Consensus Mechanism : PoAW (Proof of AI Work)

EMC adopts a Proof of AI Work (PoAW) consensus mechanism that relies on tokenized AI Agents running on a public chain to validate transactions and maintain network security. Its primary objective is to enhance the real-world efficiency of blockchain applications while promoting the decentralized nature of AI.

1. Layered Architecture

  • Execution Layer An EVM-compatible smart contract engine that supports AI task scheduling contracts.

  • Consensus Layer A modified Istanbul BFT (IBFT) consensus mechanism integrated with an AI workload verification module.

  • Storage Layer A hybrid storage framework, combining on-chain metadata with IPFS/Filecoin for off-chain data storage.

  • Cross-Chain Layer A bridging protocol that enables interoperability of assets across multiple blockchains.

2. Core Components

  • Tokenized AI Agent Each AI Agent represents a certain amount of tokenized assets or hardware resources. These AI Agents are created by EMC public chain’s computing nodes. Users can stake EMC tokens to activate an AI Agent.

  • Smart Contracts AI Agents interact with the blockchain via smart contracts, executing complex computations and making decisions on-chain.

  • AI Agent Management Contract Handles staking, activation, and weight calculation for each AI Agent.

  • Oracle Network Provides external data feeds and verification services for on-chain operations.

  • Task Validation Nodes Execute Byzantine fault-tolerant verification of AI task results.

3. Staking Rules

  • Minimum Staking Amount 3,000 EMC (dynamically adjustable).

  • Weight Calculation Formula

 Staked AmountBase Threshold×Number of TasksTotal AI Agent Stake\ \sqrt{\frac{\text{Staked Amount}}{\text{Base Threshold}}} \times \frac{\text{Number of Tasks}}{\text{Total AI Agent Stake}}
  • Staking Lock-Up Period Tied to the task execution cycle, with a minimum of 90 days.

  • Reward Mechanism AI Agents that successfully participate in consensus are rewarded based on their weight, incentivizing ongoing participation in network maintenance.

4. Workload Verification Process

  1. Transaction Submission A user submits a transaction request, which is received and verified by an AI Agent.

  2. Computation Verification The AI Agent computes and verifies the validity of the transaction, generating a hash fingerprint of the result.

  3. Data Verification The hash fingerprint is stored in a Merkle Patricia Tree structure for tamper-proof verification.

  4. Result Broadcasting The hash fingerprint is broadcast across the network and undergoes IBFT-based validation by the task validation nodes.

  5. Consensus Once the verification threshold of two-thirds is reached, the transaction is recorded on the blockchain.

5. Security and Decentralization of the Workload Verification Process

  • Diversity Encourages participation from various types of AI Agents to enhance network diversity and security.

  • Dynamic Adjustments Staking requirements and reward mechanisms are dynamically adjusted based on network conditions to ensure stability over time.

Last updated