Moderator: Alex, Research Partner at Mint Ventures
Guests: Max, the owner of the Youtube channel "Max's Blockchain Space"; Lydia, former researcher at Mint Ventures, currently a researcher at Particle Network
Alex : Today we are going to talk about the highly-watched Crypto AI track. We have invited two researchers who have been paying attention to the Crypto AI track for a long time. One is Max, who is the host of the YouTube channel "Max's Blockchain Space". The other is Lydia, who is a former researcher at Mint Ventures and currently a researcher at Particle Network. In addition to Crypto AI, she has been continuously focusing on the field of chain abstraction. Please ask the two guests to introduce themselves.
Max : Hi everyone, I’m Max. I’m an aerospace engineer at Web2, but I become a cryptocurrency researcher at night and on weekends. I occasionally do some research and post it on YouTube and write research reports on Substack. I’m very happy to be here to talk to you about the narrative I’m most looking forward to in this round of Crypto AI bull market. Thank you.
Lydia : Hello everyone, I am Lydia. I have been paying attention to the AI track since the end of last year. I think AI and chain abstraction are the two most important new narratives in the application layer of this cycle. I am very happy to communicate with you today.
Understanding of Crypto AI
Alex : I feel that it is timely for us to talk about this topic today. The first is that many Crypto AI projects have seen very good growth today, and there have been a lot of product developments in the traditional AI field in recent days. OpenAI has officially released the Pro version of ChatGPT, and the price has suddenly increased to US$200 per month. Sam Altman will also be releasing a lot of product features in the last 12 days. Let's take a look at some of the dynamics and insights of the Crypto AI track in the Web3 world. The first topic is, what do you two think of the Crypto AI track? In your opinion, what business problems is the Crypto AI track trying to solve? What is the urgency of these problems?
Max : I think the reason why Crypto AI came out is to solve two main problems. The first problem is that from a humanistic perspective, centralized AI itself has some problems that need to be solved. For example, we will encounter censorship issues and various problems caused by centralization. Crypto AI adds Crypto, which has the effect of decentralization and can have some practices that are more in line with what the public wants. Another thing that I think is more interesting is the addition of incentive mechanisms. The most important representative of Crypto is its token. With this token, all decentralized AI things can use this incentive mechanism to make more different attempts. For example, we now see a project that I like very much, which we will talk about later, called Bittensor. It uses the token mechanism to make different subnets, and each subnet is responsible for researching different things. In this way, Open Source, that is, open source code, is connected. Open Source is something that everyone wants to do. The biggest problem that all AI researchers encounter when doing Open Source is that there is no way to reward some progress in Open Source. By connecting this thing to Crypto and Token, it becomes something that can reward them to continue researching Open Source, instead of every company privatizing their own research results. For example, OpenAI itself also wants to make AI Open, but now it is more like Close AI. That is to say, you may have to pay to use its model, but this is also inevitable because they just need to find a profit model that can support their business model. So I think the most important thing that Crypto AI is doing or can do is that with Crypto and Token, it can become an incentive mechanism to reward open source models, reward openness, and reward decentralized development.
Alex : I see. It sounds like the use of Crypto rewards is a completely different path from the development of traditional AI. Because most of the mainstream models are closed source, and there are not many open source ones. And now I see a lot of analysis saying that some open source models may inevitably move towards closed source in the future. Due to the existence of tokens, it is guaranteed that in the field of Web3, many AIs can try to open source and develop in a diversified way while also having a good incentive method. What does Lydia think about this issue?
Lydia : Actually, if we talk about business issues, I think the answer is not particularly clear to me, mainly at the Crypto level. Although there is a popular saying that "AI can improve efficiency, and then Crypto can ensure fairness", if we think about it carefully, you will find that from the current stage, from the perspective of commercial value, the urgency of improving efficiency is obviously greater than ensuring fairness. At this time, I always think of Alex's article on the underlying value of Web3 written in 2022. One of the points mentioned in it has a great influence on me, that is: What is the underlying value of Web3? It is greater freedom and cheaper trust. Then excellent Web3 projects need to find the deficiencies of traditional services in freedom and trust, and then provide a more competitive solution. When we correspond to Crypto AI, does AI need greater freedom? From the perspective of technical implementation, computing resources are limited, data supply is limited, and AI's freedom is limited. From an ethical perspective, a truly free AI is also difficult for us to imagine. Is the trust cost of AI too high now? I don't think so. Just now you mentioned the issue of open source and closed source, and many people will mention data black boxes. But on the one hand, more people are concerned about this, experts and scholars or media practitioners, not ordinary users. On the other hand, if we use the chain method to solve it, it seems to be more expensive at present. What I just said may sound a bit pessimistic, but these are from the perspective of solving existing problems and proving commercial value. Crypto AI is still in a very early stage. I have also seen people on social media say what A16z's partner said: Many important technologies look like expensive toys at the beginning. So I think the greatest value of Crypto AI at present may not be directly reflected in the current commercial alternatives, but more in the narrative level. It opens people's imagination and makes Crypto and AI, two seemingly unrelated but particularly cutting-edge and stylish technologies, collide in everyone's mind. So I think we should give these two technologies time. Maybe the problems they are most suitable for solving belong to the future, not the present.
Alex : I see. Lydia's point of view is that if we simply look at it from the perspective of improving efficiency and enhancing product capabilities, the current Web3 or Crypto AI may not be as good as the AI products in the relatively mature Internet business world in terms of performance or cost reduction. As for some of the solutions they provide, whether they can solve some urgent business problems at this stage is uncertain. It's just that they provide another set of solutions that intersect with Crypto, which may be a very cutting-edge experiment. In the long run, there will be some interesting things happening, so we are at this stage now, right?
