What are the tracks that are worthy of attention from Vitalik’s article?

Author:@Charlotte0211z,@Blazingkevin_, Metrics Ventures

Vitalik published on January 30>The Promise and Challenges of Crypto + AI ApplicationsOne article discusses how the blockchain and artificial intelligence should be combined in the process, and the potential challenges that appear in the process.One month after this article was released, the NMR, Near, and WLD mentioned in the article all gained a good increase and completed a round of value discovery.Based on the four methods of the combination of Crypto and AI proposed by Vitalik, this article combed the existing AI track subdivision direction, and briefly introduced representative projects in each direction.

1 Introduction: The four ways of combining Crypto and AI

Decentralization is the consensus maintained by the blockchain to ensure that security is the core idea, and open source is a key foundation for the above characteristics to have the above characteristics from the perspective of cryptography.In the past few years, this method has been applicable in several rounds of changes in the blockchain, but when artificial intelligence participates in it, some changes have taken place.

Imagine the design of the blockchain or application through artificial intelligence, then the model is necessary to open source, but in this way, it will expose its fragility in confrontation machine learning; otherwise, it will lose decentralization.Therefore, it is necessary to think about what method and depth to complete the combination when adding artificial intelligence in the current blockchain or application.

Source: de University of Ethereum

existof>When giants coloride: Exploring the convergence of crypto x AIIn one article, the differences between artificial intelligence and blockchain in the core characteristics are explained.As shown in the figure above, the characteristics of artificial intelligence are:

• Centralization

• Low transparency

• Energy consumption

• Monopoly

• Weak monetary attributes

The blockchain is completely opposite compared to artificial intelligence.This is also the true argument of Vitalik’s article. If artificial intelligence and blockchain are combined, what kind of choices should be made in data ownership, transparency, monetization capabilities, and energy consumption costs.Guarantee the effective combination of the two.

Vitalik is divided into 4 categories in accordance with the above criteria and its own thinking to divide the application of artificial intelligence with the blockchain:

• Artificial intelligence as a participant in applications (AI AS A Player in A Game)

• Artificial Intelligence as an interface for applications (AI AS An Interface to the Game)

• Artificial intelligence as the rule of application (AI as the rules of the game)

• Artificial intelligence as a target for applications (AI as the Objective of the Game)

Among them, the first three are three ways to introduce the Crypto world in AI, which represents the three levels from shallow to deep. According to the author’s understanding, this division represents the impact of AI on human decision -making, therebyCrypto introduced different degrees of systemic risk:

• Artificial intelligence as a participant in applications: Artificial intelligence itself does not affect human decision -making and behavior, so it will not bring risks to the real human world, and therefore has the highest landing nature.

• Artificial intelligence as an application interface: Artificial intelligence provides auxiliary information or auxiliary tools for human decision -making and behavior, which will increase the experience of users and developers and reduce the threshold. However, the wrong information or operation will bring certain risks to the real world.

• Artificial intelligence as the rules of application: Artificial intelligence will fully complete the decision -making and operations instead of humans. Therefore, the evil and failure of artificial intelligence will directly lead to the confusion of the real world. Whether in Web2 or web3, artificial intelligence is currently unable to trust human beings to replace humans to perform instead of human beings.decision making.

Finally, the fourth type of project is committed to creating better artificial intelligence with the characteristics of Crypto. As mentioned earlier, centralization, low transparency, energy consumption, monopoly and currency attributes can naturally go through the attributes of Crypto to Chinaand.Although many people have doubts about whether Crypto can have influence on the development of artificial intelligence, the real world has always been Crypto’s most charming narrative through decentralized power.The hottest part of the road speculation.

2 AI as a participant

In the mechanism of AI participation, the ultimate source of incentive comes from the protocols entered by humans.Before AI becomes an interface and even before becoming a rule, we often need to evaluate the performance of different AI to enable AI to participate in a mechanism, and finally get a reward or be punished through a chain mechanism.

