Original source: IOSG Ventures
Preface
The recent rapid development of large language models (LLMs) has sparked interest in using artificial intelligence (AI) to transform various industries. The blockchain industry is not immune, with the emergence of the AI x Crypto narrative bringing it into the spotlight. This article explores three main ways to merge AI and cryptography, and explores the unique opportunities of blockchain technology to solve AI industry problems.
AIxCrypto’s three approaches include:
- Integrating AI into existing products: Companies like Dune are using AI to enhance their products, such as introducing SQL copilot to help users write complex queries.
- Building AI infrastructure for the crypto ecosystem: Startups such as Ritual and Autonolas are focused on developing AI-driven infrastructure tailored for the needs of the crypto ecosystem.
- Using blockchain to solve AI industry problems: Projects such as Gensyn, EZKL and io.net are exploring how blockchain technology can solve challenges facing the AI industry, such as data privacy, security and transparency.
What is unique about AI x Crypto is that blockchain technology has the potential to solve problems inherent in the AI industry. This unique intersection opens up new possibilities for innovative solutions that benefit both the AI and blockchain communities.
In delving into the AI x Crypto space, we aim to identify and showcase the most promising applications of blockchain technology in solving AI industry challenges. By partnering with AI industry experts and crypto builders, we are committed to promoting the development of cutting-edge solutions that leverage the strengths of both technologies.
1. Industry Overview
The AI x Crypto field can be divided into two categories: infrastructure and applications. While some existing infrastructure continues to support AI use cases, new players are launching entirely new AI-native architectures in the market.
1. 1 Computing Network
In the field of AIxCrypto, computing networks play a crucial role in providing the infrastructure required for AI applications. These networks can be divided into two types based on the tasks they support: general-purpose computing networks and specialized computing networks.
1.1.1 General Computing Network
General purpose computing networks such as IO.net and Akash provide users with access to machines via SSH and offer a command line interface (CLI) that enables users to build their own applications. These networks are similar to virtual private servers (VPS), providing a personal computing environment in the cloud.
IO.net is based on the Solana ecosystem and focuses on GPU leasing and computing clusters, while Akash, based on the Cosmos ecosystem, mainly provides CPU cloud servers and various application templates.
IOSG Ventures’ view:
Compared to the mature Web2 cloud market, computing networking is still in its early stages. Web3 compute networks fall short of Web2’s “Lego” building blocks, such as serverless functions, VPS, and database cloud projects based on major cloud providers such as AWS, Azure, and Google Cloud.
Advantages of computing networks include:
- Blockchain technology can utilize unused computing resources and personal computers to make the network more sustainable.
- Peer-to-peer (P2P) design allows individuals to monetize unused computing resources and provide lower-cost computing, potentially reducing costs by 75% -90%.
However, computing networks are difficult to put into actual production and replace Web2 cloud services due to the following challenges:
- Pricing is certainly the main advantage of general computing networks, but it is still challenging to compete with mature Web2 cloud companies in terms of functionality, security, and stability.
- Peer-to-peer style may limit the ability of these networks to quickly deliver mature and robust products. Decentralized nature will increase development and maintenance costs.
1.1.2 Dedicated computing network
Dedicated compute networks add an extra layer on top of general-purpose compute networks, allowing users to deploy specific applications through configuration files. These networks are designed to meet specific use cases, such as 3D rendering or AI inference and training.
Render is a professional computing network focused on 3D rendering. In the field of AI, new players like Bittensor, Hyperbolic, Ritual, and fetch.ai focus on AI reasoning, while Flock and Gensyn focus mainly on AI training.
Source: IOSG Ventures
While dedicated AI inference and training computing networks are still in their early stages, we anticipate that Web3 AI applications will prioritize the use of Web3 AI infrastructure. This trend is already evident in collaborations such as Story Protocol and Ritual’s partnership with MyShell to introduce AI models as intellectual property.
Although killer applications built on these emerging AI x Web3 infrastructures are yet to emerge, the potential for growth is huge. As the ecosystem matures, we expect to see more innovative applications that leverage the unique capabilities of decentralized AI computing networks.
2. Data
Data plays a vital role in AI models, and data is involved in all stages of developing AI models, including data collection, training data set storage, and model storage.
2.1 Data Storage
Decentralized storage of AI models is critical to providing inference APIs in a decentralized manner. Inference nodes should be able to retrieve these models from anywhere at any time. With AI models potentially reaching hundreds of gigabytes in size, a powerful decentralized storage network is needed. Leaders in decentralized storage, such as Filecoin and Arweave, may be able to provide this functionality.
IOSG Ventures’ view:
There are huge opportunities in this area.
- A decentralized data storage network optimized for AI models, providing functions such as version control, storage of different low-precision model quantification, and fast downloading of large models.
