Some time ago, due to the release of OpenAI's Sora and the release of NVIDIA's financial report, NVIDIA's market value approached US$2 trillion, which made the AI track in the Crypto industry soaring in popularity, with WLD, AGIX, FET and other projects All performed well. In the past few days, due to the overall improvement of the market and the NVIDIA AI Conference, the AI track has once again exploded.
At present, there are more than hundreds of AI+Web3 concept projects, covering multiple directions such as infrastructure, data, and computing power, and top institutions such as a16z and Binance Related tracks have already begun to be laid out. AI + Web3, as a new narrative, is destined to run through this bull market. Today, Baihua will take you to take a brief inventory of the AI track.
01 A similar trajectory between AI + Web3 and the Metaverse
Although new narratives have been emerging in the Crypto industry , and each bull market will be different, but there are some similarities in the development process, such as the AI track and the Metaverse track, which are both products of extending from Web2 to Web3.
The Metaverse first became popular because the first wave of the Metaverse, Roblox, was very popular at the time, and then this trend spread to the Crypto industry, and the concept of the Crypto version of the Metaverse was derived. In addition, with the FaceBook was renamed Meta, which enabled related concepts to be publicized in all directions. The metaverse-related concepts of the Crypto industry also reached their peak in this environment.
In recent times, I haven’t heard anyone mention the concept of the metaverse for a long time. After checking, the report about “Roblox, the first global metaverse concept”, has been There are all kinds of negatives, including a consecutive decline in listings, 70% of revenue relying on marketing, and a market value drop of more than 70%. At the same time that the first Metaverse stock stalled, the concept of the Metaverse in the Crypto industry has almost come to an end.
Whether the concepts related to the metaverse can be developed in the future is another matter, but the development of the AI track in the Crypto industry is very similar to it, The first wave of climax started with ChatGPT The release, coupled with the extensive coverage of domestic and foreign media, has caused BN’s only two AI projects to gain momentum. Although they have been criticized by everyone for their poor technology, at present, only AI, FET and AGIX can be mentioned. It must be among them;The second wave is due to the recent release of OpenAI’s Sora, which has caused AI-related concepts to be swept across again.
It can be seen that the current development of the AI track in the Crypto industry still relies more on OpenAI. Most projects are in the conceptual stage or are mainly based on robbing, and have not yet matured well. . Imagine if OpenAI eventually stagnates or collapses, will the AI track in the Crypto industry still be so popular?
02 AI track classification
The current combination of AI + Web3 is mainly focused on Algorithm models and data directions, look specifically at:
Decentralized computing power (GPU) strong>
As AI models become more and more complex, the demand for computing resources is also growing rapidly, especially the demand for high-performance hardware such as GPUs. From an AI perspective, the application of computing power can be divided into two directions: reasoning and training. Currently, the main focus is on the inference direction because it is relatively simple and has low hardware requirements. However, the development of training direction still faces many challenges, because it needs to be combined with neural network design and has extremely high hardware requirements, so it is very difficult to implement.
The core idea of a decentralized computing system is to enable owners of computing resources to rent out their idle resources and trade them on a secondary market through competitive pricing and easy access. This means that teams can rent computing resources on demand, and resource owners can rent out their resources at competitive prices, thereby increasing market efficiency and avoiding the problem of idle resources. In addition to economic and efficiency benefits, decentralized computing systems also have the important value of being censorship-resistant.
At present, most computing power projects are just following the trend, promoting decentralization in a high-profile manner, but ignoring the inefficiency of decentralized networks. Many projects are seriously homogeneous in design.
zkML (zero-knowledge machine learning)
Since smart contracts can only do some preset events, slightly more complex running costs will be too high or impossible to implement. Integrating AI (such as machine learning models) into smart contracts can increase functionality and security. and efficiency, and improve user experience. Of course, this integration also brings additional risks, as there is no guarantee that the model supporting smart contracts will not be attacked, and it is difficult to deal with the problem of model training in the case of insufficient data.
Machine learning requires a lot of computing resources to run complex models, which makes running AI models directly within smart contracts expensive. To reduce the high cost of running AI models in smart contracts, some projects are investigating ways to use zkML. zkML can be calculated off-chain and only needs to submit proof of verification results on-chain. This can greatly reduce the need for on-chain computation, reduce costs, and maintain trustless characteristics. This method is particularly suitable for scenarios that require large amounts of computing resources, such as revenue optimization models in DeFi protocols.
< strong>AI agents are robots that can receive, understand, and perform tasks autonomously. They use AI models to make decisions and actions. In the marriage of AI and cryptocurrency, agents play an important role. Cryptocurrencies provide a permissionless and trustless payment infrastructure, and agents can take advantage of this feature.
Once agents are trained, they can have their own digital wallets, allowing them to transact directly with smart contracts. For example, a simple agent could use an AI model to search the Internet for information and make trades based on predicted market conditions. Agents can make decisions and execute transactions autonomously through smart contracts without relying on the trust of third parties.
AI empowered DApp is a very common direction. The broad applicability of generative AI enables it to be accessed through APIs to simplify and intelligentize various applications, including data analysis platforms, trading robots, blockchain encyclopedias, and more.
In addition, generative AI can also act as a chat robot (such as Myshell) or an AI companion (such as Sleepless AI), and can even create virtual characters (NPCs) in chain games. However, due to low technical barriers, many applications simply access an API and make fine-tuning to call themselves AI projects.
03 Representative Projects
There are several strong projects currently under review. After all, this industry , in an environment where the strong will always be strong, it is still too difficult for most investors to obtain alpha returns, and it is not bad to steadily capture beta. In addition, the AI track has been relatively hot in the short term recently, so be careful about the risk of a correction.