As a participant, compared to as interfaces and rules, the risk of users and the entire system can basically be ignored. It can be said that AI has deeply affected user decision -making and behavior. ThereforeThe cost and choice required for the integration of the blockchain on this layer are relatively small, and it is also a type of product that V God believes that it is now highly available.

From the perspective of a broad sense and achievement, most of the current AI applications belong to this category, such as the TRADING BOT and Chatbot, an empowerment of AI, and the current degree of landing.Comparison and gradual optimization in BOT, encrypted users have not yet developed behavior habits using AI applications.In the article of V God, Autonomous Agent also attributes this category.

But in terms of narrow and long -term visions, we tend to divide more detailed division of AI applications or AI Agents. Therefore, under this category, we believe that the representative subdivision track includes:

2.1 AI game

To some extent, AI games can be classified as this category. Players interact with AI and train their AI characters to make AI characters more in line with personal needs.It is more combable and competitive.The game is a transitional stage of AI before entering the real world. It is also a track that is currently low -risk and is most likely to be understood by ordinary users. Lighty projects such as AI ARENA, Echelon Prime, Altered State Machine, etc.

• Ai Arena: AI Arena is a PVP fighting game that can be continuously evolved through AI to learn and train through AI. I hope that more ordinary users can contact, understand and experience AI in the form of games, and at the same time allow artificial intelligence engineers to be based on AIArena provides various AI algorithms to increase income.Each game character is an AI -enabled NFT. The core is the core of the AI ​​model, including two parts: architecture and parameters, stored on IPFS, and randomly generated when a new NFT parameter is randomly generated, which meansIt will perform random actions. Users need to improve their strategic ability by imitating the process of learning (IL). Each user trains the role and save the progress, the parameters will be updated on the IPFS.

• Altered State Machine: ASM is not an AI game, but a protocol for confirmation and trading for the AI ​​Agent to be positioned as the Yuan universe AI protocol. It is currently integrated with multiple games including FIFA and introduced AI Agent in the game and the Yuan universe.ASM uses NFT to confirm the AI ​​Agent and trade. Each agent will include three parts: Brain (the characteristics of agent), Memories, FORM (character appearance, etc.).ASM has a GYM module, including decentralized GPU cloud providers, which can provide composition support for Agent.Currently ASMs as the underlying items include AIFA (AI Football Games), Muhammed ALI (AI Boxing Games), AI League (street football games in collaboration with FIFA), Raicers (AI -driven racing games), and FLUF World’s ThingiesNFT).

• Parallel colony (prime): Echelon Prime is developing Parallel Colony, which is a game based on AI LLM. Players can interact with your AI Avatar and affect it. Avatar will have independent operations based on memory and life trajectory.Colony is currently one of the most anticipated AI games. Recently, the official has just released a white paper and announced that it has migrated to Solana, which has made Prime a new wave of rising.

2.2 Forecast market/competition

The prediction ability is the basis for AI to make future decisions and behaviors. Before the AI ​​model is used for actual prediction, the prediction competition compares the performance of the AI ​​model at a higher level.It is positive for the development of the entire Crypto × AI -by motivating the continuous development efficiency and performance, and more suitable for the model and application of the Crypto world, before AI decision -making and behavior have a deeper impact, they create better and safer to create better and saferProducts.As V God said, predicting the market is a powerful original, which can expand more other types of problems.The iconic projects in this track include: Numerai and Ocean Protocol.

• Numerai: NUMERAI is a data science competition that has been running for a long time. Data scientists predict the stock market based on historical market data (provided by Numerai) training machine learning models, and pledge models and NMR tokens for championships.NMR tokens are incentive, and the pledged token with worse models will be destroyed.As of March 7, 2024, a total of 6,433 models were pledged. The agreement provided a total of $ 75,760,979 to data scientists.Numerai is inspiring global data scientists to build new hedge funds. The currently released funds include Numerai One and Numerai Supreme.Numerai path: Market prediction competition → crowdsourcing prediction model → new hedge funds based on crowdsses models.