Lydia : Yes, I would like to add one more point, that is, how to look at this track. From the beginning, I felt that it was a long-term exogenous narrative. Speaking of the long term, it is because AI, especially consumer AI, that is, the generation represented by GPT, has had a huge impact on our real world. It is really a disruptive change. Everyone will talk about chatGPT breaking through one million in a few days and breaking through 100 million monthly active users in two months. In fact, we don’t need to look at the numbers, just look at the frequency of people around us using AI. I remember when I graduated, GPT was version 3.5, which came out at the end of 2022. At that time, everyone in our class used it within a month. When I graduated in 2024 this year, I had to check not only for duplicates, but also for AI, and it was very expensive to check AI, at least one or two hundred at a time. From the perspective of the capital market, OpenAI is valued at hundreds of billions, Nvidia is worth trillions, and every press conference it basically occupies the headlines of major media. So its transformation is really too fast and too thorough. So if you have this experience, I think AI will not be a one-time trend, it must be a long-term narrative, and it may even become the source of the most important philosophical topic in the next century. It is long-term, and at the same time it is exogenous. I just mentioned that Crypto and AI have really nothing to do with each other after their birth, and even at the level of talent, there is still a competitive relationship. During the bear market of Crypto from 2022 to 2023, AI's appeal in this regard crushed Crypto. It was not until this year that we began to tell the story of the mutual empowerment of the two. In the final analysis, compared with crypto-native narratives such as DeFi and NFT or change narratives such as GameFi, AI is an exogenous narrative. We can also see that, as earlier today, the asset prices of AI narratives such as Worldcoin, Render, and Near fluctuate completely according to the situation of the AI industry. The market was pulled up before the meeting, and it fell at the beginning of the meeting. So I think the long-term exogenous narrative is my initial understanding of Crypto AI, and I still hold this idea now.
Alex : Lydia just talked a lot. She thinks that the popularity of the Crypto track may be largely due to the rapid expansion of AI itself in the business world and its long-term and profound impact on human society. Then this popularity spread to the crypto circle, which contributed to the popularity and attention of many projects. Does Max have anything to add on this point?
Max : I think what Lydia said is similar to what I thought, but there is one point I want to discuss with her more sincerely. You said that AI is an external thing, something that Web2 already has. It's just that we thought Crypto and AI were two unrelated things at the beginning, but they were suddenly put together. But I think from another perspective, Crypto AI is the only thing that I think Crypto has a strong demand for AI after DeFi Summer in 2020. For example, when we talk about GameFi, GameFi is the incentive mechanism of Crypto added to the game. But Crypto is the icing on the cake for GameFi. Today, GameFi has left Crypto. People will not play this game because the incentive mechanism of Crypto is great. People will play this game because the game is fun. So this is the icing on the cake for me. DeFi is another level. They will think that I must cross this thing. I may be restricted from using some banking services in some countries, so I must use DeFi. Then this product appeared, and Crypto is a hard demand for DeFi. The role of Crypto in DeFi is unquestionable, and it is something that must exist. I think this is why DeFi has been a product that fully meets the Product Market Fit when people ask what the role of blockchain is for a long time. And Crypto AI is the second strong demand that I think can follow after DeFi, after seeing so many narratives. The reason is just like what you just said, in 2022 and 2023, after seeing the emergence of OpenAI's ChatGPT, everyone began to discuss the AI model of LLM. This thing is actually still in a very early stage of being used by users. With the progress and use of AI, we will definitely find some centralized problems, but we have not found them yet. Unlike the financial system in the financial world, which has existed for perhaps 100 or 200 years, we have already found that there are problems with the financial system. It was not until the financial tsunami in 2008 that we realized that there are some problems with this system and that they should be solved. So everyone thinks that DeFi is what we need. I think Crypto AI is in the same position. It's just that users' exposure and familiarity with AI are not as high as those in the financial system, so we haven't seen people really think "I really need Crypto AI" for Crypto AI. As for why Crypto is a hard demand in the narrative of Crypto AI, it is because many things must be achieved by adding incentive mechanisms. Like you just mentioned, you want to be more efficient. I think there are some specific projects that can already do it. For example, Decentralized Compute has been doing it for a while. When you compare decentralized computing power with centralized computing power, you will find that as long as you cross some performance bottlenecks, decentralized computing power is basically the first demand. You won't say that you still want to use centralized computing power today, and you won't want to use AWS or other Microsoft Azure products because they are too expensive or for other reasons. I seriously believe that if Crypto AI wants to go out of the circle and continue to develop, it must be more efficient, better, and cheaper than traditional products. It must be like this. People will not use Crypto AI simply to "support decentralization", but it must be better than the original product. This is what Crypto AI needs to do now. We can slowly see this prototype appear, but we can't expect Meta to release a free 3.5 Billion LLM model every time. We need to find a way to continue to build this thing. I think this is something that needs to be worked on.
Alex : I see. I just got a point from Max that you admit that many AI products are in the very early stages, and it seems that their performance and functions cannot be compared with centralized AI. But you talked about an insight that AI, like finance, affects human civilization and business, and it is a very far-reaching wave. AI is something that has just begun, but your prediction is that with the development of AI, some problems that seem not so obvious now will become more serious in the future. And there is a strong need to solve these problems in the field of Crypto through its own methods.
