Coin World reported:
Author: Tranks
Source: DeSpread Research
1. Introduction
With the development of the IT industry, enhanced computing power, and the widespread application of big data, the performance of artificial intelligence (AI) models has improved significantly. In recent years, AI capabilities have reached or even surpassed human levels in many fields and have been rapidly applied in industries such as healthcare, finance, and education.
A typical case of AI commercialization is ChatGPT, a generative AI model launched by OpenAI in November 2022, which can understand and respond to human natural language. ChatGPT attracted 1 million users just 5 days after its launch and reached 100 million monthly active users within 2 months, becoming the fastest-growing consumer application in history.
NVIDIA, which designs and manufactures the GPUs needed for major AI platforms, has also greatly benefited from this trend. In the first quarter of 2024, NVIDIA’s net profit increased by 628% year-on-year to reach $14.8 billion, with its stock price rising nearly threefold compared to last year, bringing its market capitalization to $3.2 trillion, showcasing exceptional performance.
The rise of the AI industry has had a significant impact on the cryptocurrency market. In June 2022, during the peak of the NFT art project boom, OpenAI released its AI model DALL-E 2, which can generate high-quality images from text, resulting in an 8-fold increase in the mention of AI keywords within major cryptocurrency Telegram channels in South Korea. Moreover, starting from the second half of 2022, there have been increasing attempts to combine AI and blockchain more directly, leading to a further 2-fold increase in mentions of AI.
The cryptocurrency community’s strong interest in AI is also reflected in the investment trends of AI-related crypto projects. According to data from the virtual asset statistics website Coingecko, as of August 20, 2024, the total market capitalization of 277 blockchain projects classified under the AI sector has rapidly grown to $21 billion since the emergence of AI and blockchain projects in the second half of 2022, which is approximately 25% higher than that of Layer 2 projects.
However, the currently highlighted blockchain projects in the AI field mainly utilize blockchain technology to address the limitations exposed during the development process of the AI industry. The main application scenarios include:
Distributed GPU Networks: These projects use blockchain technology to create a distributed GPU network where anyone can contribute GPU computing power and earn token rewards, thereby lowering the entry barriers posed by the high costs of GPU required for AI model training (e.g., IO.NET, Akash Network).
Decentralized AI Training and Model Development: These projects allow multiple participants to collaboratively engage in AI training and model development, rewarding them with tokens through blockchain technology, aiming to address AI bias issues caused by centralized AI development environments (e.g., Bittensor).
On-chain AI Markets: These decentralized AI market projects leverage blockchain technology to transparently evaluate and trade the performance and reliability of AI models or agents, catering to the demands of various industries and specific functions for AI models or agents (e.g., SingularityNET, Autonolas).
In addition to the examples mentioned above, many new attempts are emerging, utilizing blockchain infrastructure such as decentralized data markets and IP protocols to tackle the challenges currently faced by the AI industry. These attempts are creating synergies by providing a more stable infrastructure for the AI industry and expanding the application scope of blockchain technology.
At the same time, integrating AI into the blockchain ecosystem holds infinite development potential. Particularly in permissionless DeFi services, introducing AI can reduce reliance on trusted third parties, enabling functionalities that many existing smart contracts struggle to achieve.
In this article, we will explore specific use cases of AI in current DeFi protocols, the challenges faced, and the future development directions of AI in DeFi.
2. Intelligent DeFi
AI possesses excellent real-time data analysis capabilities, allowing it to draw conclusions from large volumes of data. This characteristic plays a crucial role in helping users execute financial operations and manage risks, providing vital data on the returns and risks associated with specific DeFi protocols. In this context, AI is primarily applied to the user interface of Dapps, enabling existing DeFi protocols to leverage AI without requiring substantial structural adjustments.
Yearn Finance is a typical example, serving as a yield aggregator. In order to provide users with a safer investment environment, Yearn Finance is collaborating with the AI agent platform GIZA to establish a real-time strategy risk assessment system for its v3 vaults.
However, I am more focused on the potential for DeFi protocols to gain autonomy through the integration of AI’s ability for independent thinking and action.
Current DeFi protocols typically respond passively to user transactions, meaning that the smart contracts within the protocols operate in a predetermined manner based on user interactions. However, by incorporating AI into DeFi protocols, these protocols can autonomously analyze market conditions, make optimal decisions, and proactively generate trades. This makes it possible for DeFi protocols to offer innovative financial services that were previously difficult to realize.
