AI+Web3 may become a breakthrough point for future industry integration innovation. During the research report "Analysis of the Current Situation, Competitive Landscape and Future Opportunities of the Integration of AI and Web3 Data Industry" jointly compiled by Future3 Campus and Footprint Analytics, Future3 Campus also conducted interviews with representatives from a number of well-known investment institutions. This article is compiled based on the interview content and represents only the views of the interviewed guests and does not represent Footprint’s position.
How do you view the combination of AI+Web3 data? What are the directions worth paying attention to?
Hashkey Capital-Harper: I think there are several points in the collection of AI and web3 data: First, the language is converted into SQL through the LLM model, such as Dune. There are also some projects that specialize in search engines. SQL must be strengthened to correctly access the database. The data in it can also automatically generate SQL using natural language, allowing developers to copy it and use it. The second is the chat dialogue method. The chat agent based on the transformation of chatgpt is mainly aimed at providing chat windows. It does not place so much emphasis on SQL and search optimization. It is more casual. For example, please tell me which KOL has called for orders. This security incident has a negative impact on the token. What is the impact (at this time, it may be a result of the entire network search, and the SQL optimization of the database will not be emphasized here). The third is to use AI to create appropriate models to organize off-chain and on-chain data to extract better insights.
In contrast, firstly, the project party needs to have stronger database construction capabilities, because Web3 data is very troublesome to process, but it is still difficult to achieve accuracy and speed; secondly, it is a relatively simple combination method, and the threshold is also Not too high.
SevenX Ventures-Yuxing: In fact, data is the nourishment of AI. Web3's data is open and verifiable, while the problem with AI is its black-box nature, which makes it difficult to verify. The combination of the two can produce some interesting chemical reactions. At present, I prefer to divide the combination of AI and Web3 into two categories. It is not simply AI plus Web3 data, but considering how AI can make Web3 better, and how Web3 can make AI better.
First of all, AI can effectively utilize the public and verifiable characteristics of Web3 data for Web3 data. Any AI can use Web3 data to refine and generate value. Whether it is investment advice or early warning analysis, AI can help Web3 data processing and analysis increase efficiency. On the other hand, Web3 can increase the credibility of AI because Web3 itself is a new trust mechanism. Through the data disclosure and verifiable characteristics of Web3, the transparency of AI can be improved. Even in important fields such as news reports or documentaries, key information can be stored in Web3, which can avoid some problems of AI.
The more common among these problems is the problem of AI fraud and the problem of AI black box. Some AI algorithms may be easier to understand, but some algorithms are difficult to explain, such as complex algorithms like neural networks, and GPT. People may question how his answers are generated, and its data and algorithms are not transparent enough. , it feels like magic. For example, previous facial recognition algorithms misidentified black people as gorillas because there were too few images of black people in their data samples.
If the data used by the AI model is verifiable, we can more easily identify whether the data has sample bias. Using Web3 data, the training sources and results of the entire AI model will be clearer because of its transparency. In this way, we can view AI more fairly, understand the source of its decisions, and reduce biases and errors.
The black box problem can be roughly divided into two parts. Part of it is the black box of the model algorithm itself, including how the model is trained and how the content is generated. Both the training process and the algorithm mechanism are opaque or unexplainable. The other part is the black box of the data, which does not disclose the data. Problems with the training set will also lead to deviations in the final results.
If this deviation is a problem of content accuracy, we can continue to improve it, but if it is an ideological problem, especially a political or racial discrimination problem, it may not be easy to correct. At this time, we can only control the data output. For example, the most important thing about the AI models of many national systems or state-owned enterprise systems is to control their output. What cannot be said is the most difficult thing to do. This must be The degree is similar to the ideological deviation just mentioned.
Qiming Venture Partners-Tang Yi: Regarding the combination of AI and Web3 data, I personally think that AI may be a bit hyped in this field, and the gimmicks outweigh the actual effectiveness. Because from my point of view, Crypto's data products are still in a relatively early stage, and the basic work on data is not solid enough. In this case, introducing AI or too much data analysis too early may be premature.
