ERC-8183: The answer to the AI Agent mutual trust challenge

Author:AzumaOdaily Daily Planet

 

On March 10, the dAI team under the umbrella of the Taifung Foundation, which focuses on promoting the integration of artificial intelligence (AI) with the depth of the block chain, and Virtuals Protocol, jointly launched a new standard ERC-8183。

ETHERWOOD FOUNDATION AI DIRECTORDavide CripisThe standard states that ERC-8183 is one of the missing components of the Open Agen economic system being constructed by the Taifeng community, and that it can be used in combination with x 402 and ERC-8004 to play an infrastructure role in the security interaction between Agent. The dAI team will support the adoption of ERC-8183 and work towards making it a neutral standard。

ERC-8183, WHAT ARE YOU TRYING TO SOLVE

According to the article published by Virtuals ProtocolERC-8183, designed specifically for commercial transactions between AI Agent, defines a set of chain rules that enable two mistrusted Agents to complete business processes such as “employment-delivery-settlement” without relying on a centralized platform。

The central question that ERC-8183 is trying to resolve is how can Agent conclude a transaction without a platform, without laws and without artificial arbitration when he hires and cooperates with one another

For example, if one market-oriented Agent A wants to hire another image-generated Agent B to make a series of marketing posters for it, there's a problem of mutual trust in the business -- no one knows each other, no basis for trust. If A pays first, B may strike or return unsatisfactory work results; if B works first, A may also refuse to pay..

IN THE TRADITIONAL INTERNET WORLD, USERS AND BUSINESSES WILL FACE SIMILAR COMMERCIAL MUTUAL TRUST, WITH THE PLATFORM ASSUMING A KEY INTERMEDIARY ROLE - THE PLATFORM WILL BE RESPONSIBLE FOR HOSTING A FUNDS, THE COMPLETION OF B SERVICES, AND THE FINAL LENDING. WE ARE FAMILIAR WITH TREASURE-HUNTING, KYOTO, CORPORATISM, DRIPPING, INHERENTLY PLATFORM-BASED INTERMEDIARIES。

And what the Taifung Foundation and Virtuals Protocol want to do is to use the ERC-8183 to abstract the platform ' s functions into a chain agreement, to make it executed by smart contracts, thereby assuming a decentrized intermediary role in the Agent economy。

ERC-8183 DISMANTLING OF THE WORK PROGRAMME

The operation mechanism of the ERC-8183 is not complex, and the standard introduces a new concept called Job. Each Job can be considered a complete commercial transaction with three different roles:

• Client“Customs”, simply to issue assignments, Agent
• Provider“Supplier” is Agent, who is responsible for carrying out the mission
• Evaluator“Assessors”, most specifically, is responsible for determining whether or not the task has been accomplished。

Here there is a need to focus on Evaluator, whose introduction is the ERC-8183 core design。In this standard, Evaluator is defined only as a chained address (address), but from a broader perspective, the address may correspond to a variety of different forms of execution。

• Evaluator could be an AI Agent for subjective tasks such as writing, designing or analysing, which would read the results submitted, compare them with the original mission requirements and then make a judgement
• Evaluator, on the other hand, can be a smart contract with a zero knowledge certifier (ZK verifier) for certain tasks such as computing, proving generation or data conversion. Provider submits proof that Evaluator authenticates the chain and automatically calls "complete" or "reject" to perform or reject the task
• In a high-value or high-risk mission scenario, Evaluator can also be a multiple-signed account, DAO, or certification cluster supported by a pledge mechanism。

ERC-8183 DOES NOT DISTINGUISH BETWEEN THESE DIFFERENT FORMS。The protocol layer cares only a little -- whether an address is called "complete" or "reject" , and the one that runs behind this address is an AI Agent driven by LLM, or a ZK circuit, which is not part of the protocol。

Moving on to Job, each Job has four states of life, which also corresponds to the different processes that ERC-8183 operates。

• Open: Clarent will create Job in this cycle, issue tasks and specify requirements
• Funded:Clint will transfer the commission to a smart contract hosting address instead of directly to Provider
• Submitted:Provider completes its work and provides proof
• TerminalEvaluator is responsible for the audit mission and, on the basis of the results, determines whether the mission was completed (Completed or Reprojected) and transfers the funds to Client or Provider, respectively; if no Provider responded or completed the task within the time requirement, the funds will be returned to Client。

In addition to the above standard processesERC-8183 can also achieve more derivative functions through modular extension, Hooks, to match complex business examples in the real world。Hooks is an optional smart contract attached to the creation of Job, which allows for the implementation of a self-defined logic around the life cycle of Job, such as credit thresholds, competitive mechanisms, cost allocation, or other special requirements。

What is the difference between ERC-8183 and x 402, ERC-8004

From x402 to ERC-8004, to today's ERC-8183, a less familiar reader may be in the fog and wonder why a new thing is to be done in a while. But in fact, these three are at three different stages of the AI Agent economic system, and the problems that we want to solve are different。

x402 is an HTTP payment agreement, which seeks to solve the problem of enabling AIAgent to make payments as directly as the API call; ERC-8004 is the AIAgent identity and reputation criterion, which deals with how to determine whether an Agent is reliable; ERC-8183 addresses the business transaction chain and tries to break the dilemma of how two untrusted Agents can complete the transaction。

If summed up in one sentence, x402 is responsible for solving “how to pay”; ERC-804 is responsible for knowing “who the other party is and how to rely”; and ERC-8183 is responsible for “how to trade safely”。

The three are not competitive, but complementary, and together they point to the same goal -- to create a decentralised, autonomous AI Agent economic system。