Endless Imagination on the Possibilities of Combining AI and Web3

Inception Capital

Published in

Inception Capital

16 min readJun 22, 2023

Written by michaeljin.eth and Yetta He

Possibility Unbound: When AI Meets Web3

Composing AI and Web3 to build the future of internet

A thesis on AI-Web3 composability

How can AI and Web3 be compatible?

Are AI and Web3 Compatible?

— Crypto Rises and Falls, Web3 and AI Compete and Coexist

As a Web3 practitioner caught up in the AI wave, after experiencing the explosion of information in both industries over the past few months, I have compiled some thoughts and research to share with other Web3 practitioners:

AI and Web3: one breaking through our imagination of productivity limits, the other reshaping our understanding of economic models. As representatives of the forefront technologies for future development, their combination seems natural, always inspiring endless imagination. However, when we turn our attention to reality, we find that there are very few projects that truly integrate the two. The collision of two tracks has spawned new narratives, but also given rise to many bubbles and gimmicks. Many beautiful visions that complement each other in theory may not have real demand in reality. Projects that can meet real needs may also face difficulties in implementation due to cost or technological bottlenecks.

I believe the idea that Web3 and AI compete and coexist is also proportional to the number of Web3 projects with AI content seen in the primary market, as well as the number of non-essential AI projects that are Web3-ified. AI-native entrepreneurs/project owners do not necessarily think about how to integrate with Web3, such as putting data on the chain, economic models, production relationship allocation, etc. Because in AI projects with large models from the bottom up, resource requirements are high, and the large amount of resources needed makes AI training and operation highly centralized. Therefore, I remain very cautious about the actual feasibility of Web3 projects claiming to help improve AI production relationships.

The Web3 market has encountered significant bottlenecks both at the macro policy level and the innovation level. Leaving aside new regulatory pressures, in terms of innovation, when AI attracts the attention of most users, builders, and investors with its high-speed improvement of productivity and ability to replace human thinking, Web3’s industry innovation crisis becomes even more difficult to conceal. Web3 has not had an innovation like AI for a long time. Honestly, most of the new projects that have garnered attention lately are minor improvements to past technologies/products. For example, better staking methods, multi-chain wallets with better user experience, meme coins with new gameplay, and better liquidity DEXs on new public chains. Do these so-called “innovations” really help bring in more users or increase blockchain penetration rates, or are they truly what this industry needs?

We need some new fields that can bring AI into Web3 and allow Web3 to go beyond. The practical use of blockchain’s underlying nature, such as: 1. content creation confirmation, 2. identity verification, 3. revolution of the financial system, 4. termination of trust, etc., is related to the future of the entire industry’s next paradigm shift.

With the aim of seeking organic integration, this article comprehensively reviews the new fields created by the combination of AI and Web3 from the perspective of technical adaptation and complementarity at the underlying level, and summarizes and analyzes the actual needs, development bottlenecks, and prospects of each direction in these fields.

TLDR

There are conflicts in the underlying logic between AI and Web3, as the large resources required for AI big models make AI training and operation highly centralized, while Web3 built on blockchain prioritizes decentralization and transparency. This makes it difficult for AI and Web3 to integrate at the fundamental level, and the validity of their business logic and actual demand needs further consideration.
However, it is precisely this contradictory logic at the fundamental level that allows AI and Web3 to complement each other and become solutions to each other’s pain points, without necessarily becoming each other’s narrative core, and help promote their respective development. The combination of the two technologies will also bring many new narratives and leave a huge space for imagination. Web3’s economic model design can help many AI project parties improve capital utilization and promote project growth and activity, while the benefits of blockchain, such as reducing infrastructure costs, verifying identity, injecting democracy and transparency into the data black box of artificial intelligence, and providing data contribution incentives, can provide new ideas for AI project team’s product design.
At the infrastructure level, Web3’s decentralized mechanism can solve the risks and problems that AI currently faces from the bottom up, such as privacy protection and data abuse.
Provide a decentralized market for essential elements of AI development, such as computing power and data to maximize the use of idle resources, optimize resource utilization and allocation, and help promote the development and application of AI big models.
Web3’s decentralized mechanism allows AI to become more democratic from the lowest level. By deploying, training, and using AI in a decentralized manner, users’ data privacy can be better protected while also having the opportunity to receive rewards through data sharing.
Blockchain can also be used to record and monitor AI behavior, thereby improving AI security and promoting the use of automated AI agents in various scenarios.
At the application layer, AI can help the development and popularization of Web3 applications.
Firstly, as a productivity tool, AI can help Web3 applications greatly improve development speed, while as a knowledge engine, it can also reduce the interaction and learning costs between users and dApps, helping more users enter Web3.
AI can significantly reduce the technical threshold for developing dApps and launching projects, making the project’s competitiveness more focused on innovation and operation. It is precisely in this direction that generative AI can bring new narratives to Web3 applications, such as embedding virtual people and character AI into the game and social ecology, and developing new gameplay.