In addition, from a user perspective, most scenarios where encryption projects are combined with AI are not very valid, or AI is not used very much. Because this wave of popular AI models, especially generative models, are based on large-scale Internet data, such as language processing and image generation capabilities. Although some people use generative AI to improve user experience and provide a better sense of interaction and dialogue, this may be of limited value for most scenarios. I think if we talk about broader AI (data analysis capabilities or simpler AI models), there may be some scenarios, such as price estimation for NFT based on data, or a professional trading team can use data to perform some trading operations. Overall, for the current wave of AI, I have not yet seen any opportunities that can bring special short-term benefits to the cryptocurrency industry.
Of course, I have also seen some early projects that are trying to improve data processing or analysis capabilities through AI. For example, I see early projects using AI capabilities to interpret the logic of smart contracts or perform classification and recognition tasks. These jobs require high accuracy in the field of smart contracts and cryptocurrencies as they involve critical actions such as transactions. So I can imagine that it might make sense to use some AI capabilities for data preprocessing, but ultimately human intervention might still be needed to ensure accuracy. If you want to trigger trades directly through AI capabilities, in addition to professional traders, I think a lot of progress needs to be made on the product side.
Matrix Partners-Zixi: We have observed many data projects related to Web3. For example, we invested in Footprint. I was also a loyal user of it at first, as well as Dune. I think Footprint and Dune are mainly services for VCs, developers and some small businesses. The real ordinary ones have little connection with these services.
In addition, we also looked at some data analysis companies directly related to cryptocurrency trading or profit, such as Nanson, defilama, token terminal, dappradar, and of course Dune and Footprint. These companies are great for VCs and developers, but their profitability appears to be limited. The reason is that the current overall demand for this data by VCs and developers is not large enough, and their willingness to pay is not strong, because even if some services are not free, there are always other companies providing similar free services.
We also looked at some companies similar to data cloud warehouses. We also led an investment in Chainbase with Tencent. They are actually like a data platform. They provide security, transaction, NFT, DeFi, game, social data, and some comprehensive data. Developers can combine this data on these platforms to generate the APIs they need.
In the bear market, we noticed that for companies like Chainbase, Blocksec, and Footprint, many of their customers were actually small and medium-sized startups. For example, Chainbase's revenue from some of its large customers did not decline, but the revenue from small and medium-sized customers dropped to zero after two or three months. This suggests that these projects cannot continue due to lack of funding.
Therefore, it will be difficult for data providers to make money without new developers joining in the bear market. This also reflects that currently in the Web3 field, data providers mainly rely on developers and small businesses who find the data useful. They integrate the data internally and then monetize it to balance income and output.
At the core, we still feel that the current profit model of Web3 data companies in both ToC and ToB is not very clear, which results in data providers not having a strong and stable cash flow. Especially for small and medium-sized entrepreneurs, we think this is the biggest drawback of the current Web3 data industry.
Then return to the topic of combining AI and Web data. We have also recently looked at and invested in some AI-related data companies. I think AI data companies actually face the same problem, which is the sale of data. You need to consider the balance between the cost to the client and the effectiveness of their output. At present, I am relatively optimistic about the profit prospects of AI data companies, but this is mainly limited to overseas markets.
If we only focus on the domestic market, I worry that the final result may be the same as investing in a Web2 SaaS company. There may be income, but the business scale will not be too large, and customers' willingness to pay is not very strong. You may also need to provide customized services, so your gross profit margin will not be very high. Therefore, I am relatively pessimistic about doing this domestically, and relatively optimistic about doing this overseas.
What value do you think AI can bring to Web3 data infrastructure and Web3 data companies? What is the effect of the current project that uses AI to help Web3 data? Is there any innovation in the business model?
SevenX Ventures: I think the biggest help AI can bring to Web3 data is efficiency. For example, Dune has released a large AI model tool for code anomaly detection and information indexing. Users can use natural language to query the corresponding data, and its code will be generated accordingly, and then they can also optimize the code. This It is an improvement in efficiency.
