Which Web3 projects are worth paying attention to, from AI infrastructure to application scenarios? How do the two combine?
The initial stage of Web3 social introduces AI more as a narrative tool. This article reviews 5 AI infrastructure projects and 5 AI application scenario projects in the crypto space, such as Fetch.ai as an intermediary, allowing clients to trade datasets using its native token...Author: Biscuit & alertcat.eth, ChainCatcher
OpenAI's chatbot ChatGPT reached 100 million monthly active users just two months after its launch, becoming the fastest-growing application in history. Such a powerful "follower growth" ability quickly transferred the AI hype to the crypto space. On January 10, Bloomberg reported that Microsoft is considering investing $10 billion in ChatGPT developer OpenAI, completely igniting all AI concept cryptocurrencies, with FET, AGIX, and others rising over 200% within a month.
With the boost of capital, can these two highly anticipated frontier technologies merge? Artificial intelligence uses computers to solve problems by mimicking the thinking abilities of the human brain. OpenAI provides a large amount of training data to natural language processing (NLP) models, making them more powerful. In the crypto world built on blockchain technology, the vast on-chain data generated daily can provide "fuel" for AI engines, allowing AIGC to feedback better strategies.
Additionally, as AI algorithms become increasingly intelligent, it becomes more difficult for people to understand how they make decisions and conclusions. The immutable nature of blockchain can help us access unchangeable records of the data and processes used by AI in its decision-making.
AI concept crypto projects (Source: Rootdata)
Compared to traditional AI like Stability AI and ChatGPT, which have gained significant attention and adoption, the greater imagination of blockchain lies in its potential to change the economic system of AI models. After the FOMO sentiment fades, this article will explore what characteristics do crypto projects that introduce AI technology have? What kind of chemical reactions can AI and blockchain produce?
AI Infrastructure
The common feature of AI infrastructure projects is the distribution and sale of the architecture (data, models, and computing power) of traditional AI. They generally use their native tokens as a medium of exchange. They often act as intermediaries between users and service providers, building a decentralized trading market. These projects are essentially DApps that serve as intermediary platforms for tasks that traditional AI needs to accomplish, such as NLP, AI voice, and CV. Essentially, they create a decentralized market for token pricing and exchange.
Openfabric is a platform for building and connecting AI applications. Through this platform, collaboration among AI innovators, data providers, enterprises, and infrastructure providers will facilitate the creation and use of new intelligent algorithms and services. The Openfabric ecosystem consists of four roles: algorithm creators, data providers, infrastructure providers, and service consumers, where service consumers need to pay the other three types of service providers.
- Algorithm Creators: Utilize their expertise to create AI algorithms that can solve complex business problems.
- Data Providers: Ensure the distribution of large amounts of data needed to train AI algorithms.
- Infrastructure Providers: Operate all the hardware for the AI platform.
- Service Consumers: End users who need specific business products or services.
Oraichain is an AI-driven blockchain oracle and ecosystem. In addition to data oracles, Oraichain aims to become a complete AI ecosystem in the blockchain space, serving as a foundational layer for creating smart contracts and DApps. Built on AI, Oraichain has developed many important innovative products and services, including AI price feeds, fully on-chain VRF, Data Hub, an AI Marketplace with over 100 AI APIs, AI-based NFT generation and copyright protection, Royalty Protocol, an AI-driven yield aggregator platform, and Cosmwasm IDE.
Fetch.ai is a blockchain platform based on artificial intelligence and machine learning that allows anyone to share or trade data. As an autonomous machine-to-machine ecosystem, any independent party network can become a network agent for Fetch.ai, recording any protocols generated between agents on the Fetch.ai blockchain. FET is the native token of the Fetch AI blockchain and serves as the primary medium of exchange for transaction payments.
Source: Fetch.ai Blog
SingularityNET is a decentralized AI platform and marketplace. Developers publish their services on the SingularityNET network for any internet-connected user to access. Developers can charge for their services using the native AGIX token. Services can provide cross-domain reasoning or model training, such as for images, videos, voice, text, time series, bio-AI, and network analysis.
The SingularityNET ecosystem will provide AI services for the platform and create large-scale utilization of AGIX tokens. These SingularityNET derivatives are being developed in several strategically chosen vertical markets, including DeFi, robotics, biotechnology and longevity, gaming and media, arts and entertainment (music), and enterprise-level AI.
The Gensyn protocol is a Layer 1 network for deep learning computation, allowing supply-side participants to contribute computing time to the network and execute ML (machine learning) tasks through instant rewards. The protocol does not require administrative oversight or enforcement but facilitates task allocation and payment programmatically through smart contracts. The fundamental challenge of the network is verifying completed ML work. This is an intersection of complexity theory, game theory, cryptography, and optimization. The Gensyn ecosystem consists of four roles: submitters, solvers, verifiers, and whistleblowers.