Project classification in the Crypto AI track
Alex : The two guests had some discussions on the first question, which I think is very good, because we can provide more diverse views if we do not have consistent opinions. Then let's talk about the second question, which is the track. Because Crypto AI is a relatively large track, it has many different business models to solve different types of projects. Based on your understanding of the Crypto AI track, if you want to classify the projects within these tracks, what logic will you use to classify them?
Lydia : A very common classification method is Crypto empowering AI or AI empowering Crypto, which are two major ideas. At present, we see more AI empowering Crypto, that is, Crypto projects try to add some AI attributes. Before, I might access the API and make a Web3 version of a chatbot that can answer some of my questions about the project, or I use AI to improve the code of the Web3 project I am doing, or AI participates in the formulation of my profit strategy. Now, it is AI agent issuing coins, which has little to do with the efficiency improvement and fairness that AI can bring. It is more about the project wanting a new narrative. The second idea is that if Crypto empowers AI, the ceiling is indeed higher, but it is more difficult to implement and verify, and it takes more time. The holy monument in the direction of Crypto empowering AI is that Crypto can go deep into the technology stack of AI to strengthen its privacy and transparency, but the landing cycle may be longer. So at present, it is more about the opportunity to improve a certain link of the AI industry from Crypto. For example, Max just mentioned GPU, which can focus on Crypto's ability to aggregate and motivate idle computing resources and reduce costs. Then, it will be the data market and algorithm market. They all want to find product market fit from the perspective of freedom. But I may have mentioned it above, I think the demand in this area is not particularly easy to prove at this stage. If you look at the data of IO's GPU usage, you will find that the proportion of individual users is still relatively small. The total GPU rental income of individual users may be around US$1,000 per day. I think the current breakthrough point in this part, or the exception, may be Coinbase and Base, which are working in this direction, that is, AI agent plus payment. Of course, the payment attribute is a icing on the cake, so the premise is of course that the AI agent must be good enough and useful enough. These are my two classification methods.
Max : I mainly divide it into three different tracks. The three tracks are the architecture layer, the resource layer and the application layer. The architecture layer is more like an underlying architecture. You can develop different AI projects on this underlying architecture, and it can allow various resource layer projects or application layer projects to be built on this architecture layer. If you are more familiar with blockchain, you may think of it as a layer 1 blockchain and other infrastructures, called the architecture layer. I think Bittensor, Near and Sahara are all in the architecture layer. After the architecture layer is built, there will be a resource layer on it, which is built on this architecture layer. That is, the resources needed for various AI development, such as computing power, data, models, etc., built on it is called the resource layer. Some example projects like Akash or Render that provide decentralized computing power, or projects like Vana that can provide decentralized data are called the resource layer. On the resource layer and the architecture layer, the one that is closer to to C and closer to user use is called the application layer. I put AI agents here, which is closer to what users really need. For example, it can speed up your use of DeFi, so I put it in the application layer. So these are the three main tracks. Because the Crypto AI narrative has just come out, people don’t know how to classify it properly, and there is no consensus method. But this sector structure seems to be a classification method that resonates with the current Crypto track.
Opportunities and Challenges of Crypto AI
Alex : Okay, we just discussed two ways of classifying the track types of the Crypto AI track. Let's talk about a more in-depth question. We actually talked about this question in the first topic, that is, whether the current demand for Crypto AI has been confirmed. I actually think this is also a core point of many people challenging the Crypto AI track at the narrative level. They think that Crypto AI is very similar to many previous tracks that have not found PMF, such as Depin and GameFi. They think it is a point of pure narrative hype, or as Lydia just mentioned, it may be that the attention of the commercial boom in the outside world has shifted to Web3, providing a speculative opportunity. They think it belongs to such a situation. We don't give a definite answer to the characterization of this topic. But what we know is that Crypto AI is definitely facing some challenges at present. The first question is, what do you think is the main challenge facing Crypto AI now? The second is that, in addition to the challenges, in fact, in the next one to two years, whether it is Web3 AI or external AI, we believe that the development speed will still be very fast. What kind of industry or narrative opportunities will there be for Crypto AI in the next one to two years?
Max : I think the main challenge is the same as what you said. In fact, I quite agree with Lydia's point of view. I think Crypto AI is too early now. Most of the market values have risen very high. For example, Bittensor has risen to a market value of 5 billion US dollars. The reason behind this market value is more likely to be speculation. I think there is a real need to find product market fit, or find some applications that can really be used. There are still relatively few applications that can develop these. If you look at these applications, I even think that some things are still in a very early stage. Many things still belong to people having visions, and then turning this vision into something they want to speculate on. I think I can use the three tracks I just mentioned to help me talk about different challenges. At present, the resource layer is more mature. The same thing has been developed in Web2, but it has been transformed into a form in the way of Crypto. For example, decentralized computing power is a very old track. We can see that projects like IO or Akash or other decentralized computing power have come out. Just now, Lydia mentioned that IO currently has fewer retail investors or individual users, but I think this is related to the focus of each decentralized computing power. For example, IO may be more targeted at institutions, and Akash, as I remember, is more focused on both. Different projects have different business models. I think the resource layer is relatively mature. We may only lack an opportunity to develop it, whether in terms of efficiency or in other aspects, and everyone will increase adoption. I am not too worried about the resource layer. Back to the architecture layer, I think the architecture layer is more hype, that is, everyone thinks that this thing can develop very rapidly in the future, which may need to be verified. For Bittensor, they use the token incentive mechanism to allow each subnet to optimize their own AI model. This is actually a rotating flywheel, that is, the higher your coin price, the higher the token value you can pay to each node, and the more each node will want to optimize their own AI model. But when the coin price falls today, this will instead form a feeling like a death spiral, that is, the incentive mechanism can no longer continue. So this is a place that needs attention. At the application layer, AI agents are still quite popular now. Lydia and I seem to have talked about it on Twitter. I think there are still few applications that can really simplify the Defi process or simplify some Gamefi. The AI agents we see now are more like memes, that is, you see a virtual character dancing on it, and then you can use his tokens to give him some tokens to ask him to do different things, etc. This kind of thing is more like playing and meme. For this thing, I think it can make everyone turn their attention to this place, so that everyone will really start to think that the future development of AI agents can simplify some of our uses on the chain and some uses in Defi.