There is also a project that uses AI for safety warnings. It is an AI Robot that can quickly identify safety issues after the AI has been trained accordingly. For example, there is an algorithm in the AI algorithm called anomaly detection. The effect is better than directly looking at the distribution of data and detecting an outlier through pure mathematical statistics. Therefore, this kind of AI can do security monitoring more effectively. .
In addition, I have seen some projects using AI algorithms, such as large language models, to retrieve the entire Web3 news data (not just on-chain data), perform information aggregation and public opinion analysis, and form an AI Agent. For example, users can directly check the Internet public opinion of a certain token in the last 30 days or 90 days in the dialog box. Whether the user is more inclined to be bullish or bearish, a corresponding score will be given to reflect the popularity; it will also have a Curve, through this curve we can judge whether a token is at the moment when everyone is discussing it reaches its peak, is it at a moment when its peak is falling, or is it at a moment when it is rising? These can assist users in investing, and I think it is also a very interesting application method.
But there are also some other projects that claim that their data is the data source of AI and use the concept of AI. I think this is a bit far-fetched, because any data on the chain can be the data source of AI. Because it is public, it is a bit suspected of being a hot spot. .
Matrix Partners-Zixi: Business model is a big problem in the data field now, and it is difficult to find a solution. Maybe on the ToC side, using some concepts of Web3, such as token or distributed concepts, AI data can adopt different business models. But if it’s AI technology empowering data, there aren’t many bright spots at the moment.
AI may assist in data processing and cleaning, but this is more of an internal help, such as improving functionality or user experience during product development. But from a business perspective, not much has changed.
AI bot can indeed increase some competitiveness and assist users, but currently this is not a big advantage. The core competitiveness still depends on the quality of the data source. If the data source is sufficient, I can get the information I need. The problem is that if this data is to be commercialized, then what I put together must be monetizable before I am willing to pay for the data. The problem now is that the market is not good, startups don’t know how to monetize data, and there are not enough new startups entering the market.
I think what's interesting right now is some Web2 companies that use Web3 technology. For example, a synthetic data company generates synthetic data for use through large models. The data can be mainly used in software testing, data analysis, and AI large model training. They involve many privacy deployment issues when processing data. Using the Oasis blockchain can effectively avoid data privacy issues. Later, they also want to build a data exchange to package the synthesized data in NFT for buying and selling to solve the issues of rights confirmation and privacy. I think this is a good idea. It uses Web3 technology to assist Web2 in solving problems and is not necessarily limited to Web3 companies. However, the market for synthetic data is not yet large enough, and investing in such companies at an early stage is risky. If the downstream market cannot thrive, or there are too many competitors, the situation will also be embarrassing.
In the field of AI+Web3 data, have you invested in some better projects? What are their respective directions? What are the key factors in deciding to invest in them? What do you think is the core competitiveness of this type of project? Will AI strengthen this competitiveness?
Hashkey Capital-Harper: We invested in data projects relatively early. We basically invested in them when there was no special emphasis on AI, such as space and time, 0xscope, mind network, zettablock, etc. The key to investing is to look at their positioning and data quality. . Now these projects will all have AI plans, and they basically start with chat agents. space and time and chainML have cooperated to launch the infrastructure for creating AI agents. The defi agent created is used in space and time, which is also a way to combine AI.
SevenX Ventures-Yuxing: If the project is well integrated with AI, then I might be more interested in it. One of the key factors that determines whether I will invest is whether there are market barriers to the project. I have observed many projects claiming that their combination with AI can improve efficiency, such as fast data query functions. Some projects can quickly obtain on-chain NFT data through natural language queries, such as querying the top ten NFTs with the most active transactions recently. Such a project may have a first-mover advantage, but market barriers may not be strong.
