Author: Laobai, ABCDE

After more than a year since the release of ChatGPT, the discussion about AI+Crypto has become lively in the market again. AI is considered one of the most important tracks in the bull market of 2024-2025. Even Vitalik Buterin himself published an article "The promise and challenges of crypto + AI applications" discussing the possible exploration directions of AI+Crypto in the future.

This article will not make too many subjective judgments, but will simply summarize the entrepreneurial projects that have combined AI and Crypto observed in the past year from the perspective of the primary market, to see from which angles entrepreneurs have entered the market, what achievements have been made so far, and what areas are still being explored.

I. The Cycle of AI+Crypto

Throughout 2023, we have discussed nearly dozens of AI+Crypto projects, among which clear cycles can be seen.

Before the end of 2022 when ChatGPT was released, there were few blockchain projects related to AI in the secondary market, and the main ones that come to mind were FET, AGIX, and a few other well-established projects. Similarly, there were not many AI-related projects that could be seen in the primary market.

From January to May 2023, it can be said that it was the first concentrated outbreak period for AI projects, after all, ChatGPT brought a significant impact. Many established projects in the secondary market pivoted to the AI track, and almost every week in the primary market, AI+Crypto projects could be discussed. Similarly, during this period, the AI projects gave people a relatively simple feeling, many of which were "skin-wrapped" + "blockchain transformation" projects based on ChatGPT, with almost no core technical barriers. Our in-house development team could often replicate a project framework in just one or two days. This also led to a lot of discussions about AI projects during this period, but ultimately no action was taken.

From May to October, the secondary market started to turn bearish. Interestingly, the number of AI projects in the primary market also decreased significantly during this period, until the recent one or two months when the quantity became active again, and discussions and articles about AI+Crypto in the market also became rich. We have once again entered the "boom" period where AI projects can be encountered every week. After a six-month hiatus, it is clearly felt that a new batch of AI projects has appeared, with a better understanding of the AI track, the landing of commercial scenarios, and a significant improvement in the combination of AI+Crypto compared to the first wave of AI hype. Although the technical barriers are still not strong, the overall maturity has taken a step forward. It was not until 2024 that we finally made our first bet on the AI+Crypto track.

II. The Track of AI+Crypto

In the article about the prospects and challenges, Vitalik Buterin gave predictions from several relatively abstract dimensions and perspectives:

• AI as a participant in the game
• AI as the game interface
• AI as the game rules
• AI as the game goal

We will summarize the AI projects seen in the primary market from a more specific and direct perspective. Most AI+Crypto projects are mainly centered around the core of Crypto, that is, "decentralization on the technical (or political) level + assetization on the business level."

There is not much to say about decentralization, it's all about Web3… According to the category of assetization, it can be broadly divided into three main tracks:

• Assetization of computing power
• Assetization of models
• Assetization of data

Assetization of Computing Power

This is a relatively intensive track, as in addition to various new projects, there are many old projects that have pivoted, such as Akash from Cosmos, Nosana from Solana, and after the pivot, the tokens have all surged in value, which also indirectly reflects the market's optimism about the AI track. Although RNDR is mainly focused on decentralized rendering, it can also serve AI, so many classifications also include all these computing power-related projects in the AI track.

Assetization of computing power can be further divided into two directions based on the use of computing power:

• One is represented by Gensyn, which is "using decentralized computing power for AI training";
• The other is represented by most pivots and new projects, which is "using decentralized computing power for AI inference."

In this track, an interesting phenomenon can be observed, or rather, a disdainful hierarchy:

• Traditional AI → Decentralized inference → Decentralized training
• Traditional AI professionals do not think highly of decentralized AI training or inference
• Those focused on decentralized inference do not think highly of decentralized training

The main reason is that in terms of technology, AI training (especially for large models) involves massive amounts of data, and even more exaggerated than the data requirement is the bandwidth demand formed by high-speed communication. In the current environment of large Transformer models, training these large models requires a matrix of computing power composed of a large number of 4090-level high-end graphics cards/professional AI cards and a hundred-gigabit communication channel composed of NVLink and professional fiber optic switches. Can these things be decentralized? Hmm…

The demand for computing power and communication bandwidth for AI inference is far less than that for AI training, so the possibility of decentralized implementation is naturally much greater for inference than for training. This is also why most computing power-related projects focus on inference, and there are basically only a few major players like Gensyn and Together that have raised over a billion in funding. However, from the perspectives of cost-effectiveness and reliability, at least at the current stage, centralized computing power for inference is still far superior to decentralized computing power.

This explains why those focused on decentralized inference look down on decentralized training and think "you can't make it," while traditional AI professionals think that "training is not realistic technically" and "inference is not reliable commercially."

Some say that when BTC/ETH first came out, everyone said that the distributed node model was relatively unreliable compared to cloud computing, but in the end, it worked out, right? This depends on the future demand for correctness, immutability, redundancy, and other dimensions of AI training and inference. Purely in terms of performance, reliability, and price, it is currently indeed not possible to surpass centralized methods.