Infrastructure Layer

Token Incentives and Governance Mechanisms: Empowering AI Infrastructure with Decentralized Markets

In the era of AI large language models (LLM), every aspect of the infrastructure that supports AI development will become particularly important.

In the process of building and developing AI infrastructure, a key challenge is how to effectively incentivize and coordinate participants to jointly promote the development and operation of the system. Decentralized markets and token incentives provide a novel and powerful approach to addressing this issue. In such markets, tokens play an important role as a digital asset and value medium. Tokens can represent specific rights, functions, or resources, and their transactions and transfers are conducted through smart contracts, achieving a secure, transparent, and automated transaction process.

For AI infrastructure, token incentives can play multiple roles. First, tokens can serve as a means of incentive to reward and encourage participants who contribute to AI infrastructure. These contributions can include providing computing resources, datasets, algorithm models, computing power, and more. For example, the recently popular character AI project, MyShell, has achieved data flywheel effects through chatbot creation workshops and data analysis. Users can customize the voice, function, and knowledge base of chatbots on the MyShell platform and interact with them. The data collected from these interactions is used to improve the robot’s performance and personalized services, attract more users to the platform, further increase data and value, and form a virtuous cycle of growth.

By providing token incentives to participants, Web3’s economic model can also attract more people to participate in the construction of AI infrastructure, promote resource sharing and cooperation. Tokens can be used to realize the flow and exchange of value in decentralized markets. Participants can buy and sell resources, services, and algorithm models, etc. using tokens to achieve transactions and collaborations in the market. This mechanism of value flow can provide a more flexible and efficient way for the development of AI infrastructure, enabling participants to better meet their own needs and interests.

Homomorphic Encryption and Federated Learning: Integrating Privacy Protection into AI Training

Ensuring effective model training while protecting personal privacy and data security has been a long-standing challenge. In this regard, homomorphic encryption technology provides a powerful privacy protection method that can integrate privacy protection into the bottom-level training of AI to ensure the security of sensitive data.

Homomorphic encryption is a special encryption technology that allows data to be calculated under encryption without decryption. This means that model training and computation can be performed on encrypted data without revealing the content of the original data. By applying homomorphic encryption to the bottom-level training process of AI, privacy protection can be achieved without leaking sensitive data.

When using homomorphic encryption for AI training, there are several key steps and considerations:

Data encryption: Encrypt the data involved in AI training using homomorphic encryption algorithms. This ensures the privacy and confidentiality of the data during the training process.
Encrypted computation: Perform calculation operations including model training, optimization, and inference under encryption. Homomorphic encryption technology makes these calculations possible without decrypting the data.
Secure parameter sharing: The parties involved in the training need to share and exchange the secure parameters required for encrypted computation. These parameters are used to control the homomorphic encryption process and decrypt the results.
Encrypted result processing: After completing the encrypted computation, the results can be decrypted to obtain the final model weights or prediction outputs. Appropriate security measures need to be taken when decrypting the results to prevent data leakage or unauthorized access.

