The Trust Layer of AI: The Innovative Path of the Mira Network

Recently, a network public beta called Mira was officially launched, aiming to build a reliable foundation for artificial intelligence. The emergence of this project has sparked deep contemplation about the reliability of AI: why does AI need to be trusted? How does Mira address this issue?

When discussing AI, people often focus more on its powerful capabilities. However, the issues of "hallucinations" or biases in AI are often overlooked. The so-called "hallucinations" of AI simply refer to the fact that AI sometimes fabricates content that seems reasonable but is actually not true. For example, when asked why the moon is pink, AI might provide a series of seemingly reasonable but completely fictional explanations.

This phenomenon is closely related to the current path of AI technology. Generative AI achieves coherence and rationality by predicting the "most likely" content, but this method is difficult to verify for authenticity. In addition, the training data itself may contain errors, biases, or even fictional content, all of which can affect the quality of AI output. In other words, AI learns human language patterns rather than pure facts.

The current probability generation mechanisms and data-driven models almost inevitably lead to AI producing "hallucinations". While this may not cause serious consequences for general knowledge or entertainment content temporarily, in highly rigorous fields such as healthcare, law, aviation, and finance, AI errors could trigger significant issues. Therefore, addressing AI hallucinations and biases has become one of the core challenges in the development of AI.

The Mira project was born to solve this problem. It attempts to reduce AI bias and hallucinations and enhance its reliability by constructing a trust layer for AI. Mira's core approach is to validate AI outputs through the consensus of multiple AI models and verify them through a decentralized consensus mechanism.

In Mira's architecture, content is first converted into independently verifiable statements. These statements are verified by node operators in the network, ensuring the honesty of the verification process through cryptoeconomic incentives and penalty mechanisms. Multiple AI models and decentralized node operators participate together to ensure the reliability of the verification results.

The operational process of the Mira network includes content transformation, distributed verification, and consensus mechanism. The content submitted by clients is broken down into different verifiable statements, which are randomly assigned to different nodes for verification, and finally, the results are aggregated to reach a consensus. To protect client privacy, the content is distributed in a randomly sharded manner to prevent information leakage.

Node operators participate in network operations by running validator models, processing claims, and submitting validation results, earning profits from this. These profits come from the value created for clients, especially in reducing AI error rates. To prevent node operators from gaming the system, nodes that continuously deviate from consensus face the risk of their staked tokens being deducted.

Overall, Mira provides a new approach to achieving reliability in AI: building a decentralized consensus verification network based on multiple AI models to offer customers more reliable AI services, reduce AI bias and hallucinations, and meet the demands for high accuracy and precision. This not only creates value for customers but also brings benefits to network participants, promoting the in-depth development of AI applications.

Currently, users can participate in the Mira public testnet through the Klok app. Klok is an LLM chat application based on Mira, allowing users to experience verified AI outputs and have the opportunity to earn Mira points. Although the specific uses of these points have not been announced, this undoubtedly opens a new path for exploring the credibility of AI.