Let us specifically examine some intelligent DeFi protocols that apply AI in their primary operational mechanisms.
2.1. Fyde Treasury: AI Token Fund
Fyde Treasury is a protocol that offers a basket fund service called Liquid Vault, which operates multiple tokens and is managed by AI. Users can receive and utilize liquidity tokens $TRSY corresponding to the assets deposited into the Liquid Vault.
2.1.1. Asset Selection and Fund Operation
The core mission of Liquid Vault is to increase the proportion of low-volatility tokens during market downturns, thereby providing users with a lower loss rate and outperforming other asset classes in the long term.
Fyde Treasury selects assets for inclusion in the Liquid Vault investment portfolio through the following three steps:
1. Assessing whether trading liquidity is sufficient
2. Checking the background of the protocol founders and the audit status of the protocol code to identify any issues
3. Analyzing on-chain data with AI to evaluate the existence of wash trading, token concentration, and natural growth trends
Tokens that meet these criteria will be included in the Liquid Vault investment portfolio. Furthermore, Fyde Treasury utilizes AI during the asset management process of the Liquid Vault, specifically including:
Market Analysis and Forecasting: Analyzing on-chain transaction data, market trends, and news to predict future market movements
Weight Calculation and Rebalancing: Calculating the optimal token weights and rebalancing based on predicted market trends and the recent performance and volatility of tokens in the portfolio
Risk Management and Response: Rapidly identifying governance attacks, liquidity pool depletion, and abnormal transactions in specific wallets for each token in the portfolio, and adjusting the portfolio or isolating related tokens in a timely manner
Advanced Asset Management Strategies: Continuously assessing the performance of the portfolio, analyzing the effectiveness of strategies, and extracting data to modify and develop new strategies. Existing strategies are then compared against new strategies to measure their performance and applied to actual operational strategies
As of the writing date, August 23, there are 29 tokens in the Liquid Vault portfolio, all of which are various industry tokens based on the Ethereum network.
Liquid Vault Dashboard, Source: Fyde
Additionally, Fyde Treasury provides a feature that allows users who deposit specific protocol governance tokens into the Liquid Vault to maintain their governance voting rights through liquidity tokens. The governance tokens deposited into the Liquid Vault are sent to their wallets in the form of $gTRSY-token, which can be used to execute governance votes for the corresponding protocols in Fyde Treasury’s governance tab.
However, voting rights are affected by the token weights in the portfolio, so they may change with each adjustment of the portfolio.
2.1.2. Liquidity Mining Activities
Fyde Treasury rewards liquidity providers who enhance the market liquidity of $TRSY (Liquid Vault liquidity token) with Fyde points and promises to distribute its governance token $FYDE based on these points in the future.
Unlike other projects that typically require users to deposit trading pairs directly into decentralized exchanges to obtain tokens or points through liquidity mining activities, Fyde Treasury allows users to deposit $FYDE into the liquidity mining contract within the protocol and directly provide liquidity on Uniswap v3. Uniswap v3 is a decentralized exchange that allows users to set supply ranges when providing liquidity.
When providing liquidity on Uniswap v3, the system calculates and executes the optimal path to convert part of the $FYDE deposited into the liquidity mining contract into $ETH through an AI-driven simulation environment. Moreover, AI manages and optimizes the liquidity deposit ranges on Uniswap v3 in real time based on market conditions, achieving approximately 4 times the capital efficiency compared to providing liquidity with the same capital on general decentralized exchanges.
AI Simulation Dashboard, Source: Fyde Docs
In this manner, Fyde Treasury is building a basket fund. This fund utilizes AI for real-time management of the assets deposited by users in the protocol, thereby reducing human judgment and mitigating various risks in the market.
2.1.3. Protocol Performance
Since its launch in January 2024, Fyde Treasury’s TVL has steadily grown to approximately $2 million. However, due to the sustained weakness in the market since late May, the return rate of the $TRSY token has been -35% over the past three months.
However, compared to other major tokens in the Ethereum ecosystem, the price volatility of $TRSY has been relatively stable, with a smaller decline.