• Ocean Protocol: Ocean Predictoor is paying attention to predictions, and began crowdsourcing predictions that begin with cryptocurrencies trend.Players can choose to run the Predictoor Bot or Trader Bot, Predictoor Bot uses the AI ​​model to predict the price of cryptocurrencies (such as BTC/USDT) in the next time (such as five minutes), and pledge a certain number of $ Ocean. The agreement will be based onThe pledge volume is weighted to calculate the global prediction. Traders purchase prediction results and can be traded according to it. When the accuracy of the prediction results is high, traders can make a profit from it.The purchase fee for this part of the tokens and traders is a reward.On March 2nd, Ocean Predictoor announced the latest direction in the media-World-Worm Model (WWM), and began to explore predictions on real worlds such as weather and energy.

3 AI as the interface

AI can help users understand what is happening in a simple and easy -to -understand language, act as a mentor of the user in the Crypto world, and prompts the possible risks to reduce the threshold and user risks of Crypto and improve the user experience.The specific functional products are rich in functions, such as risk prompts when wallets interact, AI -driven intent trading, AI Chatbot, which can answer ordinary user Crypto, and so on.The expansion of the audience, except for ordinary users, developers, analysts, etc., almost all groups, will become AI service objects.

Let us reiterate the common points of these projects again: have not yet replaced humans to perform certain decisions and behaviors, but are using the AI ​​model to provide human beings with information and tools for humans to assist in decision -making and behavior.From this level, the risk of AI evil has begun to expose to the system -it can interfere with the final judgment of humans by providing wrong information, which has also been analyzed in detail in the article of V God.

There are many items that can be attributed to this category, including AI Chatbot, AI Smart Contract Audit, AI code writing, AI Trading Bot, etc. It can be said that most of the current AI applications are in this categoryThe primary level, representative projects include:

• Paal: Paal is currently the leading project of AI Chatbot. It can be regarded as ChatGPT trained by Crypto related knowledge. Through integrated TG and Discord, it can provide users with: token data analysis, token fundamentals, token economics analysis, and textFor other functions such as pictures, PAAL BOT can be integrated into group chat to automatically reply to some information.Paal supports customized personal bot. Users can build their own AI knowledge base and custom BOT by feeding data collection.Paal is marching towards AI Trading Bot. On February 29th, the Crypto research & amp; transaction terminal PAALX supported by its AI support., Artificial Intelligence Assistant can reduce the threshold for user use.

• Chaingpt: Chaingpt has developed a series of Crypto tools by artificial intelligence, such as Chatbot, NFT generator, news collection, smart contract generation and auditing, trading assistant, Prompt market and AI cross -chain exchange.But Chaingpt’s current direction is project incubation and Launchpad. At present, the IDO and 4 Free Giveaways of 24 projects have been completed.

• Arkham: Ultra is a special AI engine of Arkham. The use case is to match the address with the entity in the reality through the algorithm to improve the transparency of the encrypted industry.Ultra is based on the chain provided by the user and its own collection, which merges it and exports it into an extended database, which is finally presented in a chart.However, the Arkham document did not discuss the Ultra system in detail. The reason why Arkham was concerned about this round was that the founder of OpenAI Sam Altman had a personal investment of its personal investment.

• Graphlinq: Graphlinq is an automated process management solution. It aims to deploy and manage various types of automated functions without programming, such as pushing the price of bitcoin in the Coingecho to TG BOT every 5 minutes.The solution of Graphlinq is to visualize the automation process with Graph. Users can create automated tasks by dragging nodes and perform using Graphlinq Engine.Although no code is required, the process of creating Graph still has a certain threshold for ordinary users, including choosing the right template and choosing the right and connected in hundreds of logical blocks.Therefore, Graphlinq is introducing AI. Users can use dialogue artificial intelligence and natural language to complete the construction and management of automated tasks.