- Decentralized vector databases, as they are often bundled with models, provide more accurate answers by inserting necessary knowledge relevant to the question. Existing SQL databases can also be added with vector search support.
2.2 Data collection and labeling
Collecting high-quality data is critical for AI training. Blockchain-based projects such as Grass use crowdsourcing to collect data for AI training, leveraging personal networks. With appropriate incentives and mechanisms, AI trainers can obtain high-quality data at a lower cost. Projects such as Tai-da and Saipen focus on data labeling.
IOSG Ventures’ view:
Some of our observations on this market:
- Most data labeling projects are inspired by GameFi, attracting users with the “tag to earn” concept and developers with the promise of lowering costs for high-quality labeled data.
- There is currently no clear leader in this area, and Scale AI dominates the Web2 data labeling market.
2.3 Blockchain Data
When training AI models specifically for blockchain, developers need high-quality blockchain data that they want to be able to use directly in their training process. Spice AI and Space and Time provide high-quality blockchain data with SDKs, allowing developers to easily integrate the data into their training data pipelines.
IOSG Ventures’ view:
As the demand for blockchain-related AI models grows, the demand for high-quality blockchain data will surge. However, most data analysis tools currently only provide the ability to export data in CSV format, which is not ideal for AI training purposes.
To facilitate the development of blockchain-specific AI models, it is crucial to enhance the developer experience by providing more blockchain-related machine learning operations (MLOP) capabilities. These features should enable developers to seamlessly integrate blockchain data directly into their Python-based AI training pipelines.
##3.ZKML
Centralized AI providers face trust issues due to incentives to use less complex models to reduce computational costs. For example, there were times last year when users thought ChatGPT was underperforming. This was later attributed to an OpenAI update aimed at improving model performance.
Additionally, content creators have raised copyright concerns with AI companies. It is difficult for these companies to prove that specific data was not included in their training process.
Zero-knowledge machine learning (ZKML) is an innovative approach that solves the trust issues associated with centralized artificial intelligence providers. By leveraging zero-knowledge proofs, ZKML enables developers to prove the correctness of their AI training and inference processes without revealing sensitive data or model details.
3.1 Training
Developers can perform training tasks in a zero-knowledge virtual machine (ZKVM), such as the one provided by Risc Zero. This process generates a proof that verifies that training was performed correctly and only authorized data was used. This certification serves as evidence that the developer adhered to appropriate training specifications and data usage permissions.
IOSG Ventures’ view:
- ZKML provides a unique solution for proving authorized data usage in model training, which is often difficult to achieve given the black-box nature of AI models.
*This technology is still in its early stages. The computational overhead is huge. The community is actively exploring more use cases for ZK training.
3.2 Reasoning
ZKML takes significantly longer to infer than its training counterpart. There are already several well-known companies emerging in this space, each with a unique approach to making machine learning inference trustless and transparent.
Giza is focused on building a comprehensive Machine Learning Operations (MLOP) platform and building a vibrant community around it. Their goal is to provide developers with tools and resources to integrate ZKML into inference workflows.
On the other hand, EZKL prioritizes the development experience by creating a user-friendly ZKML framework that provides good performance. Their solution aims to simplify the process of implementing ZKML reasoning, making it easy for more developers to use.
Modulus Labs takes a different approach, developing their own proof system. Their main goal is to significantly reduce the computational overhead associated with ZKML inference. By reducing the overhead by a factor of 10, Modulus Labs attempts to make ZKML inference more practical and efficient for real-world applications.
IOSG Ventures’ view:
- ZKML is particularly suitable for GameFi and DeFi scenarios, where trustlessness is crucial.
- The computational overhead introduced by ZKML makes it difficult to run large AI models efficiently.
- The industry is still looking for DeFi and GameFi pioneers to heavily use ZKML in their products to demonstrate its practical application scenarios.
4. Proxy network + other applications
4.1 Proxy network
An agent network consists of numerous artificial intelligence agents equipped with the tools and knowledge to perform specific tasks, such as assisting with on-chain transactions. These agents can collaborate with each other to achieve more complex goals. Several well-known companies are actively developing chatbot-like agents and agent networks.
Sleepless, Siya, Myshell, characterX, and Delysium are important players that are building chatbot agents. Autonolas and ChainML are building proxy networks for more powerful use cases.
IOSG Ventures’ take:
Agents are crucial for real-world applications. They can perform specific tasks better than general artificial intelligence. Blockchain offers several unique opportunities for artificial intelligence agents.
- Has an incentive mechanism: Blockchain provides an incentive mechanism through technologies such as non-fungible tokens (NFT). With clear ownership and incentive structures, creators are incentivized to develop more interesting and innovative agents on-chain.