Worldcoin is The project founded by OpenAI co-founder Sam Altman aims to build a global identity and financial network to achieve widespread adoption around the world. At the heart of the project is World ID, which enables users to verify their identity online while protecting their privacy.
Worldcoin’s goal is to solve some of the problems existing in the existing financial system and provide financial services and identity verification to one billion people around the world. By building a global identity network, it aims to provide a safe, reliable and privacy-preserving way to verify a person's identity.
With the recent surge in popularity of the AI track, Worldcoin’s various data are also rising. According to official reports, as of February 21, the number of verified World ID users has exceeded 3.5 million, and the number of daily users of its World App has also exceeded one million. The market value of WLD is close to 1.2 billion, and the full circulation market value is 88 billion. In terms of background and other aspects, Worldcoin can be said to be a perfect candidate for the AI track, but due to its terrifying full-circulation market value, it can scare away most investors.
< p>
Render is focused on meeting the growing computing needs in media production, especially in areas such as augmented reality, virtual reality and AI-enhanced media. By connecting global decentralized graphics processing units (GPUs), sharing computing power, completing AI video rendering and contributing AI computing power. Render's core product is its integration with AI, helping artists generate assets and optimize digital artwork by supporting AI-related tasks. It also expands the ability to manage large-scale art collections and optimize rendering workflows.
Render’s vision is to build a decentralized computing power pool that connects the created computing power with applications that require computing power. The investors behind Render include Microsoft, Apple and other companies, and its parent company OTOY also fully supports it. OTOY is a company with a high reputation in the field of video rendering, providing video rendering technology services to top technology companies such as Microsoft, Apple and Google.
Fetch.AI is an open platform designed to build AI applications and services. It provides users with an innovative way to leverage AI agents for various tasks and transactions by combining AI and blockchain technology.
AI agents are the core of Fetch.AI. They are modular components that can perform specific tasks. These agents can connect, search, and trade autonomously, interact with other agents, and create dynamic markets. By integrating with these agents, traditional products can access AI, optimize business processes, and create new business models. Its goal is to promote the development of the smart economy, realize a decentralized digital economy, and improve the automation and efficiency of various processes.
SingularityNET is A decentralized AI platform and marketplace designed to connect various AI tools and services and facilitate cooperation and transactions between them. It provides an open platform for AI agents to communicate and cooperate with each other, while also providing users with a convenient way to obtain and use AI technology.
On the SingularityNET platform, users can delegate tasks to different AI agents through transaction information, negotiated payments and ratings. Such a decentralized structure allows agents to cooperate and compete freely, providing users with more choices and high-quality AI services.
The goal of SingularityNET is to promote the development and application of AI technology and promote the widespread application of AI in various fields. By establishing an open platform and ecosystem, it is committed to strengthening the interoperability of AI tools and services and bringing more innovation and convenience to users.
FET and AGIX are often discussed together. After all, they have been rising and falling at the same time. Generally speaking, when talking about the AI track, BN will definitely come to mind. Although they are considered by professionals It’s not good to criticize others’ technology, but because of its early advantage, it does not prevent them from following each wave of AI. The market value of both projects is approaching the US$1 billion mark.
Bittensor is An open source protocol, a blockchain-based decentralized machine learning network that aims to democratize AI by creating a platform for multiple decentralized commodity markets, or "sub-networks". Unified under a single Token system. Bittensor’s mission is to build a network comparable to large AI super companies such as OpenAI through a unique incentive mechanism and advanced sub-network architecture.
Bittensor’s system can be regarded as a machine that efficiently transfers AI capabilities to the chain through the blockchain. The network is governed by two key actors: miners and validators. Miners submit pre-trained AI models to the network and receive rewards for their contributions. Verifiers ensure the validity and accuracy of model outputs. This setup creates a competitive environment that incentivizes miners to continuously improve their models for better performance and more rewards (in the form of native token TAO). Users interact with the network by sending queries to validators, who distribute the queries to miners. The validator ranks the output of the miners and returns the highest ranked reply to the user.
Bittensor does not train models because model training is complex and costly. Instead, the network relies on a decentralized training mechanism. Averifier is tasked with evaluating miner-generated models using a specific dataset and scoring each model against specific criteria such as accuracy and loss function. This decentralized evaluation ensures continuous improvements in model performance.
Coingecko data shows that Bittensor’s current market value is about US$3.6 billion. Because it occupies the two major tracks of PoW and AI, its performance was relatively outstanding last year, but its overall liquidity is poor.
Arkham is A crypto-intelligent analytics platform that provides detailed data and analysis by connecting blockchain addresses and real-world entitieswhile leveraging the concepts of token economics to create an intelligence trading platform that enables users to buy and sell zones Information about the owner of the blockchain address. It uses an AI algorithm engine called Ultra to connect blockchain addresses with real-world entities to reveal the individuals and companies behind blockchain activity and provide relevant data and analysis.
The main product is Profiler, which provides a comprehensive view of entity or address activity, including information such as transaction history, positions, balance history, profit and loss, trading platform usage and top counterparties. By using Profiler, users can obtain detailed information about entity activities and conduct deeper analysis.
Its goal is to systematically analyze and de-anonymize blockchain transactions,to establish a decentralized intelligence-based money-making economy, and to promote transparency and efficiency in the crypto market. It is worth mentioning that OpenAI co-founder Sam Altman is one of the investors.
04 Summary
The above is a summary of the AI track, as a new narrative for this round, plus With the help of Web2, it is foreseeable that in this round of bull market, the AI track will inevitably occupy a place. In addition, we can focus on the AI progress of the Web2 industry. Generally, in major events, there will be linkage between Web2 and Web3.