Although Fyde Treasury has been launched for less than a year, its AI model has continuously learned and evolved through market data. With the accumulation and optimization of AI learning, better performance may be expected in the future, making it worthwhile to pay attention to the future development directions and performance of Fyde Treasury.
2.2. Mozaic Finance: AI Yield Optimizer
Mozaic Finance is a yield optimization protocol that uses AI to optimize yield farming strategies through specific DeFi protocols. It provides users with various asset management strategies in the DeFi ecosystem, presented in the form of vaults, and utilizes two types of AI for strategy optimization:
Conon: Real-time analysis of on-chain data to predict market conditions and changes in APY of yield farming strategies
Archimedes: Calculating the optimal investment strategy based on the predictive data from Conon and executing fund allocation
In Mozaic Finance, the AI agent Conon serves as the “analyst,” while Archimedes acts as the “strategist,” working together to manage the assets deposited by users.
2.2.1. Vault Types
Hercules: This is a vault that uses stablecoins for yield farming, where depositors receive MOZ-HER-LP tokens as liquidity tokens.
The assets deposited by users in the vault are used to provide liquidity and generate returns through the bridge protocol Stargate. AI will real-time bridge the vault assets and rebalance them to liquidity pools with higher yields. The characteristic of Stargate is that even the same asset may have different APYs across different networks due to liquidity disparities.
Stargate Farm Dashboard, Source: Stargate
Theseus: This is a vault that generates returns through various volatile assets, where depositors receive MOZ-THE-LP tokens as liquidity tokens.
Users’ assets are deposited into the GM pool of the GMX protocol, a decentralized perpetual futures exchange that provides liquidity to traders and earns incentives. When deploying liquidity, the volatility and interest rates of the trading assets in each GM pool are considered. Depending on market conditions, the proportion of stablecoins may be increased and deposited into Stargate to generate additional interest.
GMX GM Pool Dashboard, Source: GMX
Perseus: This is a vault that actively utilizes the PoL (Proof of Liquidity) consensus mechanism to earn network rewards by providing liquidity for the ecosystem protocols of the upcoming mainnet Berachain. The Mozaic Finance team is developing and preparing to launch a strategy using the Berachain testnet, with more details to be announced later.
For more information on Berachain and the PoL consensus mechanism, please refer to the article “Berachain — The Bear Catching Two Rabbits: Liquidity and Security.”Mozaic Finance is a protocol that optimizes liquidity supply strategies and processes, manages risks, and employs AI when users deposit assets into DeFi protocols.
As of January 2024, the Hercules and Theseus treasuries have performed well, with expected APYs of approximately 11% and 50% respectively. However, both treasuries are currently suspended due to a funds theft incident at Mozaic Finance.
On March 15, 2024, Mozaic Finance experienced a funds theft incident while transitioning to a new security solution developed by Hypernative to enhance on-chain risk and security. Before the security update was completed, an internal developer discovered that the treasury funds could be stolen by using the private key of a core team member. They invaded the member’s computer, obtained the private key, and stole approximately $2 million worth of treasury assets, which were then liquidated on a centralized exchange.
In response to the incident, Mozaic Finance team suspended the operations of the Hercules and Theseus treasuries. The value of the governance and protocol fee collection token $MOZ dropped by around 80%. The team immediately transparently disclosed the progress of the incident and collaborated with a security company to track the stolen assets. They also applied for the freezing and return of the stolen funds from the centralized exchange where the developer had stored them, striving to restore the protocol’s normal operation.
Fortunately, the process of returning all stolen funds is currently underway. While waiting for the stolen funds to be returned from the centralized exchange, the team is preparing to launch Mozaic 2.0. The new version includes the following improvements:
1. Enhanced Security: Code audits and security enhancements through Trust Security, Testmachine, and Hypernative, among other security professional companies.
2. AI Model Improvements: Comprehensive upgrade of the existing Archimedes model and prediction and learning of yet-to-happen black swan events based on expert knowledge. In addition, detection of abnormal decisions and the establishment of flags for manual review and model improvement.
3. Improved User Experience: Improvement of the Dapp’s UI/UX and enhanced user access to the Dapp in various chain environments through account abstraction and bridging services.
Therefore, despite experiencing a major funds theft crisis, Mozaic Finance is actively preparing to launch Mozaic 2.0, aiming to provide users with more secure and efficient asset management services.