- Submitters: Provide tasks to be computed and pay for completed work units.
- Solvers: Execute model training and generate proofs for verifiers to check.
- Verifiers: Link non-deterministic training processes to deterministic linear computations, replicate solver proofs, and compare distances with expected thresholds.
- Whistleblowers: Check the work of verifiers and raise challenges for the chance to earn cumulative rewards.
Gensyn's vision is to provide essential infrastructure components for Web3 applications through decentralized ML computation, reducing DApps' reliance on Web2 infrastructure.
Application Scenarios
In such application scenarios, projects aim to address emerging demands arising from recent blockchain developments using AI.
These demands may include enabling blockchain game users to skip tedious operations, allowing developers to quickly create blockchain games, facilitating social interactions on blockchain platforms, generating virtual personas with unique personalities, or detecting fraudulent NFT projects, among others. Unlike traditional AI platforms, these projects have strong demand inimitability, creating a deep moat. However, as platforms that leverage emerging demands as selling points, their development challenges lie in customer acquisition—how to attract enough customers to prove that the demand for their platform is sustainable and objectively exists has become a significant issue in the development of such platforms.
Blockchain Gaming Direction
Under the mainstream financial system of the crypto game "P2E" model, users face constantly changing gameplay and a large number of repetitive basic operations. AI can provide players with stable automated processes and develop higher win-rate game strategies. rct AI is a project that uses AI to provide a complete solution for the gaming industry, with its core technology, the Chaos Box, being an AI engine based on deep reinforcement learning. rct AI has developed a DRL (Deep Reinforcement Learning) model for Axie Infinity, where the number of combinations of all cards is approximately 10^23, along with the game's competitive features, enhancing efficiency and win rates through a large amount of simulated battle data.
Additionally, AI can provide action prototypes for developers. Mirror World is a Solana-based gaming matrix virtual world that has launched AI technology to create Mirrama, which combines Roguelike gameplay, and the PVP-based arena game Brawl of Mirrors. Furthermore, Mirror World has introduced a series of NFTs that can interoperate within the game, with these NFTs' prototypes completed using AI action algorithms.
Related Reading: 《Dialogue with rct AI: It's Time to Consider How Blockchain Will Change Game Publishing》
Social Direction
PLAI Labs focuses on building the next-generation social platform using AI and web3, allowing users to play, chat, battle, trade, and adventure together. The platform secured $32 million in funding from a16z in January 2023. Currently, PLAI Labs has showcased two products:
- Champions Ascension, a massively multiplayer online role-playing game (MMORPG) where players can own their characters as NFTs and engage in battles in large arenas, complete quests, and build and compete in custom dungeons while trading digital items.
- An AI protocol platform that will help process everything from user-generated content (UGC) to matching 2D to 3D asset rendering.
PLAI Labs plans to launch the V2 white paper this year, detailing the core economic cycle (enhancing experiences using NFTs and blockchain), UGC toolkit (including AI) plans…
Related Reading: 《Veteran Entrepreneur Takes a New Step: PLAI Labs Discusses Why They Chose Web3》
NFT Direction
Alethea AI proposed the concept of iNFT, which is a technology that combines artificial intelligence and blockchain. With the integration of AI, NFTs gain interactivity, generativity, scalability, and unique personality traits.
In simple terms, if an NFT is a digital human work, after integrating AI, it becomes an iNFT, capable of chatting with users. On June 10, 2021, the world's first iNFT, Alice, was auctioned at Sotheby's for $478,800.
Altered State Machine (ASM) is an innovative project that combines NFT, artificial intelligence, and machine learning, providing training power for AI-driven NFTs. Its vision is to become a protocol for ownership and monetization of AI using NFT technology. In the ASM ecosystem, AI-based avatars are called Agents, consisting of a brain and an avatar. The project also issued the ASTO token to power the ASM ecosystem.
Related Reading: 《Detailed Explanation of Altered State Machine: Innovative Exploration of Evolving NFTs Using AI and Machine Learning》
Optic is establishing an AI NFT verification protocol focused on NFT fraud analysis and value discovery within the community, aiming to help the entire NFT market achieve higher authenticity and transparency. The Optic smart engine learns from real NFT collections and then retrieves NFT collections from the market. Afterward, Optic returns a matching score indicating the degree of match between the checked NFT and the real NFT.
Optic completed a $11 million funding round led by Pantera Capital and Kleiner Perkins in July 2022, with participation from Circle Ventures, Polygon Ventures, and others. Currently, OpenSea has adopted Optic's Copymint detection service.