• ** 0x0.AI: ** 0x0 related businesses related to AI are mainly: AI smart contract audit, AI anti -RUG detection, and AI developer center.Among them, AI anti-RUG testing will detect suspicious behaviors, such as excessive taxes or remove liquidity to prevent users from being deceived. The AI ​​Developer Center uses machine learning technology to generate smart contracts to achieve NO-Code deployment contracts.However, at present, only AI smart contract audits have been initially launched, and the other two functions have not been developed.

• Zignaly: Zignaly was born in 2018, which aims to allow individual investors to choose fund managers to manage assets for themselves, similar to Copy-Trading.Zignaly is using machine learning and artificial intelligence technology to establish an indicator system for system evaluation of fund managers. The first product currently launched is Z-SCORE, but it is still primary as an artificial intelligence product.

4 AI as a game rule

This is the most exciting part -allows AI to make decisions and behaviors for humans. Your AI will directly control your wallet and make trading decisions and behaviors for you.Under this category,The author believes that it can be divided into three levels: AI applications (especially applications with independent decision -making as the vision, such as AI automated transactions BOT, AI DEFI income BOT), Autonomous Agent protocol, and ZKML/OPML.

The AI ​​application is a tool for specific decisions in a certain field of problems. They have accumulated knowledge and data in different segments, and they depend on the decision -making decision -making by AI Model tailored according to the segment.It can be noted that AI applications are attributed to two categories at the same time in this article: interfaces and rules. From the perspective of development vision, AI applications should become an agent of independent decision -making, but at present, whether it is the effectiveness of the AI ​​model and the security of integrated AI, the security of AI models is currentlyIt is impossible to meet this requirement, and even as a slightly reluctant interface. The AI ​​application is in a very early stage. The specific projects have been introduced in the previous article.

Autonomous Agent mentioned in the first category (AI as a participant). From the perspective of the long -term vision, this article attributes it as the third category.Autonomous Agent uses a lot of data and algorithms to simulate human thinking and decision -making process, and perform various tasks and interactions.This article mainly focuses on the infrastructure of agent’s communication layers, network layers. These protocols define the ownership of Agent, establishes the identity, communication standards and communication methods of Agent, connecting multiple agent applications, and can cooperate with decision -making and behavior.

ZKML/OPML: Through the method of cryptography or economics, it is guaranteed to provide a credible output after the correct model reasoning process.Safety issues are very fatal to introducing AI into smart contracts. Smart contracts rely on input to generate output and automatically perform a series of functions. Once AI has given wrong inputs, it will introduce great systemic risks to the entire Crypto system. Therefore/OPML and possible series of potential solutions are the basis for allowing AI to perform independent operations and decisions.

Finally, the three constitute the three basic layers of AI as the running rules: ZKML/OPML as the bottom infrastructure to ensure the security of the protocol; the Agent protocol establishes an Agent ecosystem to cooperate with decision -making and behavior;The specific AI Agent will continue to improve the ability in a certain field and actually make decisions and actions.

4.1 Autonomous Agent

The application of AI Agent in the Crypto world is natural. From smart contracts to TG Bots to AI Agents, the encrypted world is moving higher and lower automation and lower user thresholds.Although the smart contract is automatically executed by the unable to be tampered with, it still needs to be awakened by relying on external triggers, and it cannot run and continuously operate. TG BOTS reduces the user threshold. Users do not need to directly interact with the encryption front end. Instead, they pass through nature.The language completion of the linguistic chain, but can only complete extremely simple and specific tasks, and still cannot achieve transactions centered on user’s intention; AI AGENTS has a certain independent decision -making ability, understands the natural language of users, and found and combines other combinations to set up other other combinations.The agent and the tool on the chain complete the target specified by the user.

AI Agent is committed to greatly improving the experience of encrypted products, and the blockchain can also help the operation of AI Agent more decentralized, transparent and safe. The specific help is:

• Inspired more developers through token to provide Agent

• NFT confirmation to promote Agent -based charges and transactions

• Provide the Agent identity and registration mechanism on the chain

• Provide an inexplicable Agent activity log to trace the source and accountability of its behavior in a timely manner

The main items of this track are as follows:

• Autonolas: Autonolas supports the asset rights and combined assets of Agent and related components through the chain protocol, so that code components, agents, and services can be discovered and reused on the chain and inspire developers to obtain economic compensation.After the developer develops a complete agent or component, it will register and obtain NFT on the code on the code, representing the ownership of the code; Service Owner will jointly create a service with multiple agents and register on the chain, and attract Agent Operators to come toActual execution services, users use paid services.