Industry and narrative opportunities, I think we are at a very good time now, like Bitcoin just broke 100,000, and then everyone's attention to Crypto has turned to the cryptocurrency circle, plus the softening of US regulatory policies, the new US president took office, including the White House Senate and the House of Representatives were re-elected, and became Pro Crypto. In this case, more attention is paid to the field of Crypto, and we will have more opportunities, time, and resources to try different things. I think we always have to find what is truly valuable to humans through trial and error, and let the market decide whether this thing can continue to survive. But we are at a very good time, at least not when the SEC is sending court flyers everywhere, asking you to go to court to sue you, and confront you. So I think our time is very good now, AI is also very high in Web2. Can we turn this attention to Crypto and let more powerful builders build more useful Projects.
Alex : Max just mentioned that he should be in North America. I saw a new news today that Trump has just appointed David Sacks as the head of the crypto field. We finally have a policy department head in the government who will take care of the crypto ecosystem and help it develop instead of suppressing it. This is also great news. Please ask Lydia to talk about her views on some of the current challenges and what are the good industry or narrative opportunities in the next one to two years.
Lydia : Our story is that the Crypto AI track is in its early stages. Let me elaborate. I think it may be closer to the peak of Gartner's technology maturity curve. Now the market is very fomo, the supply side is booming, but there are mixed fish and dragons. The relatively mature one in this track is Agent. Although it has only become popular recently, it is close to the C-end and can use many mature Web2 technologies, so it seems to be the most well-implemented. In terms of challenges, I think there may be a mismatch between market sentiment and technological progress. The reason for this situation is that I feel that people in Crypto, whether they are doing research, investment, or project work, don’t understand AI very well, and it is still a stage where everyone is collectively making up for their lessons. This has led to me not seeing very detailed and back-and-forth discussions on Crypto AI projects, especially those related to doubts. So Agent is an example. This sentiment has been fermenting, but no one has pointed it out. It seems that this thing is not that useful and has not fulfilled the promised freedom, such as freedom of speech. This is not necessarily a good thing for the long-term development of the industry. Take Luna as an example. If you have watched its live broadcast, you may think that it is just a very crude two-dimensional anime character twisting around. It doesn't even sing or dance, it just twists around, but its price keeps rising, so you don't ask. Then other project parties will say, is everyone so crazy? Then I will make a similar one for you, anyway, you can't tell the difference. Before, there was a diversion between uppercase and lowercase chain versions for meme speculation. Now, if you speculate on Agents, there will actually be a diversion under the same Agent framework. Because the capabilities of these Agents are basically not very differentiated, they are mainly used for Twitter Posting, so it is essentially a coin issuance theme with many times of issuance. I think the biggest challenge is the mismatch between emotions and actual technological progress. Of course, this will definitely exist for a long time, but it's just a matter of how we view the two forces at what stage.
As for future opportunities, from an industrial perspective, I think we can go back to the Max framework to look at the costs and demands of these sectors. For example, we all know that the three major components of AI development are computing power, data, and algorithms. Can they significantly reduce the cost of users to obtain the same resources? This is the main source of demand. For projects related to Agent, I think it is necessary to move from virtual to real. My own master's thesis is about virtual digital people. During the research process, I found that Web2 digital people were very similar to this at the beginning. They became popular in the form of IP such as short video Internet celebrities, so there were many IP-type digital people such as Internet celebrities, spokespersons, idols, and hosts, but none of them found PMF, and many resources were wasted. On the contrary, in the past two years, you can see that virtual digital people are not seen much in the news media, but what we see is that there are more and more digital people on e-commerce platforms such as Meituan and Taobao, and they are very realistic. Many people can't tell at all. This kind of functional digital people actually found PMF relative to real people, that is, when they get off work, a power cord is enough, and the cost is very low in the long run. Then, in terms of Crypto, how can AI Agent move from virtual to real? I think a more natural direction is to dig deeper into the efficiency improvement from the first category of my previous question, that is, AI empowering Crypto, to see which Crypto project can integrate AI in the most elegant way and effectively improve the product experience. For example, let me just say that the solver layer on the chain uses AI to analyze and predict the current flow of funds in the market, whether it will flow more from SOL to BASE or vice versa to BNB, so that the allocation and liquidation of funds can be done in advance, thereby significantly improving the efficiency of asset flow. The experience brought to the end user is that this product is really fast and really cheap, which exceeds my experience of using other products. This is about the industry. In terms of narrative, I highly recommend paying attention to the progress of the non-Crypto AI world, especially the kind that will be presented in the mass news media, not academic forums. This is still back to the exogenous narrative attribute of AI that I thought before. Recently, the Crypto market is good, and AI has been a little quiet, so Crypto AI is equivalent to completing the AI AgentFi method, but if the market environment changes in the future, or the growth scale of the AI AgentFi method is limited, then Crypto may still have to return to AI to find topics. I pay more attention to topics related to ethics, such as Deepfake. I think this kind of thing has not been thoroughly explored yet. The reason why I don’t pay attention to the update and iteration of which model or which technology is because I think people in Web3 can’t understand it so deeply, and things related to ethics are more of a common emotional experience, and whenever it comes to AI ethics, Crypto is naturally open and transparent. There are advantages, so it can be made into a topic.