Assetization of Models

This is also a track where projects are concentrated, and it is relatively easier to understand compared to the assetization of computing power, because after the popularity of ChatGPT, one of the most well-known applications is Character.AI. You can seek wisdom from sages like Socrates and Confucius, chat with celebrities like Musk and Ultraman, or even engage in romantic conversations with virtual idols like Hatsune Miku and General Raiden, all thanks to the charm of large language models. The concept of AI Agent has become deeply ingrained in people's minds through Character.AI.

What if sages, Musk, and General Raiden were all NFTs?

Isn't this AI X Crypto?!

So, rather than saying it is the assetization of models, it is more like the assetization of Agents built on large models. After all, large models themselves cannot be put on the chain. There are various types of Agents that can teach you English or engage in romantic conversations with you, and derivative projects such as Agent search and Market Place can also be seen.

The common problem in this track is that there is no technical barrier, it is basically just the NFT-ization of Character.AI. Our in-house technical genius can create an Agent that talks and sounds like BMAN using existing open-source tools and frameworks in just one night. Secondly, the degree of integration with blockchain is very light, somewhat similar to Gamefi NFTs on ETH. Essentially, what is stored in the metadata may only be a URL or hash, and the models/Agents are all on cloud servers, and the on-chain transactions are only for ownership.

The assetization of models/Agents will still be one of the main tracks of AI x Crypto in the visible future. Hopefully, we can see projects with relatively strong technical barriers and a more native integration with blockchain itself in the future.

Assetization of Data

Logically speaking, the assetization of data is the most suitable for AI+Crypto, because traditional AI training can mostly only utilize visible data on the internet, or to be more precise, data from the public domain, which may account for only 10-20% or less. More data actually lies in private domain traffic (including personal data). If this traffic data can be used for training or fine-tuning large models, we can certainly have more professional Agents/Bots in various vertical fields.

The most famous slogan of Web3 is "Read, Write, Own!"

So, through AI+Crypto, under the guidance of decentralized incentives, the release of personal and private traffic data, and its assetization, can provide better and richer "food" for large models. It sounds like a very logical approach, and indeed there are several teams deeply involved in this field.

However, the biggest challenge in this track is that data is difficult to standardize, unlike computing power. In decentralized computing, the model of your graphics card can directly be converted into computing power, while the quantity, quality, and usage of private data are difficult to measure. If decentralized computing power is like ERC20, then the assetization of decentralized AI training data is somewhat like ERC721, and it's like having many projects and traits of PunkAzuki monkeys mixed together, making liquidity and market development much more difficult than ERC20. Therefore, projects that aim to assetize AI data are currently facing significant challenges.

Another noteworthy aspect of the data track is decentralized labeling. Data assetization works at the "data collection" stage, and the collected data needs to be processed before feeding it to AI, which is where data labeling comes in. This step is currently mostly centralized and labor-intensive. Through decentralized token rewards, turning this labor work into decentralized, "Label to Earn," or a similar approach like distributing work through a crowdsourcing platform, is also a viable strategy. A few teams are currently working in this area.

III. Missing Puzzle Pieces in AI+Crypto

Let me briefly explain from our perspective the missing puzzle pieces in this track at the moment.

First is the lack of technical barriers. As mentioned earlier, the vast majority of AI+Crypto projects have almost no barriers compared to traditional AI projects in Web2. They rely more on economic models and token incentives in the front-end experience, and focus on the market and operations. This is certainly understandable, as decentralization and value distribution are the strengths of Web3. However, the lack of core barriers inevitably gives an "X to Earn" feeling. We still hope to see more teams like RNDR, with a core technology, making a big impact in the Crypto space.

Second is the current status of practitioners. Based on the current observations, some teams in the AI x Crypto track are very knowledgeable about AI but have a shallow understanding of Web3. On the other hand, some teams are very Crypto Native but have limited expertise in the AI field. This is very similar to the early days of the Gamefi track, where some were very knowledgeable about games and wanted to transform Web2 games into blockchain, while others were very knowledgeable about Web3 and wanted to innovate and optimize various gold-making models. Matr1x is the first team we encountered in the Gamefi track that has a deep understanding of both games and Crypto. This is why I previously mentioned that Matr1x is one of the three projects that I "sealed the deal" on in 2023. We hope to see more teams in 2024 that have a deep understanding of both AI and Crypto.

Third is the commercial scenarios. AI x Crypto is in an extremely early exploration stage, and the various types of projects seen in the market that combine AI and Crypto feel somewhat "awkward" or "rough." They have not fully leveraged the optimal competitiveness or combinability of AI or Crypto, which is also closely related to the second point mentioned above. For example, our in-house R&D team has thought of and designed a better way of integration, but unfortunately, after seeing so many AI track projects, we still haven't seen any teams entering this niche area, so we can only continue to wait.

What? You're asking why our VC team can think of certain scenarios before entrepreneurs in the market? Because we have 7 geniuses in our in-house AI team, 5 of whom have a background in AI with a PHD. As for the understanding of Crypto by the ABCDE team, you know…

Finally, I want to say that although from the perspective of the primary market, AI x Crypto is still very early and immature, this does not prevent us from being optimistic about 2024-2025, when AI x Crypto will become one of the main tracks of this bull market. After all, is there a better way to combine productivity-liberating AI and relationship-liberating blockchain? :)

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。