Source: Cointelegraph
Original: “How Decentralized Finance (DeFi) Can Achieve Safe Scalable Development in the Age of Artificial Intelligence (AI)”
Views from: Karan Sirdesai, Co-founder and CEO of Mira Network
In all smart contract blockchains, when decentralized finance (DeFi) transactions enter the memory pool, it is like stepping into a "dark forest"—anyone can peek and take advantage of it. The most well-equipped and informed users often profit from this.
For the industry, this arbitrage motivation is known as miner extractable value (MEV). MEV bots have emerged, operating under simple preset rules with very low flexibility. To date, these rule-based bots have profited millions of dollars from Ethereum transactions. In the new era of AI and agent technology, this number is expected to grow exponentially—along with the accompanying opportunities and risks.
It is estimated that by the end of 2025, over 50% of on-chain transaction volume will be driven by various AI agents, whose functions will far exceed those of current bots. The launch pools of DeFi AI (DeFAI) platforms such as HeyAnon and Griffain have already shown significant spikes in usage and token prices.
The Role of AI Agents in DeFi
We are in the nascent stage of AI agents, and industry participants are assessing the potential of these new tools. The current research and development phase has observed that AI possesses at least three precise functions in DeFi:
Just as ChatGPT has driven the mainstreaming of AI by lowering the usage threshold, AI agents will shorten the learning curve, making DeFi more accessible to the public. Clearly, the first beneficiaries will be industry professionals, who are always the earliest hunters equipped with new weapons in the "dark forest."
MEV agents and market-making agents are about to emerge. Looking to the future, all AI optimists hope that agents can evolve to have independent goals and decision-making capabilities, completely free from human intervention.
Given the potential impact and high-risk characteristics of AI agents in DeFi, the security risks they bring cannot be ignored. No one wants to repeat the AI black swan event like the Terra/LUNA collapse.
The Illusion Problem
Known security risks introduced by AI agents include "illusions"—the phenomenon where large language models (LLMs) produce absurd or erroneous outputs. Imagine if AI misreports high fees or incorrect contract addresses while executing a simple exchange in DeFi, leading to significant user losses.
In addition to accidental illusions, some AIs may be manipulated by hackers to perform unauthorized operations. In a censorship-resistant DeFi environment, this risk will directly translate into user fund losses.
The proliferation of AI agents in DeFi may amplify systemic risks across the entire crypto market, making "illusions" a key bottleneck in the development of DeFAI.
Strategies for DeFAI Builders
In response to the real challenge of AI illusions, there are various existing research solutions. If focusing on the DeFAI field, builders should consider the following strategies:
Establish a robust rule set: Detailed and precise operational rules must be formulated for agents to reduce the risk of errors, but flexibility must be balanced—overly rigid rule sets will weaken the agents' ability to adapt to new information.
Implement extreme stress testing: It is necessary to simulate massive interaction scenarios for testing, clarifying the capability boundaries of agents through data, and setting reasonable goals.
Deploy protective systems: The best solution is to use real-time verification systems, conducting automated stress tests through multiple LLMs, and dynamically assessing each operation before execution.
Safe DeFAI Will Ultimately Be Achieved
The potential of AI agents in DeFi is just beginning to emerge. Their ability to be on standby 24/7, make autonomous decisions, and execute massive transactions in milliseconds is a deadly attraction for any trader. Many predictions indicate that DeFAI will soon handle the majority of on-chain transactions.
To realize this vision, it is essential to have a clear understanding of the current limitations of LLMs and to tackle the issues of illusions and biases to achieve the safe scaling of DeFAI. With the continuous evolution of AI models and the collaborative efforts of builders, the integration of agents and blockchain will come naturally.
Views from: Karan Sirdesai, Co-founder and CEO of Mira Network
Related: Blockchain Interoperability Will Accelerate Institutional Success
This article is for informational purposes only and does not constitute any legal or investment advice. The views expressed are solely those of the author and do not necessarily reflect the position of Cointelegraph.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