• Fetch.ai: FETCH.AI has a strong team background and development experience in the AI ​​field, and is currently paying attention to the AI ​​Agent track.The agreement consists of four key layers: AI AGENTS, Agentverse, AI Engine, and FETCH Network.AI AGENTS is the core of the system, and the others are the framework and tools for the construction of Agent services.Agentverse is a software, that is, a service platform, mainly used to create and register AI Agent.The goal of AI Engine is to convert it into operably by reading the natural language input of the user, and select the most suitable AI Agent in AgentVerse to perform the task.FETCH Network is the blockchain layer of the protocol. AI Agent must register in the Almanac contract on the chain to start working with other Agent.It is worth noting that Autonola currently focuses on the agent construction of the Crypto world, introducing the agent operation under the chain to the chain; the scope of FETCH.AI includes the web2 world, such as travel booking and weather prediction.

• Delysium: Delysium is transformed from the game to the AI ​​Agent protocol, which mainly includes two layers: communication layer and blockchain layer, and the communication layer is the main trunk of Delysium.The blockchain layer has authenticated Agent and realized an unmotivated record of Agent behavior through smart contracts.Specifically, the communication layer establishes a unified communication protocol for Agent. The standardized message system is used to allow agent to communicate without obstacles through a common language. In addition, a service discovery agreement and API can be established, so that users and users and users and APIs can be established.Other agents can quickly discover and connect available agents.The blockchain layer mainly includes two parts: Agent ID and Chronicle smart contracts. Agent ID ensures that only legal Agent can access the network. Chronicle is a log repository for all important decisions and behaviors made by Agent.Make sure the trusted traceability of Agent behavior.

• Altered State Machine: The standards of assets and transactions of Agent through NFT have set standards. Specific analysis shows part 1. Although ASM is currently mainly connected to games, it may also be expanded to other Agent fields as basic specifications.

• Morpheous: It is building an AI Agent ecological network. The agreement aims to connect Coder, Computer Provider, Community Builder, and Capital to provide the network with AI agent, the computing power of supporting Agent, the front end and development tools and funds.The form of FAIR LAUNCH provides incentives to miners, STETH pledgers, Agent or smart contract development contributors, and community developers.

4.2 ZKML/OPML

Zero knowledge proves that there are currently two main application directions:

• It proves on the chain with lower costs to obtain the correct operation (ZK-Rollup and ZKP cross-chain bridge are using the characteristics of ZK);

• Privacy Protection: You don’t need to know the details of the calculation, you can also prove that the calculation has been implemented correctly.

Similarly, the application of ZKP in machine learning can also be divided into two categories:

• Reasoning verification: That is, through ZK-Proof, the process of dense computing this dense computing of AI model reasoning on the chain is correctly executed under the chain.

• Privacy protection: It can be divided into two categories. One is the protection of data privacy, that is, using privacy data on public models for reasoning, you can use ZKML to protect privacy data; the second is the protection of model privacy, hoping to hide the modelSpecific information such as weights and other information, calculate the output results from the public input.

The author believes that the current more important thing about Crypto is reasoning verification, and we will further explain the scene of reasoning verification.Starting from AI as a participant, to the rules of AI as the world, we hope to turn AI into part of the process on the chain, but the calculation cost of the AI ​​model reasoning is too high to operate directly on the chain. Put this process under the chain.It means that we need to endure the trust issues brought by this black box -a AI model operator tampered with my input?Do you use the model I specified to reason?By converting the ML model into a ZK circuit, it can be achieved: (1) smaller models on the chain, store the small ZKML model into smart contracts, and directly go to the chain to solve opaque problems;At the same time, the ZK proves that the correctness of the reasoning process can be proved by running ZK on the chain. The infrastructure will include two contracts-the main contract (the output result of the ML model) and the ZK-PROOF verification contract.