The real barrier is the application of AI itself and how engineers apply AI to specific scenarios. Engineers who are skilled at fine-tuning their models can often achieve good results. For those projects that improve efficiency, the market barrier mainly lies in the data source. Not only the data on the chain, but also how the project party processes and parses this data. For example, the projects mentioned before can quickly retrieve important data through AI algorithms. However, the effect of engineers fine-tuning the model is limited, and the real sustained advantage lies in the quality of the data source and its ability to continuously optimize. This is also the reason why some data analysis companies can stand out in the market. They not only provide data sources, but also include data processing and analysis capabilities. The difference often lies in the technical capabilities and talents of the team. These factors are directly related to the final effect of AI combined application.
In addition, I also pay attention to Web3 technology projects that can make AI better, because the AI market is very huge. If Web3 technology can enhance the capabilities of AI, the application scenarios will be very wide. This is why the ZKML project is so popular. However, I've noticed that Web3 projects tend to have their value overstated or understated. Although projects like ZKML have attracted much attention, their return on investment is not as fast as people expect, and the exit mechanism is not clear because it is difficult for them to issue tokens. So while these projects are innovative and potentially valuable, investors need to carefully consider whether they are worth investing in now and how much return they will ultimately bring.
Matrix Partners-Zixi: We have invested in a company that combines AI and Web3. It is a data annotation company called Questlab. They use blockchain technology to provide crowdsourcing services for data annotation. Data annotation was originally a direct operation or subcontracted industry, and it was difficult to achieve full coverage of the knowledge field.
As far as traditional data annotation is concerned, it is generally divided into three types: direct operation, subcontracting and crowdsourcing. But in fact, there are relatively few people doing crowdsourcing. The factors that companies in these three models need to consider when choosing data annotation services are: whether the price is cheap, whether the annotation quality is high, and how efficient it is. Another is whether they can cover the industry they are in. If you just do some general model language or image annotation, it is actually very simple, just to recognize English words or images. No matter how difficult it is, for example, if you need to distinguish cats, dogs, moons, strollers, etc., this is not difficult. But if what you need to do is more professional annotation, such as the annotation required by the voice robot community, it is much more complicated. They may need to label various dialects and multiple languages, including Chinese dialects, English dialects, and languages from various niche areas. Few traditional studios are willing to do such work.
A more complex example is a legal plus AI company, which needs to label a large amount of legal knowledge to train various models. It is very difficult to find people who understand both law and professional labeling. You need to understand the laws of various countries at the same time, and understand various professions. Legal fields, such as contract law, lease law, civil law, criminal law, etc. There is almost no data annotation company on the market that can provide such professional services. Law is professional, as are finance, biology, medical care, education, etc. Therefore, annotation work in these fields can generally only be completed by internal teams, who use crowdsourcing methods to solve the problem of professional coverage of knowledge.
We believe that using blockchain for crowdsourcing is a good direction, just like what YGG is doing in the Gamefi field. This is what we think is a promising direction.
In addition, we feel that there will be some good opportunities in the open source model community. For example, a project invested by Polychain is a hugging face similar to web3, which is used to solve the economic problem of model content creators.
Regarding the combination of other AI and Web3, I think if the ToC direction can be combined with some token gameplay to improve the stickiness, daily activity and emotion of the entire community, we think this is feasible. This also makes it easier for investors to cash out, but the market size is not yet certain. These are some of my views on AI and Web3. I think if it is a pure ToB business, there is no need to use Web3, and it is good to use Web2.
Qiming Venture Partners-Tang Yi: Currently, some of the data projects we have invested in are working in security scenarios through on-chain data. I think some basic AI pattern recognition or feature discovery work is involved, and the results are okay. However, more advanced work, such as inputting large amounts of activity data into the model and identifying multiple types of information, is still in the trial stage, and the effect needs to be verified. In addition to the security field, similar situations exist in many other fields.
A recent example is our investment in NFTGo, which is based on big data analysis to price NFT, has a certain degree of accuracy, and plans to use it for price Oracle and other purposes. Although this system sounds interesting, it still needs to be validated in terms of product and user acceptance. Because even though it may currently be possible to achieve an accuracy of 90 points or 85 points, users may require a higher level, such as 98 points or 95 points, so further verification is needed. Therefore, although some projects are applying simple AI capabilities such as data analysis and pattern recognition to products, whether it will become a key factor has not yet been verified.