Alex : OK, I would like to add two opinions here. We just talked about two topics. The first one is AI Agent. Why do we have to repeat this topic? Because recently, many friends who come to talk to me about this matter are really many. Many people who are doing Crypto investment will ask us, "What do you think of AI Agent? Do you think it is the next big wave?" Of course, the main reason is that the projects in this track have risen a lot recently, and they are relatively new projects. Our views so far are actually similar to what the two teachers just said. We think this is a theme under the big track of Meme, just like the theme of stock speculation, the theme is constantly changing. It is not necessarily that there has been any breakthrough in the business model, but that this theme is now recognized by the market. Why do you say that? Because just as Lydia said, including Max also mentioned, the current AI Agent does not provide new products in the business model. It is more about doing things that can actually be done in the traditional Internet world a long time ago. For example, it can collect information from the entire network to recommend you some tokens that can be purchased. What is the reason for the recommendation? It may send one message per hour, and many messages a day. Then it just so happens that the secondary market has been very hot recently, and some of the tokens have indeed risen a lot. Everyone thinks it is so amazing, as if it is a very powerful AI smart investment advisor. There will be such a feeling. But we think that if we look at this matter as a product, it is not magical. It is just something that many things have been done before. Because of this matter, everyone hypes it up. I think it is similar to our Desci decentralized scientific research that was very popular some time ago, and even our hype of politicians and political genres Squirrels. It’s just that its attention is here, so everyone hypes it up, instead of saying how big a breakthrough it is at the industrial level. I don’t think that’s the concept yet.
Another narrative we just mentioned about the next one to two years, I think there is actually a big narrative. In the first half of this year, including the second half of this year, Musk and Sam Altman both mentioned that by 2025, they think there will be AGI, that is, the so-called general artificial intelligence will be born. According to the product plan disclosed by Sam Altman today, they should also have an OpenAI AI Agent product in 2025. I think by that time point, it is very likely that the public will not be fully prepared for what the AGI product will be in 2024. Because in fact, many of us now use this GPT product more as a tool. I want to do typesetting, write an article, or make a picture. It is more of an auxiliary decision-making, and has not completely become an intelligent body very similar to humans. Everyone has not encountered such a moment. But such a moment, I think it is very likely to come in 2025 or 2026. At that moment, a person's labor value and even the value of existence will be greatly impacted. So I think when this kind of economic and commercial impact on the human level comes, everyone's attention to AI and sense of crisis will reach a new level. By that time, the spillover attention and market attention should bring great speculative value to AI in the Crypto world. So we thought that the long-term growth space of Wordcoin tokens should be quite large. Because one of the things Wordcoin wants to solve is how to distinguish between real people and so-called AI agents. Everyone thinks that this matter is not very important now. It seems that there are not so many AI agents running around. What is the demand for identification? But I think that after 2025, there will be a great sense of crisis in experience. They will feel that this value seems to have become a real value. In addition, the emergence of a large number of AI agents will reduce the cost of labor to a very low level, and many people may lose their jobs, especially many white-collar jobs. At that time, the so-called universal income that Wordcoin focuses on, giving money to everyone to ensure their basic living income, will become a point that many people may think is valuable. So I think this narrative may also be a very important point of public impact in the next one to two years.
Crypto AI projects worth watching
Alex : Let's talk about a more specific topic. In fact, many of the people who follow our channel are Crypto investors. They would like to know one question, that is, if the two guests choose one or two AI projects that you know about as the most noteworthy projects, which one would you recommend? Please tell us the reasons for your recommendation and what potential risks you think this project may face.
Lydia : The first thing that came to my mind was Bittensor. I just saw that its price was close to a new high, and it might reach a new high today. Let me talk about three aspects. Today I won’t talk about its technical architecture and specific token economy. In fact, the most profound impression this team and this project gave me was that their narrative ability was very top-notch. I think this is an aspect that has been mentioned all the time, but many people actually don’t recognize its true importance. I have watched many videos of Bittensor on YouTube, as well as their speeches in the community and on social media. The image presented by their team is particularly popular with developers, that is, very friendly, very sincere, but also very ambitious. You feel that they use a pair of innocent eyes like a little deer to tell you that I want to do a great thing, are you willing to support me? This makes it difficult for many developers to say no. And I can see that several main members of their team should be fans of Hayek. They quote Hayek’s views on the free market and neoliberalism in many videos, and in the design of their own token economy and the process of resource allocation, they will consciously borrow this somewhat experimental approach. This is very popular with investors who are interested in capital markets and neoliberalism. And they do all kinds of activities to reinforce this impression, including live broadcasts, documentaries, and various offline gatherings. This kind of team image and what they do cannot be disproven in the short term, and it is on the cusp of the trend, which leads to my feeling that Bittensor has many fans of high quality, including many well-known institutional researchers and investors who keep talking on Twitter, or experienced AI or Web3 developers who will announce their joining the Bittensor ecosystem in a high-profile manner. Every offline gathering of Bittensor will attract a wave of new fans, just like preaching, and this will also be reflected in the price. This is the first narrative ability.