ZKML is also in a very early stage, facing the technical problems of the ML model transformation to the ZK circuit, as well as the cost of extremely high computing and password science.Like the development path of Rollup, OPML has become another solution from the perspective of economics. OPML uses Arbitrum’s Anytrust assumption that at least one honest node is advocated to ensure that the submitter or at least one verifier is honestof.However, OPML can only be an alternative to reasoning verification and cannot achieve privacy protection.

The current projects are constructing ZKML infrastructure, and they are working hard to explore its applications. The establishment of the application is equally important. Because it is necessary to clearly prove to the encrypted user’s important role in ZKML, it proves that the final value can offset huge costs.In these projects, some focus on ZK technology research and development related to machine learning (such as Modulus Labs), and some are more common ZK infrastructure construction. Related projects include:

• Modulus is using ZKML to apply artificial intelligence to the inference process on the chain.Modulus launched the ZKML proofer Remainder on February 27. Compared with the traditional AI reasoning on the same hardware, it achieved 180 times the efficiency improvement.In addition, Modulus cooperates with multiple projects to explore the actual cases of ZKML, such as cooperating with UPSHOT, collect complex market data, evaluate NFT prices by using artificial intelligence proved by ZK, and pass the price to the chain;Cooperation proves that Avatar, who is fighting, is trained by players.

• Risc Zero puts the model on the chain. By running a machine learning model in RISC ZERO’s ZKVM, it can prove that the exact calculation involved in the model is executed correctly.

• Ingonyama is developing hardware dedicated to ZK technology, which may reduce the threshold to enter the field of ZK technology, and ZKML may also be used for model training.

5 AI as the target

If the previous three categories focus on how AI empower Crypto, then “AI as the goal” emphasizes Crypto’s help to AI, that is, how to use Crypto to create better AI models and products, which may include multiple judgment standards: More efficient, more accurate, more decentralized and so on.

AI includes three cores: data, computing power and algorithms. In each dimension, Crypto is committed to providing AI with more effective help:

• Data: Data is the basis for model training. The decentralized data protocol will inspire individuals or enterprises to provide more private domain data, and at the same time use cryptographic to ensure data privacy and avoid leakage of personal sensitive data.

• Computing power: The decentralized computing track is currently the hottest AI track. It is agreed that through the provision of the matching market of supply and demand, it promotes the matching of long -tail computing power and AI companies, and is used for model training and reasoning.

• Algorithm: Crypto’s empowerment of the algorithm is the core link to realize decentralization AI. It is also the main content of “AI as the goal” in the V Section.The problem of confrontational machine learning will be solved, but a series of obstacles will face extremely high cryptographic expenses.In addition, “use encryption incentives to encourage and make better AI” can also be achieved without being completely encrypted into the rabbit hole that is not completely encrypted.

The monopoly of large technology companies in data and computing power has jointly caused monopoly for model training process. Closed sources models have become the key to profitability of large enterprises.From the perspective of infrastructure, Crypto inspires the decentralized supply of data and computing power through economic means, and at the same time to ensure data privacy in the process through cryptographic methods, and use this as a basis to help decentralized model training to achieve realizationMore transparent and decentralized AI.

5.1 Decentralization data protocol

The decentralized data protocol is mainly carried out in the form of data crowdsourcing, which inspires users to provide data sets or data services (such as data labels) for enterprises for model training, and open the Data Marketplace to promote the matching of both parties to supply and demand. Some protocols are also exploring and passing.DEPIN incentive protocol, obtain user browsing data, or use user equipment/bandwidth to complete network data crawling.