- Composability of smart contracts: Smart contracts on the blockchain are highly composable and operate like Lego blocks. The open API provided by smart contracts enables agents to perform complex tasks that are difficult to achieve in traditional financial systems. This composability enables agents to interact with and leverage the functionality of a variety of decentralized applications (dApps).
- Intrinsic openness: By building agents on blockchains, they inherit the inherent openness and transparency of these networks. This creates significant opportunities for composability between different agents, allowing them to collaborate and combine their capabilities to solve more complex tasks.
4.2 Other Applications
In addition to the major categories discussed previously, there are several interesting AI applications that are receiving attention in the Web3 space, although they may not be large enough to form separate categories. These applications span a variety of fields and demonstrate the diversity and potential of artificial intelligence in the blockchain ecosystem.
- Image generation: ImgnAI
- Image prompt realization: NF
- Community-trained AI image generation: Botto
- Chatbots: Kaito, Supersight, Galaxy, Kn n3, Awesome QA, Qna3
- Finance: Numer AI
- Wallet: Dawn_wallet
- Game: Parallel TCG
- Education: Hooked
- Security: Forta
- DID：Worldcoin
- Creator Tools: Plai Lab
5. Promote AIxCrypto to Web2 users for mass adoption
AI x Crypto is unique because it solves the most difficult problems in artificial intelligence. Despite the gap between the current AIxCrypto product and Web2 AI products and its lack of appeal to Web2 users, AIxCrypto still has some unique features that only AIxCrypto can provide.
5.1 Cost-effective computing resources:
A major advantage of AIxCrypto is the provision of cost-effective computing resources. As the demand for LLM increases and there are more developers in the market, GPU availability and price become more challenging. GPU prices have increased significantly, and there are shortages.
Decentralized computing networks such as the DePIN project can help alleviate this problem by leveraging idle computing power, GPUs in small data centers, and personal computing devices. Although decentralized computing power may not be as stable as centralized cloud services, these networks provide cost-effective computing equipment in a variety of geographies. This decentralized approach minimizes edge latency, ensuring a more distributed and resilient infrastructure.
By leveraging the power of decentralized computing networks, AIxCrypto can provide Web2 users with affordable and accessible computing resources. This cost advantage is attractive for Web2 users to adopt AIxCrypto solutions, especially as demand for AI computing continues to grow.
5.2 Grant ownership to creators:
Another important benefit of AI x Crypto is the protection of creators’ proprietary rights. In the current field of artificial intelligence, some agents are easily copied. These agents can be easily replicated by simply writing similar prompts. Additionally, proxies in GPT stores are often owned by centralized companies rather than creators, limiting creators’ control over their works and their ability to monetize effectively.
AI x Crypto solves this problem by leveraging mature NFT technology that is ubiquitous in the crypto field. By representing agency as NFTs, creators can truly own their creations and receive actual revenue from them. Every time a user interacts with an agent, creators can earn incentives, ensuring a fair reward for their efforts. The concept of NFT-based ownership applies not only to agents, but can also be used to protect other important assets in the field of artificial intelligence, such as knowledge bases and tips.
5.3 Protect privacy and rebuild trust:
Users and creators have privacy concerns about centralized AI companies. Users worry about their data being misused to train future models, while creators worry about their work being used without proper attribution or compensation. Additionally, centralized AI companies may sacrifice service quality to reduce infrastructure costs.
These problems are difficult to solve with Web2 technology, and AIxCrypto leverages advanced Web3 solutions. Zero-knowledge training and inference provide transparency by proving the data used and ensuring the correct model is applied. Technologies such as Trusted Execution Environment (TEE), federated learning, and fully homomorphic encryption (FHE) enable secure, privacy-preserving AI training and inference.
By prioritizing privacy and transparency, AIxCrypto enables AI companies to regain public trust and provide AI services that respect user rights, setting them apart from traditional Web2 solutions.
5.3 Protect privacy and rebuild trust:
Users and creators have privacy concerns about centralized AI companies. Users worry about their data being misused to train future models, while creators worry about their work being used without proper attribution or compensation. Additionally, centralized AI companies may sacrifice service quality to reduce infrastructure costs.
These problems are difficult to solve with Web2 technology, and AIxCrypto leverages advanced Web3 solutions. Zero-knowledge training and inference provide transparency by proving the data used and ensuring the correct model is applied. Technologies such as Trusted Execution Environment (TEE), federated learning, and fully homomorphic encryption (FHE) enable secure, privacy-preserving AI training and inference.
By prioritizing privacy and transparency, AIxCrypto enables AI companies to regain public trust and provide AI services that respect user rights, setting them apart from traditional Web2 solutions.
