DeFAI: A New Path to Intent

DeFAI is AI+DeFi in a succinct way. The market has hyped AI over and over again, from AI computing power to AI Meme

1. DeFAI tells the story of AI+DeFi

1.1 What is DeFAI

DeFAI is a concise way of saying AI+DeFi. The market has hyped AI over and over again, from AI computing power to AI Meme, from different technical architectures to different infrastructures. Although the overall market value of AI Agents has generally declined recently, the concept of DeFAI is becoming a new breakthrough trend. The current DeFAI can be roughly divided into three categories: AI abstraction, autonomous DeFi agent, and market analysis and prediction. The specific divisions within the categories are shown in the figure below.

1.2 How DeFAI works

In the DeFi system, the core behind AI Agent is LLM (Large Language Model), which involves multi-level processes and technologies, covering all aspects from data collection to decision execution. According to the research of @3sigma in the IOSG article, most models follow the data. The six specific workflows of collection, model reasoning, decision making, hosting and operation, interoperability, and wallet are summarized below:

1. Data Collection: The first task for an AI Agent is to gain a comprehensive understanding of the environment in which it operates. This includes acquiring real-time data from multiple sources:

On-chain data: Obtain real-time blockchain data such as transaction records, smart contract status, and network activities through indexers, oracles, etc. This helps Agents keep in sync with market dynamics;

Off-chain data: Obtain price information, market news, and macroeconomic indicators from external data providers (such as CoinMarketCap, Coingecko) to ensure that the Agent understands the external conditions of the market. This data is usually provided to the Agent through an API interface;

Decentralized data source: Some agents may obtain price oracle data through a decentralized data feed protocol to ensure the decentralization and credibility of the data.

2. Model Reasoning: After data collection is completed, the AI Agent enters the reasoning and calculation phase. Here, the Agent relies on multiple AI models for complex reasoning and prediction:

Supervised and unsupervised learning: AI models can analyze the behavior of markets and governance forums by training on labeled or unlabeled data. For example, they can predict future market trends by analyzing historical trading data, or predict future market trends by analyzing governance forum data. , speculate the outcome of a voting proposal;

Reinforcement learning: Through trial and error and feedback mechanisms, AI models can autonomously optimize strategies. For example, in token trading, AI Agents can determine the best time to buy or sell by simulating multiple trading strategies. This learning method allows Agent to continuously improve under changing market conditions;

Natural Language Processing (NLP): By understanding and processing user natural language input, Agents can extract key information from governance proposals or market discussions to help users make better decisions. This is especially useful in scanning decentralized governance forums or processing user instructions. It is especially important when.

3. Decision making: Based on the collected data and the results of reasoning, the AI Agent enters the decision-making stage. In this stage, the Agent not only needs to analyze the current market situation, but also make trade-offs between multiple variables:

Optimization Engine: Agent uses the optimization engine to find the best execution plan under various conditions. For example, when performing liquidity provision or arbitrage strategies, Agent must consider factors such as slippage, transaction fees, network latency, fund size, etc. in order to find The optimal execution path;

Multi-agent system collaboration: In order to cope with complex market conditions, a single agent sometimes cannot fully optimize all decisions. In this case, multiple AI agents can be deployed, each focusing on different task areas, to improve the overall system's decision-making through collaboration. Efficiency. For example, one agent focuses on market analysis and another agent focuses on executing trading strategies.

4. Hosting and operation: Since AI Agent needs to process a lot of calculations, it is usually necessary to host its model on an off-chain server or distributed computing network:

Centralized hosting: Some AI agents may rely on centralized cloud computing services such as AWS to host their computing and storage needs. This approach helps ensure the efficient operation of the model, but it also brings potential risks of centralization;

Decentralized hosting: To reduce the risk of centralization, some agents use decentralized distributed computing networks (such as Akash) and distributed storage solutions (such as Arweave) to host models and data. Such solutions ensure the model Decentralized operation while providing persistence of data storage;

On-chain interaction: Although the model itself is hosted off-chain, the AI Agent needs to interact with the on-chain protocol in order to perform smart contract functions (such as transaction execution, liquidity management) and manage assets. This requires secure key management and transactions. Signing mechanisms such as MPC (Multi-Party Computation) wallets or smart contract wallets.

5. Interoperability: The key role of AI Agent in the DeFi ecosystem is to interact seamlessly with