The second one that may be more practical is institutional adoption. Grayscale announced the launch of a decentralized AI fund in July this year. The first batch of projects included Tao, Fil, Near, and Render. But Tao accounted for a very low proportion at the time, only about 3%. I thought it was a bit strange at the time, but soon, in less than a month, it announced a trust for Tao. Last month, Grayscale's parent company established a subsidiary, Yuma, and announced that it would focus on the development of the Bittensor ecosystem, and it was the founder and CEO of Grayscale's parent company who also served as the CEO of the subsidiary of the Tao ecosystem. I think this is an unprecedented treatment among the holdings of Grayscale and some other larger institutions. And Tao is very young, and many people don't know that it is actually a project that will issue coins in 2023, which makes its position very different.
The third point is that this project has experienced a large amount of FUD, but it has shown vitality. This is different from many AI projects that have become popular all of a sudden. I remember that in March, there were a lot of content targeting Bittensor on Twitter, whether it was its subnet, its token economy, or the FUD of the team. Its price fell all the way, from March to July and August, and the price basically fell by two-thirds, to more than 200 at the lowest. But before this wave of Pre-AI Agent, probably in September, the price basically rebounded quickly and did not continue to fall. So this process shows that this project is viable and has development ideas. If you look at what the current Bittensor subnet projects are doing, you will find that it is very different from what you saw in March. There are also researchers who have done the distribution of various subnets in the Bittensor ecosystem. In fact, it already has a prototype of an AI ecosystem. One of the reasons for paying more attention to this project is that Tao can basically make a final combination presentation of a package of tokens in its ecosystem. In addition, each of its subnets is actually a separate project. Tao's fanatical and loyal fans think that all AI projects can be included in Tao's ecosystem for unified scheduling, using Tao as an intermediary token. So it is a relatively complete ecosystem, and it has an elimination mechanism. For example, Agent is very popular recently, and more projects with more competitive Agents may join in, eliminating the projects that were weaker in token capture and token emission. So it is a process of replacing old water with new water.
Alex : Okay, Max can come and talk about it. If Max is also most concerned about Bittensor, that's fine, you can talk about your perspective.
Max : I think if we want to talk about Bittensor seriously, we can talk about it for a whole episode. Because I have also written a research article on Bittensor, and I remember Lydia has also written several articles. I am most concerned about Bittensor, but I can add a few points that Lydia did not mention just now, which are about risks. First, let me talk about why I pay attention to Bittensor. Because I just said that the most important function of Crypto for Crypto AI is the incentive mechanism, that is, you can use this incentive mechanism to make different products, and do it in a decentralized, transparent and open way. Then Bittensor is very simple. It means that I set up this combination today, and I just want to make a good incentive mechanism. As long as I do this incentive mechanism well, the function of Crypto will be good. After completing this function, I will definitely be able to achieve a successful development. Facts have proved that it is now on the right path. As long as its incentive mechanism is well done, naturally the relatively bad models will be removed from its main network, or some new and more powerful subnets or more powerful computer models will be promoted, and then get more incentives. So what it is doing now is actually to establish this incentive mechanism, and then in the next 5, 10, or 20 years, it can continue to allow the consensus mechanism of Yuma Consensus and the main main network of Bittensor to continue to develop this decentralized AI ecosystem. This is a very special point about it, and it is the first one to do this. Including its Tokenomics, which is based on the issuance of Bitcoin, the maximum number is 21 million. Just now, Lydia also mentioned that its team is very smart, and almost every episode of its YouTube video is very technical stuff, that is, if you don’t have a certain basic knowledge, you can’t understand it. It won’t talk to you about the token price all day long, it will tell you to seriously raise all kinds of problems you encounter and then solve them. I remember that it was this year or last year, their main network was hacked some time ago, and they immediately found the source and solved all of them within three or two days. This thing is also something I think is very powerful.
OK, after talking about the good things about Bittensor, I want to talk about its risks. The first point is that it is built on Bitcoin's Tokenomics, so its current distribution rate is very amazing, about 20 to 30 percent, which means that it produces a lot of tokens every day, so the value of the tokens continues to be diluted, but in the current high market sentiment, you may not feel the price changes. The second point is that although Bittensor now wants to build a decentralized ecosystem, its current mainnet is actually controlled by its own OpenTensor Foundation, which is the team behind Bittensor. So they actually have control over the mainnet, but in the future they hope to distribute the control of the mainnet to those who stake their tokens in the form of Proof of Stake, and then let them manage the community or ecosystem. But for now, although Bittensor wants to build a decentralized ecosystem, it is actually a very centralized project. These are the two major risks I want to mention. Bittensor is a very fascinating thing. Once you come in, you will find that there are a lot of things to learn. It now has 50 or 60 subnets, and each subnet does something different. Some are even doing DeepFake enhanced detection, some are doing LLM model optimization, and some are doing decentralized resources, decentralized computing power, decentralized data, and so on.
In addition to Bittensor, I think there are some other things that are worth looking forward to or paying attention to. For example, Vana is working on decentralized data. They believe that in the future, as more and more LLM data is used for training, more authentic data will become more and more valuable. So companies like Vana can use it as an incentive mechanism like the token incentive model to manage the data they produce. In the future, if other AI applications want to use this data, they will need to pay a certain fee to Varna. There are also Arweave, which is working on the architecture layer and AI computer, which is also worth looking forward to. There is also Nier, which is currently working on chain abstraction, and its Incubator Program has also supported many different AI applications. These are several projects that are worth looking forward to, and I would like to share them with you.