5.4 Tracking content sources
As AI-generated content becomes increasingly sophisticated, it becomes more difficult to differentiate between human-authored and AI-generated text, images, or videos. To prevent misuse of AI-generated content, people need a reliable way to determine the source of the content.
Blockchain excels at tracing the provenance of content, just as it has done successfully in supply chain management and NFTs. In the supply chain industry, blockchain tracks the entire life cycle of a product, and users can identify the manufacturer and key milestones. Likewise, blockchain tracks creators and prevents piracy in the case of NFTs, which are particularly vulnerable to piracy due to their public nature. Despite this vulnerability, utilizing blockchain can minimize losses from fake NFTs because users can easily differentiate between real and fake tokens.
By applying blockchain technology to trace the origin of AI-generated content, AIxCrypto provides users with the ability to verify whether content creators are AI or human, thereby reducing the potential for abuse and increasing trust in the authenticity of the content.
5.5 Develop models using cryptocurrencies
Designing and training models, especially large models, is an expensive and time-consuming process. There is also uncertainty surrounding the new model, and developers cannot predict its performance.
Cryptocurrencies provide a developer-friendly way to collect pre-training data, gather reinforcement learning feedback, and raise funds from interested parties. This process is similar to the life cycle of a typical cryptocurrency project: raising funds through private investment or a takeoff platform, and distributing tokens to active contributors at launch.
Models can take a similar approach, raising funds for training by selling tokens and airdropping tokens to contributors of data and feedback. With a well-designed token economic model, this workflow helps individual developers train new models more easily than ever before.
6. Challenges of Tokennomics
AI x Crypto projects are starting to target Web2 developers as potential customers, as crypto has a unique value proposition and the Web2 AI industry has a sizeable market. However, tokens can be a barrier for Web2 developers who are unfamiliar with tokens and are reluctant to get involved in token-based systems.
In order to cater to Web2 developers, reducing or removing the utility of tokens may cause confusion for Web3 enthusiasts, because it may change the fundamental stance of the AI x Crypto project. When working to integrate valuable tokens into AI SaaS platforms, finding the balance between attracting Web2 developers and maintaining the utility of the token is a challenging task.
To bridge the gap between Web2 and Web3 business models while maintaining token value, there are several potential approaches that could be considered:
- Leverage tokens within the project’s distributed infrastructure network. Implement staking, reward and penalty mechanisms to protect the underlying network.
- Use tokens as a payment method and provide access for Web2 users
- Implement token-based governance
- Share profits with token holders
- Use proceeds to buy back or destroy tokens
- Services provided to the project, offering discounts and additional features to token holders
By carefully designing a token economic model that aligns with the interests of Web2 and Web3, the AI x Crypto project can successfully attract Web2 developers while maintaining the value and utility of its token.
7. Our favorite AI x Crypto scenarios
Our favorite AI x Crypto scenario leverages the power of user collaboration to accomplish tasks in the field of artificial intelligence with the help of blockchain technology. Some specific examples include:
-
Collective data contribution for AI training, alignment and benchmarking (such as Chatbot Arena)
-
Collaborate to build a large shared knowledge base that can be used by various agents (e.g., Sahara)
-
Use personal resources to capture network data (for example, Grass)
By leveraging the collective efforts of users based on blockchain incentives and coordination, these models demonstrate the potential of a decentralized, community-driven approach to AI development and deployment.
in conclusion
We are at the dawn of AI and Web3, and the integration of AI and blockchain is still in its early stages compared to other industries. Among the top 50 Gen AI products, there are no products related to Web3. The top LLM tools are related to content creation and editing, primarily for sales, meetings, and notes/knowledge bases. Considering the extensive research, documentation, sales, and community efforts in the Web3 ecosystem, there is huge potential for the development of custom LLM tools.
Currently, developers are focusing on building infrastructure to bring advanced AI models to the chain, although we are not there yet. As we continue to develop this infrastructure, we are also exploring the best user scenarios for conducting AI inference on-chain in a secure and trustless manner, which provides unique opportunities in the blockchain space. Other industries can directly use existing LLM infrastructure for inference and fine-tuning. Only the blockchain industry needs its own native AI infrastructure.
In the near future, we expect blockchain technology to leverage its peer-to-peer advantages to solve the most challenging problems in the artificial intelligence industry, making AI models more affordable, accessible, and profitable for everyone. We also expect the crypto space to follow the AI industry narrative, albeit with a slight delay. Over the past year, we have seen developers combine Crypto, proxy and LLM models. In the coming months, we may see more multi-modal models, text video generation, and 3D generation impact the crypto space.
The entire AI and Web3 industry has not received sufficient attention at present. We are eagerly looking forward to the explosive moment of AI in Web3, a killer application of CryptoxAI.
