Onchain AI Infernece: DPrompt

At the heart of the REVOX ecosystem, DPrompt emerges as the pivotal component that enables on-chain AI inferences for all agents within the network. Designed to integrate effortlessly with smart contracts, DPrompt facilitates the convergence of AI capabilities with blockchain technology, offering a robust foundation for developing dApps. REVOX introduces two solutions to address the challenges associated with on-chain AI computation, ensuring both integrity and efficiency in AI service provision:

1. Proof-of-Stake Off-Chain Computation Drawing inspiration from oracle networks, this model entails the logging of AI requests on the blockchain, which are then addressed by designated AI providers. These providers, required to stake REVOX tokens as a guarantee of their reliability, pick up requests and return responses to DPrompt’s smart contract. The staked tokens serve as a form of security, with the accuracy and integrity of computation results subject to challenge and verification through reruns. While this approach introduces the possibility of overhead due to AI reruns, it stands as a pragmatic initial solution, leveraging token staking to ensure the fidelity of AI providers.

2. Proof-Based Machine Learning with Efficient Challenge In pursuit of more advanced verification mechanisms, REVOX explores cutting-edge research in generating proofs for machine learning computations. Concepts such as ZKml (Zero-Knowledge Machine Learning) and OPml (Optimistic Proofs for Machine Learning) represent promising directions for the ecosystem. These methodologies focus on creating proofs that allow for efficient challenges or verifications of AI computations, minimizing the need for resource-intensive reruns while enhancing trust and integrity in the system’s AI outputs.

The Dual Approach to Computation Verification

Through these two solutions, REVOX addresses the dual challenge of maintaining computational integrity and minimizing operational overhead in AI-based blockchain applications. The Proof-of-Stake off-chain computation model provides a solid starting point, leveraging the economic security model of token staking. Meanwhile, the exploration of proof-based machine learning methodologies signals REVOX’s commitment to advancing the frontier of efficient and verifiable AI computations on the blockchain.