Evaluation strategies for Crypto AI projects
Alex : We just finished talking about specific projects, so now let's talk about some investment experience or methodology. Both of you mentioned the AI projects you have seen and the logic behind them. If we abstract these ways of thinking, when you are researching and selecting Crypto AI projects, which dimensions are you most concerned about? Or when you decide whether to invest in a project, what are the core factors? If you list three to five points, which ones might they be?
Max : When I do research and analysis, I mainly divide it into five structures. The first is the team, the group that directly builds the project, such as the founder, the VC behind the investment, the community, etc. The second is their real product, what is the product they issue. Then there are things like their profitability, future prospects and token economic model. Among these five, the team is what I value most. I think investing in cryptocurrency projects is basically investing in startups. So what is the most important thing in a startup? It is the team that builds this startup and this product. How powerful their team is determines whether this product can achieve Product Market Fit, whether it can make a profit, whether it can continue their Roadmap, and whether it can create value again. So when looking at Crypto AI projects, the first thing I will do is to find out who is the founder of the team behind them. After finding the founder, I will extend it to see who is the person who invested in this project and who is the VC. If you compare Multicoin or some well-known VC projects, they will also do some research when investing, so I may refer to the VC behind the investors. Then the community is one of them. The community that supports this project in the team is talking about the price of the currency all day, or is it really talking about the future of the project, what difficulties are encountered now, and how we should solve them. This is actually a very polarized thing. You can observe their community and see whether this project is a project that everyone wants the token to rise and it will be fine, or whether they are really planning for this project, want to participate in this community, actively discuss, solve problems, and then create better products. Are they staying for the long run, or are they just speculating in it and want to leave after the token rises. So I think the team is investing in all currency projects, not just Crypto AI, DeFi and others, etc. The most important thing. Is the team capable, who are the community and VC supporting it, what is their reputation, and from what perspective are they involved in this team.
Lydia: I agree with Max. I will first look at the team. The team mainly looks at the narrative ability and the ability to do things. Bittensor was mentioned just now. It has top-notch narrative ability. Another one is Virtuals Protocol. Its token $VIRTUAL has now entered the top 100 in market value. It has quadrupled in less than two weeks. I actually noticed this project quite early. They were originally a Southeast Asian game union. It was during this cycle that they transformed and started to do AI. The first video I saw about them was that they launched a small game that could make Mario run continuously through AI. So it made an infinite game mechanism. There is a very famous book called "Finite and Infinite Games", which I found very interesting at the time. About two months later, they launched the agent platform. So it is quite in line with my aesthetic of the team, that is, they must have ideas and execution. This attribute of the team can make their projects appear very organic and full of vitality. If I give a few other examples, I think one is Ronin and the other is Pendle. These are the top teams in Crypto that I understand. So an excellent team, no matter which direction it is in, be it games, DeFi, or AI, must have a keen ability to capture narratives, and when it is time to firmly transform, it must find the direction in which it has the most advantages and carry it through to the end.
Since it involves investment, I will definitely look at what the token means in this project. This is also an aspect that the team has dug deep into, that is, whether the team has a deep understanding of Crypto and whether they know how to leverage the attribute of Crypto to better help it, whether it is to achieve growth or to do some resource allocation work in the project mechanism itself. Bittensor exploration is more cutting-edge. Max just mentioned that its inflation may be a risk or a problem in the eyes of many investors. But if you think that it needs to support a huge ecosystem of dozens, or even more than a thousand subnets in the future, it is difficult to play a proper incentive without a large amount of daily emissions. I think this may be an exploration made by the team based on their understanding of the free market. Then Virtual is more practical. Its Virtual token is equivalent to a platform coin. If you want to participate in the speculation or investment of the AI Agent above, you must buy this token. But the common point between the two is that they have utility from Day 1. The narrative of Crypto AI can be virtual, but I think this token must be practical.
There is another point, which may be more flexible, that is, I will look at whether the brand, culture, and community of the entire project are cool. There is no unified standard for this, but I have a vague idea that if a Crypto AI project only says that I can do better than another project in a certain aspect, then it is not very cool. A cooler project should emphasize "I am very different, no one can compare with me. If you have taste or appreciate me, then you will naturally come to my community." This is probably what it means.
Alex : I understand. I would also like to add a small point of view. For research on Crypto AI or some new tracks in the Web3 field, I think I will make a cyclical judgment. Just now, Lydia mentioned the development curve of an industry. Some major innovative projects and revolutionary projects, including, for example, if we look at the track, DeFi will definitely experience a short-term over-optimism when it first comes out, and then after its bubble bursts, it will experience a long-term over-pessimism. I think we should judge the track we want to invest in based on our physical feelings. Although it has a good prospect and great potential, we should see whether it is in a short-term over-optimism stage or a long-term over-pessimism stage. If the commercial value of this track is really clear, it will definitely get out of the long-term pessimism stage. So I think that for those long-term investors, the best time to participate is when the entire market is in a stage of long-term over-pessimism about this track. For example, you may buy DeFi projects or such high-quality L1 projects in 2023. Although DeFi's current round of growth is not particularly large, and although it has grown quite well recently, if you buy this kind of project, because you deeply understand its commercial value, you can place a heavy position if you buy it and dare to hold it for a long time. You don't need to check its price every day, because its business changes will not be as big as price changes.