• Ocean Protocol: Data confirmation and toblyize, users can complete the NFT creation of data/algorithms in Ocean Protocol through unclean way. Colleagues create corresponding DATATOKEN to control access to data NFT.Ocean Protocol uses Compute to Data (C2D) to ensure the privacy of the data. Users can only get the output results of data/algorithms and cannot be completely downloaded.Ocean Protocol was established in 2017. As a data market, the AI ​​express is naturally on this round of upsurge.

• Synesis ONE: This project is the Train2earn platform on Solana. The user obtains the $ SNS reward by providing data and data labeling of natural language. Users can support mining by providing data. The data will be stored and chain after verification.Come to train and reason.Specifically, miningers are divided into three categories: Architect/Builder/Validator, Architect is responsible for creating new data tasks, Builder provides corpus in the corresponding data task, and Validator verifies the dataset provided by Builder.The completed data assembly is stored in IPFS, and the data source and IPFS addresss will be stored on the chain and the IPFS address will be stored in the database under the chain for AI (currently AI).

• Grass: The decentralized data layer, known as AI, is essentially a decentralized network capture market, and obtains data for AI model training.Internet websites are an important source of AI training data, and many websites, including Twitter, Google, and Reddit, have important value, but these websites are constantly restricting data crawling.GRASS uses unused bandwidths in the personal network to reduce the impact of data blockade by using different IP addresses to capture data in public websites and complete data to clean up data.Grass is currently in the beta testing stage, and users can provide bandwidth acquisition points to receive potential airdrops.

• AIT Protocol: AIT Protocol is a decentralized data label agreement, which aims to provide developers with high -quality data sets for model training.Web3 enables global labor to quickly access the network and obtains incentives through data marking. AIT data scientists will notice the data, and then the user will be further processed. After data scientists checkFor those.

In addition to providing and data labeling protocols, the former decentralized storage infrastructure, such asFilecoin, arweaveWaiting will also help more decentralized data supply.

5.2 Decentralization computing power

In the AI ​​era, the importance of computing power is self -evident. Not only is Nvidia’s stock price climbing the peak day, but in the Crypto world, decentralized computing power can be said to be the most enthusiastic direction of the AI ​​track hype -Among the 11 AI projects, there are 5 projects for decentralized computing power.In the small market value projects, many decentralized computing platforms have also appeared. Although they have just started, with the wave of Nvidia Conference, as long as they are sideways with the GPU, they quickly gain a wave of rising.

From the characteristics of the track, the basic logic of the project is highly homogeneous in the direction of the project -through token incentives to provide resources or enterprises with idle computing resources, and thus significantly reduce the cost of use, establish an calculated supply and demand supply and demand supply and demand supply and demandAt present, the main computing power supply comes from data centers, miners (especially after Ethereum to POS), consumer computing power, and cooperation with other projects.Although homogeneity, this is a track that has a high care of the city. The main competitive advantage of the project comes from: computing power resources, computing power lease prices, computing power usage and other technical advantages.The leading projects of this track include Akash, Render, IO.NET, and Gensyn.

According to the specific business direction, the project can be rough into two categories: AI model reasoning and AI model training.Because the requirements for computing power and bandwidth in AI model training are much higher than reasoning, it is more difficult to propose to distributed reasoning, and the market for model reasoning is rapidly expanding. The predictable income will be significantly higher than the model training in the future. ThereforeMost projects are mainly attacked (Akash, Render, IO.NET), and the leading leader of the main training direction is Gensyn.Among them, Akash and Render were born earlier, not for AI calculations. Akash was originally used for general computing, render was mainly used in video and picture rendering. IO.NET was specially designed for AI computing, but in AI, the computing power will be computing power.After the demand has improved one level, these projects have tended to develop AI.

