of tens of billions or even hundreds of billions of Agents, there will be very complex economic relationships between these Agents and between Agents and humans. Such a huge economic network, even if only a part of it is supported by an encrypted network, will greatly increase the volume of encrypted transactions by hundreds or thousands of times. This in itself is a huge breakthrough.

In the future, Agents may truly make the blockchain their habitat. Agents will grow out of the encrypted network, live in the encrypted network, and obtain a near-perpetual state through the encrypted network, continuing to work according to their free will. This may be a bit further away, but I believe this is the future.

Now there is already a specialized DAO like Metropolis DAO that specifically invests in AI. Will there be DAOs specifically dedicated to developing AI in the future?

I believe there will be. First, if we expect organizations like OpenAI to operate entirely in a DAO manner, such as Bankless DAO, where all work and resources are placed within the DAO framework, this goal is still very distant. Because it is very difficult for DAOs to quickly handle such complex work. It's not impossible, but it will take a long time.

It took companies 400 years to develop into the highly developed forms of the 20th century, and DAOs still have a long way to go to reach that level. However, if it is a somewhat decentralized organizational model that delegates governance to a DAO, I think it is possible to achieve this in the medium term. Although OpenAI has not fully done this, its structure already embodies decentralized thinking, involving the entire society (although the actual effect is not fully achieved). And the community-based approach of Anthropic goes even further than OpenAI.

For example, Stability, Emad (founder of Stability AI, former CEO) initially wanted to make Stability a DAO, but later found that it was not feasible given the current development status of DAOs, so he ultimately chose a centralized approach. If you look at the first round of investors in Stability, they are those few crypto funds from Seed Club Ventures, and Lightspeed Ventures came in the second round. I believe he initially thought of doing it this way, but in the end, he compromised.

In addition, if we do not pursue complete collaboration through DAO, we can assign some modules, such as data collection, to a data collection DAO, which is also an important form of AI development. Or in the training and fine-tuning of models in specific fields, whether a DAO in a specific industry can gather talent to collectively contribute to a specific vertical model, these are all possible directions. I am quite optimistic about these directions.

"You need to raise enough money to survive the cycle"

Currently, the market value of Crypto AI is 30 billion, while DeFi is 88 billion, Ethereum is 400 billion, Bitcoin is 1 trillion, and NVIDIA is 3 trillion. If Crypto AI develops well in the future, what kind of market value level do you think it will eventually reach?

I am not good at predicting the secondary market, so I definitely do not have a clear answer to this question. First, I am not good at it, and secondly, I think it will not be easy to distinguish between AI and encryption in the future.

I believe that in the future, encryption and AI will definitely be deeply integrated, which will bring about a phenomenon. For example, a project we recently invested in provides encrypted payments for AI Agents to hire human labor. It is not a crypto project, it is just an AI project that uses encryption technology, and it will not issue coins. But if this project succeeds, it may create a huge market where a large amount of economic activity flows through it every year, and its valuation may increase by hundreds or even thousands of times, reaching tens or even hundreds of billions.

Whose market value would you consider this to be? Is it considered part of the crypto market or not? If it becomes a huge market, bringing in a large amount of capital flow, it may cause Ethereum to rise by 50% and Polkadot to rise by 30%. Whose part of the rise would you consider this to be? Would it be considered part of the Crypto AI market or the traditional AI market? I think it will be difficult to distinguish the two when encryption and AI are deeply integrated and widely used in the future, making it even more challenging to calculate market value separately.

If we analyze it from a data perspective, we can look at it this way. People expect that in the next ten to fifteen years, AI technology will significantly increase global GDP. Currently, the global GDP is approximately 100 trillion US dollars per year. If AI technology can create another 100 trillion GDP within ten to fifteen years, then I think it is reasonable for AI technology, as the underlying support, to capture a 20% share of that, which is 20 trillion.

In this 20 trillion, electricity and computing power as the underlying support may take 10 trillion; the model layer, such as companies like OpenAI, may take 5 trillion; the application layer may take 4 trillion; and the remaining 1 trillion may go to the service layer. Many of the products created within this 20 trillion will use encryption technology. Even if the combination of encryption and AI only penetrates 1% of the overall AI market, this means that Crypto AI may capture 200 billion from it.

Currently, the value-capturing ability of blockchain networks is very weak, and price support comes more from narrative and meme-driven factors. The real value capture from network usage, I guess, may only be around 2 billion US dollars per year. If encryption and AI are deeply integrated and begin to be widely used, this may transform the crypto market from one that relies on narrative and emotions to one that is truly supported by value-capturing ability. However, this will take time, possibly ten to fifteen years. Do not expect to see such changes within one or two cycles. The current cycle may end in one or two years, and there may not be particularly strong projects in the next cycle three to four years later. But as long as there are teams working seriously in this field, enduring industry cycles, they may become industry leaders in eight years. Even if they don't make it through, their innovation and validation in technology will still make a huge contribution to the industry's development.

Recently, the new project Sentient, co-founded by Polygon, raised 85 million in its first round, close to the funding scale of traditional AI. However, it has sparked continuous controversy, and many people think it's just a PPT project. How do you view such a strong lineup of investors and star founders in a Crypto AI project?

From an investment perspective, I hold a neutral to slightly positive attitude towards such projects. Why neutral to slightly positive? Because I have also seen many projects without star investors that ended up doing very well and had their own innovations. I don't think it is necessary to have star investors or a glorious background to succeed. But undoubtedly, a good investment background can help a project gain consensus. A team that has done outstanding things in the past can be called a star entrepreneurial team, and this is definitely a validation of their capabilities.

Secondly, my current expectation for all Crypto AI projects is that it is not possible to create truly useful things within this cycle. It may all be concept validation, or gaining a small PMF, but its market value and narrative mainly rely on future expectations. A small team that only raised 3 million US dollars may not last until that day, but a large team that is not there to raise money but is serious about their work, if they raised 85 million US dollars in the first round, may be able to raise 300 million US dollars in another round within one or two years, giving them the ability to endure the funding winter. If they keep up with technology with each iteration, they may succeed in a few years or even further into the future.

If you look at the early days of the internet, many companies successfully weathered the winter and eventually succeeded. For example, Amazon was founded in 1995, and when the internet bubble burst in 2000, many companies went bankrupt, but Amazon survived and is now a giant in the industry.

Therefore, I think if such a project is serious about its work, I would give it a slightly positive evaluation. Whether to invest is another matter; it may not be invested in because the valuation is too high, but it will not be considered a failed project just because it raised a lot of money. On the contrary, if they really want to do something, it is definitely a good thing.

Including investment DAOs and personal investments, which AI investment are you most satisfied with?

In terms of AI, it's difficult to describe it as "most satisfied," "excited" is more appropriate. Currently, the AI project that excites me the most is Altera. The team's background and goals are particularly attractive to me, and its funding is also very strong. It was founded less than 100 days ago and has already raised two rounds. The first round was invested by a16z SPEEDRUN, and the second round was invested by First Spark Ventures, a deep tech fund led by Eric Schmidt, former CEO of Google. The Altera project itself involves simulating the human brain on a computer, which may have a good fit with such deep tech funds.

I have high expectations for it and believe that many of its experiments will definitely be helpful for humanity to enter the next generation. If successful, this investment may bring returns of several hundred or even several thousand times. Even if it is not successful, I will not regret it, and I will also be happy to have supported such innovation.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。