Homomorphic encryption technology has some advantages and potential applications in integrating privacy protection into the bottom-level training of AI:

Privacy protection: Homomorphic encryption enables model training on sensitive data without actually accessing or exposing the data. This helps maintain personal privacy and data owner control.
Data collaboration: Multiple data owners can participate in AI training together without sharing their original data. Homomorphic encryption technology makes this data collaboration possible, promoting opportunities for cooperation and sharing.
Legal compliance: Homomorphic encryption provides a compliance-compliant method for AI training for sensitive data subject to legal and regulatory restrictions, such as medical records or financial data.

This type of privacy can also be achieved through decentralized computing platforms. For example, Fluence (backed by OP) is a decentralized computing platform that can run many programs including AI, aiming to achieve digital innovation through peer-to-peer applications. It provides an open Web3 protocol, framework, and tools for developing and hosting applications, interfaces, and back-ends on an unlicensed peer-to-peer network.

zkML and On-Chain AI Inference: Monitoring AI Agent Behavior and Responsibility Constraint

In the context of the rapid development and wide application of artificial intelligence (AI) technology, ensuring that the behavior of AI systems complies with ethical and legal requirements has become particularly important. AI systems are often viewed as agent entities capable of performing tasks and making decisions that may have profound impacts on humans and society. Therefore, monitoring AI agent behavior and constraining its responsibilities has become a key issue in safeguarding public interests and individual rights. zkML, as an innovative approach, provides a secure, verifiable, and transparent solution for monitoring AI agent behavior and responsibility constraint by combining zero-knowledge proof and blockchain technology. Through a decentralized monitoring and constraint framework established by zkML, the decision-making process and behavior path of AI agents can be monitored and audited in real-time.

This decentralized monitoring mechanism ensures transparency and traceability, enabling non-compliant behavior or inappropriate decisions to be discovered and corrected in a timely manner. zkML also provides a mechanism for AI agent behavior responsibility constraints. By combining smart contracts with the operation and decision-making process of AI systems, a series of rules and conditions can be set to limit the scope of AI agent behavior and ensure compliance with ethical standards and legal regulations. This responsibility constraint mechanism makes AI systems a reliable tool that can create value for human society without abusing power or causing harm to human interests. This technology lays an important foundation for building sustainable, ethical, and responsible artificial intelligence systems.

Execution Layer

Improving Productivity, Accelerating Web3 Development

In the development of Web3, artificial intelligence (AI) plays an important role in combining with various fields to improve productivity and create better user experiences. Here are several key areas where AI is combined with Web3:

AI and On-Chain Data Collection and Analysis
AI technology plays an important role in on-chain data collection and analysis. As a distributed database, blockchain records a large amount of transactions and information. By utilizing AI technology, on-chain data can be better understood and utilized. For example, Web3 Analytics is an AI-based analysis platform that uses machine learning and data mining algorithms to collect, process, and analyze on-chain data. It can help users gain insights into on-chain transactions, market trends, and user behavior patterns, providing more accurate data analysis and decision-making support. Similar platforms include MinMax AI, which provides AI-based on-chain data analysis tools to help users discover potential market opportunities and trends.
AI and Automated dApp Development
AI technology is also very important in the process of automated dApp development. Smart contract and dApp development usually require writing a lot of code and performing tedious testing and deployment work. By combining AI with smart contract and dApp development tools, a more efficient and intelligent dApp development process can be achieved. AI can help automate code generation, smart contract verification and testing, as well as dApp deployment and maintenance. This can save time and resources and improve the efficiency and accuracy of the development process. For example, some AI-assisted development tools use natural language processing and machine learning technologies to help developers write smart contracts faster and automatically detect and fix potential errors.
AI and On-Chain Transaction Security
In the Web3 world, on-chain transaction security is crucial. Due to the openness and transparency of blockchain, there are risks of malicious attacks, fraudulent behavior, and data leaks. AI technology can be used to enhance the security and privacy protection of on-chain transactions. For example, the Web3 security platform SeQure uses AI to detect and prevent malicious attacks, fraudulent behavior, and data leaks, and provides real-time monitoring and alert mechanisms to ensure the security and stability of on-chain transactions. Similar security tools include AI-powered Sentinel.