Challenges: The Decentralization and Scalability Dilemma of AI
Through the cases of Fyde Treasury and Mozaic Finance, we have learned how AI can be a core component of intelligent DeFi protocols. The advantages that AI brings to intelligent DeFi protocols include:
1. Establishing new DeFi protocol models through autonomy.
2. Improving capital efficiency by analyzing and optimizing fund operations.
3. Real-time analysis and response to risks such as abnormal transactions.
Currently, the integration of blockchain and AI mostly focuses on building blockchain infrastructure to overcome the limitations of AI. However, given the aforementioned advantages, it is expected that more attempts will be made to incorporate AI into DeFi protocols. Of course, in the process of integrating these two fields, there are challenges that need to be addressed.
AI requires an environment that can handle large amounts of data quickly, but the current blockchain infrastructure cannot achieve this level of data processing speed. For example, the ChatGPT-3 model is estimated to require processing trillions of data per second to answer questions, which is about ten million times faster than Solana’s maximum TPS (transactions per second) of 65,000.
Furthermore, even if the blockchain infrastructure develops to support AI computations, the transparency of public blockchains may expose the training data and decision weights of AI models to the public. This means that transactions generated by AI may become predictable and face various external attack risks.
Therefore, DeFi protocols like Fyde Treasury and Mozaic Finance, which aim to leverage AI, currently choose to run AI on centralized servers and interact with the blockchain based on the results. However, this approach requires users to trust the honesty of the team responsible for managing the AI when depositing assets into the protocol. This weakens the core principle of DeFi, which is to eliminate the need for trusted third parties through smart contracts to provide a trustless trading environment.
The challenges of decentralization and scalability in applying AI to blockchain are considered as challenges that DeFi applications must solve when utilizing AI. zkML (zero-knowledge machine learning) technology is emerging as a solution to address these challenges.
zkML is a technology that combines zero-knowledge proofs (ZKP) with machine learning (ML). Zero-knowledge proofs are encryption methods that can verify the authenticity of data without revealing the data itself, thereby achieving privacy protection and data integrity verification. zkML utilizes these characteristics of zero-knowledge proofs and applies them to the field of machine learning, enabling the verification of model outputs without disclosing inputs, parameters, and internal mechanisms of AI models.
In addition, by designing smart contracts of DeFi protocols to verify zero-knowledge proofs, on-chain transactions are only generated when AI models operate honestly as expected and without external interference, thus securely integrating AI into DeFi protocols.
For example, Mozaic Finance, mentioned earlier, plans to introduce zero-knowledge proof technology into its protocol in the future. They state in their documentation that this technology will enhance the real-time verification of Archimedes’ honest decisions and treasury management.
However, zero-knowledge proof technology is still in its early stages and requires extensive discussion and development for practical applications. Especially for complex AI models, generating zero-knowledge proofs, although more efficient than executing AI models directly on the blockchain, still requires computational power and storage space beyond the current blockchain infrastructure’s capabilities. Therefore, to make zkML truly practical, further technological advancements and optimizations in zero-knowledge proofs and blockchain infrastructure are necessary.
Conclusion
In this article, we have explored the emerging service protocols that combine AI and blockchain technology, the challenges faced by these protocols, and the future of AI agent-based blockchain ecosystems.
In the future, AI and blockchain technologies will continue to develop and integrate, overcoming each other’s limitations. Through this integration, it is expected to provide individuals with a more convenient environment to access and utilize AI and blockchain technology.
Especially in future AI agent-based on-chain economic ecosystems, people will be able to easily use and provide financial services without requiring deep financial knowledge. This will contribute to significantly improving the liquidity of on-chain ecosystems and enhancing the inclusiveness of the financial industry.
Furthermore, AI and blockchain not only have the potential to influence each other but also have the potential to become the infrastructure of various industries. Therefore, the development of these technologies will have profound effects on the entire human society, not limited to individual industries.
However, AI-related regulations such as data privacy protection and AI accountability issues, as well as blockchain-related regulations such as the security attributes of tokens, will have a significant impact on the future development direction and industry structure of these technologies. Therefore, we need to closely monitor the upcoming AI and blockchain industry regulations.
Ultimately, we hope that the development of these technologies will create a better environment for humanity and help address many societal issues.