Related Reading: 《Analysis of Optic: AI NFT Verification Protocol》
Trend Analysis
From the current development paths of blockchain AI projects, the AI infrastructure consists of three parts: data, algorithms, and computing power. A normal AI project aiming to achieve generative or analytical capabilities requires models, datasets, and software entities that call the models along with their GUIs. Therefore, the distribution of models and datasets in this field, the training of models (computing power leasing), and the development of software front ends have formed intermediaries, which will give rise to blockchain AI projects aimed at efficiently meeting customer needs.
For example, as mentioned above, Fetch.ai acts as an intermediary, allowing customers to trade datasets using its native token. SingularityNET allows customers to purchase computing power training services from developers, while Openfabric AI's customers need to obtain services such as models (algorithms), datasets, and infrastructure (software) from providers. Humans.ai essentially packages trained AI models with datasets as NFTs, which users purchase using the native token.
Gensyn is essentially a decentralized computing power leasing platform. These are all projects that serve as intermediary platforms for tasks that traditional AI needs to accomplish, such as natural language processing, AI voice, and image generation.
As decentralized applications in blockchain generate new demands, AI projects based on blockchain gaming, social direction, and NFT direction aim to address user pain points in blockchain. For instance, rct.ai solves the problem of manual repetitive operations for blockchain game users, Mirror World addresses blockchain game development, while other projects focus on blockchain social and NFT development.
Currently, in the early stages of Web3 social, the introduction of AI is more of a narrative tool. In the future, AI projects may explore several possible directions:
- Enhancing Data Privacy: Web3 can maximize data privacy protection through zk technology, while AI can analyze data without compromising privacy.
- Smart Contracts: Web3 technology can integrate AI applications into Web3 applications through smart contracts, achieving controllability over AI models. Such applications can be used for trading models and datasets, automating trading processes while using ZK technology to protect user data. However, this type of project faces challenges from open-source datasets and models. Consider this: if users can obtain open-source data and models on Hugging Face and use auto-train for training, why would they trade on blockchain platforms? Facing competition from Web2 companies, Web3 AI model and dataset trading lacks sufficient moats.
- More Efficient Machine Learning: Web3 technology can enhance the efficiency of machine learning through decentralization, making AI applications faster and more reliable. This has already been applied in traditional AI training; for example, the improved version of AlphaGo, KataGo, used distributed training technology, allowing global participants who wish to update this AI to voluntarily provide computing power for training. In blockchain applications, this could be similar to Gitcoin, where donating computing power can earn POAP, or like AMM providing liquidity incentives, becoming a paid computing power leasing platform. However, due to the high volatility of cryptocurrency prices, such applications do not hold advantages over traditional GPU computing power leasing unless the platform itself engages in financial business, sufficiently subsidizing users from the value captured by the protocol, like Numerai, which profits from the stock market using AI technology, thus attracting enough users to provide the three essential elements of AI to the platform.
Conclusion
Currently, both blockchain-native AI infrastructure and crypto projects that leverage AI engines to realize application scenarios are in their infancy, with the main goal of creating a suitable underlying infrastructure and refining the integration of token economics with hardware providers, data providers, and AI algorithm solutions.
However, the integration of the two also faces numerous challenges. First, the complexity of blockchain technologies such as Rollup and ZK will pose challenges for AI to acquire data. Secondly, there is insufficient continuous experimental data to support the applicability of AI in the blockchain ecosystem, as well as the AI engine's ability to adjust in response to emergencies. Finally, the frequent emergence of fraudulent projects that exploit AI concepts in the crypto space can easily lead to a loss of confidence in exploring this field.
All blockchain AI projects that aim to solve traditional AI problems need to answer one question: why does this platform need to introduce tokens on the blockchain? This creates a disadvantage for platforms that trade existing assets in the Web2 market, such as models, data, and computing power.
Token economics, like a flywheel, can change the rise and fall cycle of a project. Currently, if one hopes for a positive flywheel, they need to consider the actual users of the platform, i.e., the customer acquisition issue. The inimitability of demand is a moat for a project; projects lacking a moat may achieve short-term success but will not have enough users and a robust developer ecosystem. When demand is a false proposition, economic incentives are unsustainable, and the project's lifecycle will shorten. We look forward to more AI + Web3 projects based on real users and inimitable demands emerging. They aim to fulfill needs that are either absent or poorly addressed in Web2, thus necessitating the introduction of Web3.
Regardless, the integration of AI into Web3 is a technological trend for the future, and some Web3 application instances that combine artificial intelligence have already emerged at this stage. Over time, more related Web3 infrastructure and new models will follow.