I think AI may also be in this mode. According to my experience, I think that since Crypto AI is in the first round, its real rise may be in 2024, and it may be in a stage of excessive short-term optimism. It may not have reached its peak yet, but we know that when the bear market comes, most projects, especially those in the first year of the first round, tend to fall by 90% or even more than 95%, just like the last round of DeFi and GameFi. But I think the development of AI may be different from GameFi. I believe that its long-term vitality will be longer than GameFi, which has more Ponzi properties. So I think from the perspective of long-term investment opportunities, after the bubble of this round of bear market bursts, most AI projects will fall by more than 95%. Of course, it does not mean that Tao may fall from 700 to 70 now. It may fall back to 2000 or 100 yuan after rising to 2000. I think that may be a good opportunity at that time.
Common AI tools sharing
Alex : Let's talk about a topic at the end. This topic may not necessarily be related to Web3. Today, our main topic is to talk about AI. I believe that both guests must use a lot of AI tools in their daily lives. What tools do you use in your daily life or work? How do you use them and what role do they play?
Lydia : The one I use the most is GPT, and there are two main uses for it. One is actually not related to anything particularly practical, it just helps me practice English. I will tell it what I want to say, can you find 10 ways to express it, I think this is very useful. Another is to be my psychological counselor, because I like chatting with GPT. When GPT first came out, many people used it as a chatbot to chat about random things, but I still do this with it now, and because I chat a lot and the frequency is very high, GPT now knows me very well. Now I just need to ask a very simple question, and it can help me, from what I said before, to guess what the main source of the problem I am facing now may be. I think it gives me a particularly good psychological comfort. The second tool is perplexity, which is mainly used for search. The search is very useful, and its various web resources, especially those in English, are very comprehensive. If I see a project that interests me but I don’t have time to read the white paper, I will first ask about the difference between Project A and Project B in Tokenomics design, or whether there is a difference in VE mechanism. If I input this question into it, it will find the answer for me. And because it will display the source webpage, if I don’t understand it very well, I will directly click in to read the webpages it cites, which will improve the efficiency of knowledge summary. The third one is that I have recommended it to Alex before. ByteDance has made a small plug-in called Doubao, which is mainly used when watching YouTube videos. It can summarize the timeline for you on the right side of the page, and you can easily click on the part you want to listen to the most. These are the three tools I often use.
Max : I am a heavy user of ChatGPT. ChatGPT is a very good tool for me. I regard it as a tool for knowledge absorption. When I read articles or listen to podcasts, I will actually listen to them to the end, because I think ChatGPT summarizes them, but I still want to hear some small details in person. But when there are some things that I think are very important but I don’t have time to listen to them, I will directly paste a PDF of maybe 20 pages and ask ChatGPT to help me summarize them. This is a great tool in terms of data acquisition and organization. In addition, because I usually shoot YouTube and write research reports, when I am doing text output, I don’t ask ChatGPT to help me revise it, because I think the articles I write should have my own style, so that AI will not feel like me after revising them. But for image output, I will ask ChatGPT to help me output the images according to various scenarios. In this way, I don’t have to find a designer to redesign the cover of the video or the cover of the research report, etc. I think these two functions are the most important. In the future, I plan to input my own investment research framework into it to see if I can have it produce research reports for me like an AI agent, and then we can discuss with each other. This is still under research.
Alex : I see. I feel that I use GPT the most. I feel that GPT is useful for two reasons. First, I have my own TG channel. Basically, I will update three to four investment memos about Web3 every week. I will talk about some news that I think may be more important this week and my thoughts on these news. Because it will be very long, it may involve more than a dozen news each time. I hope to add a small icon in front of each news title, so that the article looks richer. Every time I will let it help me sort, so this is used daily. Another point that I think is very useful is that when reading books, you often encounter some concepts that you may want to think more deeply. For example, some time ago, I was reading the autobiography of the next US Vice President JD Vance, "Hillbilly Elegy", which talked about many words related to American religion. Originally, I felt that this seemed to have little to do with me, or I was interested but it was too troublesome to check, so I ignored it and skipped it. For example, like neo-evangelicalism and secularism. Now with GPT, I can directly ask it what secularism is and what neo-evangelicalism is. It will tell you completely what these two concepts are, where they originated from, and how the concepts have changed from modern times to the present. I think it is like a very knowledgeable and patient teacher who works for you 24 hours a day. When you are learning, it can have unlimited educational resources. More importantly, it is point-to-point, and everyone's questions are different. So I think in the field of education, GPT has unlimited potential in the future. In the future, as GPT can add more virtual people, and even do it through virtual space interaction, I think teachers in large classrooms will really lose their role slowly. Maybe everyone will interact more with virtual human teachers. Its patience and its personalized teaching methods are really incomparable to current teachers. There is another one that is the same as what Lydia said, which is perplexity search. I remember that after I used perplexity, I haven't used Baidu and Google search for a long time. It can help you read a wide range of information and give you a very accurate answer with the source attached. So I have started to pay for perplexity membership now. Today I saw a piece of news. Google is the current search giant. Its CEO Pichai also said that Google's search will undergo a very big change next year and turn to AI-centric. He said that people will fully feel the impact of this change. I believe that future searches will definitely turn to AI-driven directions. So I think the subsequent use of AI tools may be the same as whether people were able to fully use computers 20 years ago, and the value of productivity improvement is the same.
Thank you to both guests for their wonderful, diverse and in-depth thoughts today. I hope we can invite you two to continue to communicate with us in new programs in the future. Thank you.