As for investment willingness, I personally would not be more inclined to invest just because the project has some AI gimmicks, because I think the actual effect and whether the project can achieve its goals and bring benefits are more important. I can understand if a project just has a splash in the name or marketing as a marketing ploy to attract more attention or exposure. But in investment decisions, I think what's more important is the actual results.
For example, some projects are working on ZKML. This track seems to have attracted much attention, but at the same time, there is also a big question, which is what scenario it is used for. I think the uncertainty is particularly strong at the moment, and it’s more about a grand narrative.
From the perspective of overall industry development, what are the potential opportunities or development directions for the AI + Web3 data track in the future? In the future, is it possible for AI to completely upgrade data products and introduce new concepts? Will it increase users’ willingness to pay?
Hashkey Capital-Harper: There are definitely potential opportunities. In fact, the future development direction still lags behind web2 AI, where creativity is obviously stronger. The AI on web3 is likely to be a mapping implementation of web2 AI.
Matrix Partners-Zixi: I think the recent Miaoya Camera has made everyone realize that people are actually willing to pay for AI products. This is unlike traditional SaaS products or games, where people expect to use them for free. Users’ willingness to pay for AI is actually quite strong.
I can provide some ideas in the future. There is a key step in our data labeling process called pre-labeling, which means we train a model and let the model perform primary labeling. This step is very valuable and can save a lot of labor costs. We put the raw data into a pre-trained model for pre-annotation, then perform semi-automated data processing, and finally manually perform precise annotation. Pre-annotation can significantly improve efficiency. A job that may have originally required 100 people may now only require 50 to 70 people.
In addition, pre-labeling also involves the collaboration between AI and humans. Through your feedback, the model’s pre-labeling capabilities can be continuously improved, thus reducing the number of people required for the data labeling team. As AI and people collaborate better and better, a team of 100 people may only need 30 people. However, this process has a lower limit. Even if AI collaboration is done very well, a certain amount of manual work is still required for final annotation and review.
In other fields, since I am not a data scientist, I have not personally cleaned the data or used the data to perform SQL queries, so I don't know how much help AI can provide in these fields.
Qiming Venture Partners-Tang Yi: I think there should be some intersection with Web3 and AI in the long term. For example, from an ideological perspective, the value system of Web3 can be combined with AI, and is very suitable as a bot's account system or value conversion system. Imagine a robot that has its own account and can earn money from its smart parts, as well as pay to maintain its underlying computing power, etc. These concepts are a bit science fiction, and practical applications may still be a long way off.
The second possible direction is to verify whether the output of the AI model is based on a specific category or a specific model, or specific data, and whether it is trustworthy. These areas may have some use in trustworthy AI models. These are very interesting from a technical perspective, but whether there is sufficient market demand is uncertain.
On the other hand, the emergence of AI has made data content generation widespread and cheap. For content such as digital works, it is difficult to determine the quality and creator. In this regard, the verification of data content may require a completely new system, including the roles of creators and agents. But overall, these issues may still need to be resolved, and the narrative content may take longer to develop. In the short term, we should continue to focus on the quality of the underlying data and expect the models to become more powerful.
In addition, in terms of commercialization, it is indeed very difficult to commercialize data products. But I think from a business perspective, AI may not be the solution to the problem of commercializing data products in the short term. Commercialization requires more productization efforts, not just data capabilities. Therefore, these projects may require the development of additional products for commercialization.
——————
Footprint Analytics is a blockchain data solutions provider. With the help of cutting-edge artificial intelligence technology, we provide the first code-free data analysis platform and unified data API in the Crypto field, allowing users to quickly retrieve NFT, GameFi and wallet address fund flow tracking data of more than 30 public chain ecosystems.
Footprint 官网：https://www.footprint.network
Discord：https://discord.gg/3HYaR6USM7
Twitter：https://twitter.com/Footprint_Data
Telegram：https://t.me/Footprint_Analytics