Optimizing Resource Allocation, Navigator in the Web3 World

In the Web3 world, optimizing resource allocation is a key challenge. With the combination of blockchain technology and artificial intelligence, we can use AI as a navigator to achieve more effective resource allocation and utilization. Here are several areas where AI plays a navigational role in the Web3 world:

AI and On-Chain Activity Optimization:
On-chain activities on the blockchain include transactions, contract executions, and data storage. Through AI’s intelligent analysis and prediction capabilities, we can better optimize on-chain activities, improve overall efficiency and performance. AI can help identify transaction patterns, detect abnormal activities, and provide real-time recommendations to optimize the resource allocation of the blockchain network through data analysis and model training.
AI and On-Chain Advertising Mechanism:
In the Web3 world, advertising is also a resource. AI can play a key role in on-chain advertising mechanisms, helping advertisers more accurately target their audience and provide personalized advertising content. By analyzing the data and behavior patterns of on-chain users, AI can achieve more accurate advertising placement, improve click-through rates and conversion rates, and optimize resource allocation and utilization.
AI and DAO Governance:
Decentralized Autonomous Organizations (DAOs) are a new type of organization in the Web3 world. AI can be an important tool for DAO governance, assisting in decision-making, voting mechanisms, and community governance. AI can help DAO members better understand community needs and opinions through data analysis and prediction, and provide decision-making support. With AI’s participation, DAOs can operate more efficiently, optimize resource allocation, and promote community development and growth.

Application Layer

Lowering the Barrier to Entry, the Propeller of Web3 Popularization

Friendly user interface embedded with AI
For example, the Web3 auditing platform Fuzzland uses AI to help code auditors check for code vulnerabilities and provide natural language explanations to assist in auditing expertise. Fuzzland also uses AI to provide natural language explanations for formal specifications and contract code, as well as some sample code to help developers understand potential issues in the code. By combining AI technology with auditing expertise, Fuzzland makes it easier for Web3 industry developers to understand and explain code, improving audit efficiency and accuracy.
AI-embedded smart contract interpretation
AI-embedded smart contract writing
In the development of Web3, lowering the barrier to entry is the key to achieving popularization. In order to achieve this goal, AI technology plays an important role in providing friendly user interfaces, smart contract interpretation, and smart contract writing. The friendly user interface embedded with AI provides a more intuitive and convenient user experience for users of the Web3 platform. Traditional blockchain technology usually requires users to learn complex commands and syntax to interact and execute operations. However, by applying AI technology to user interface design, natural language processing, graphical interfaces, and other functions can be achieved, allowing users to easily use Web3 platform for various operations without having to understand technical details. AI also provides users with the ability to better understand and explain smart contracts. By applying AI technology, smart contracts can be automatically parsed and visually displayed, clearly presenting logic flow and condition expressions in smart contracts to users, improving user understanding and trust in smart contracts.

Plots enriched, the creative library of the Web3 world

AI and generative NFT
AI automatic trading agent
Character AI and game NPC
AI and automatic rendering of metaverse scenes

The rise of generative AI has brought new possibilities to the creative industry and a more diverse and innovative experience to the Web3 world, allowing users to participate in rich plots and gameplay. In the past NFT bull market, AI has injected unlimited creativity into generative NFT. Generative NFT (Non-Fungible Token) is a kind of artwork or digital asset based on algorithms and data. Through AI technology, various unique and diverse artworks and characters can be generated. These generative NFTs can become characters, props, or scene elements in games, virtual worlds, or metaverses, providing users with rich choices and personalized experiences. In the DeFi boom, AI automatic trading agents have also brought convenience and efficiency to the economic transaction process in the creative library. In the Web3 world, users can obtain profits by owning, trading, or participating in digital assets in the creative library. AI automatic trading agents use intelligent algorithms and machine learning technology to automate asset trading, helping users to obtain the best trading opportunities and maximize profits.