The two most important competition indicators still come from the supply side (computing power resources) and demand side (computing power).Akash has 282 GPUs and more than 20,000 CPUs. It has completed 160,000 leases. The utilization rate of the GPU network is 50-70%. This track is a good number.IO.NET has 40,272 GPUs and 5958 CPUs, and 4318 GPUs and 159 CPUs and FileCoin’s 1024 GPUs of RENDER have permits, including about 200 H100 and thousands of A100. The reasoning has been completed 151,879Second, IO.NET is expected to attract computing resources with extremely high airdrops. The GPU’s data is growing rapidly. It is necessary to re -evaluate the ability to attract resources after the tokens are launched.Render and Gensyn did not publish specific data.In addition, many projects are increasing their competitiveness on supply and demand through ecological cooperation. For example, IO.NET uses the computing power of Render and Filecoin to improve its resource reserves. Render has established a computing client plan (RNP-004).Allow users to indirectly access the computing power resources of Render by calculating the client -IO.NET, Nosana, Fedml, and Beam, so as to quickly transition from the field of rendering to artificial intelligence computing.

In addition, the verification of decentralized calculation is still a problem -how to prove that workers with computing power resources have correctly executed the computing task.Gensyn is trying to establish such a verification layer to ensure the correctness of the calculation through probability learning certification, accurate positioning protocol of graph -based, and incentives.It provides computing power support, and the verification mechanism it has also has unique value.FLUENCE, located on the SLANA, also adds verification of computing tasks. Developers can verify whether its application is running and calculating whether the application is operated and calculated correctly by checking the certification issued by the provider on the inspection chain.However, the actual demand is still “feasible” greater than “credible”. The calculation platform must first have enough computing power to compete. Of course, for excellent verification agreements, you can choose to access the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms and become become the computing power of other platforms.The verification layer and the protocol layer to play a unique role.

5.3 Decentralization model

It is still very far away from the ultimate scene described by Vitalik (shown in the figure below). At present, we cannot realize the creation of a trusted black box AI through blockchain and encryption technology to solve the problem of confrontation machine learning. Data training will be trained in data training.The encryption process of the entire AI operation of the query output is a very large expense.However, there are projects that are currently trying to create a better AI model through the incentive mechanism. First of all, the closed state of different models has been opened, creating a pattern of learning, collaboration and benign competition between models.project.

• Bittersor: Bittersor is promoting the combination between different AI models, but it is worth noting that Bittersor itself does not conduct model training, but mainly provides AI reasoning services.Bittersor’s 32 sub -nets focus on different service directions, such as data capture, text generation, text2image, etc. When completing a task, the AI ​​models in different directions can cooperate with each other.The incentive mechanism promotes the competition between the subnets and the inside of the subnet. At present, the award is issued at the speed of 1 TAO. The total daily distribution of about 7,200 TAO tokens, SN0 (root network) 64 verification deviceAccording to the performance of the subnet, the distribution ratio of these rewards between different subnets, and the subnet verification device decides the distribution ratio between different miners through the work evaluation of the miners.Better models have obtained more incentives and promoted the improvement of the overall reasoning quality of the system.

6 Conclusion: MEME speculation or technological revolution?

From SAM Altman’s move to the price of ARKM and WLD, to the Nvidia Conference to bring a series of participation projects, many people are adjusting the investment concept of the AI ​​track. Is the AI ​​track speculation or the technological revolution?

Except for a few celebrities (such as ARKM and WLD), the AI ​​track is more like “MEME” dominated by technical narrative.

On the one hand, the overall speculation of the Crypto AI track must be closely linked to the progress of Web2 AI. The external speculation headed by Openai will become the fuse of the Crypto AI track.On the other hand, the story of the AI ​​track is still mainly technical narrative. Of course, here we emphasize that “technical narrative” instead of “technology”, which makes the selection of the AI ​​track subdivision and the fund’s fundamental aspect of the project fundamentalsFollowing is still important. We need to find a narrative direction with hype value, and we also need to find projects with medium and long -term competitiveness and moat.

From the possibility of the four categories proposed by God V, it can be seen that the charm of narrative and the possibility of landing.In the first and second categories represented by AI applications, we have seen many GPT WRAPPER.Stories that can be told in homogeneous competition.The third and fourth categories represent the grand narrative combined with AI and Crypto, such as the cooperation network, ZKML, and decentralized reshaping AI on the Agent chain.It’s just a very early landing display.