AIGC has also brought new gameplay and creativity to content platforms and UGC communities. For example, Yodayo (backed by OP) is an AI art platform for virtual anchors and anime fans to share and create more of what they love. Yodayo makes content creation and interaction on the platform easier and more accessible by integrating the AIGC engine, allowing most users who are typically “silent” on traditional platforms to become creators and up-and-coming people from content consumers. This helps establish closer connections with the community and make contributions.

The combination of character AI and game NPC has brought a more realistic and interactive experience to the game plot in the creative library. By applying AI technology to game characters and non-player characters (NPC), they can be given intelligent behaviors, autonomous decision-making, and emotional expression capabilities. This makes the game plot more diverse, and players can interact with characters with realistic artificial intelligence to explore the game world and solve various challenges together. The combination of AI and automatic rendering of metaverse scenes creates a more realistic and vivid environment for the virtual world in the game. For example, Inward AI systematically analyzes the behavior and preferences of players, based on their previous interactions, allowing key characters in the game to provide unique tasks or information, shaping personalized storylines for each player. The real-time combat AI provided by rctAI makes every battle lifelike. The characters that battle with players can learn continuously from their battle tactics, improve their skills, and adjust their strategies, making battles more uncertain and exciting. The integration of these AI technologies creates a dynamic and interactive narrative, realistic and challenging battle scenes, and makes the game world more immersive and attractive.

Conclusion

As a Web3 practitioner swept by the AI wave, after experiencing the information explosion of these two industries in the last few months, we have had a deeper reflection on the combination of AI and Web3. Although the two have conflicting logic in the underlying logic, and the centralization characteristics of AI seem difficult to reconcile with the decentralization principle of Web3, it is precisely this contradictory logic that enables AI and Web3 to complement each other and become solutions to each other’s pain points, promoting each other’s development. The decentralized mechanism of Web3 can fundamentally solve the problems that AI faces such as privacy protection and data abuse. The application of Web3 and blockchain technology can also monitor and record the behavior of AI, improve the security of AI, and promote the promotion and application of automatic AI agents in various fields.

Although the combination of AI and Web3 in the underlying layer is facing many challenges and restrictions, many new possibilities and narratives can be created at the application level. AI can become an important help for Web3 applications, greatly improving the development speed of Web3 applications, reducing the interaction and learning costs of users with dApps, and helping more users enter the Web3 world. At the same time, while reducing the technical threshold for dApp development and project release, AI can also bring more gameplay and competitiveness to projects in innovation and operation, such as embedding virtual people and character AI in game and social ecological environments, bringing a new narrative and experience to Web3 applications, and further promoting the development and promotion of the Web3 industry.

Although the combination of AI and Web3 faces some challenges and limitations, we believe that only the organic combination of the two can support the narrative and ideals of the next generation of the Internet. We look forward to seeing more innovative projects that can bring AI into Web3 and push Web3 to a wider field, and we hope that the development of these two cutting-edge technologies can continue to help each other break through technical bottlenecks, overcome cost limitations, and create a more intelligent and open future.

ABOUT OP CRYPTO

OP Crypto is a leading high conviction, early-stage venture capital firm in the crypto and blockchain industry, specializing in pre-seed and seed stage investments. The fund successfully raised $50M in September 2021 and has since invested in over 30 projects, including companies such as Scroll Tech, Snackclub, Merit Circle, Omni, and Fyde.

With the support of prominent investors like Bill Ackman and Alan Howard, as well as institutions like Galaxy Digital, Huobi, and DCG, OP Crypto has access to a global network of venture funds, scouts, and ecosystem partners to source the most competitive deals in the market.

With a core team based in the United States and strong ties to the APAC region, the fund serves as a bridge between East and West. Additionally, a dedicated portfolio team provides post-investment support to founders in areas such as marketing, tokenomics, and networking.

Learn more about OP Crypto at: Website