The Trust Road of AI: How the Mira Network Addresses AI Bias and Hallucination Issues

Recently, the public testnet of the Mira network has officially launched, aiming to build a trust layer for AI. This has sparked thoughts on the credibility of AI: why does AI need to be trusted? How will Mira address this issue?

In discussions about AI, people often focus more on its powerful capabilities while overlooking the issues of "hallucination" or bias that exist within AI. AI's "hallucination" refers to instances where AI may "make things up" and seemingly provide reasonable explanations for phenomena that do not exist. For example, when asked why the moon appears pink, it may offer a series of seemingly reasonable but actually unfounded explanations.

The "hallucinations" or biases of AI are 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. Moreover, errors, biases, and even fabricated content in the training data can affect AI's output. In short, AI learns human language patterns rather than the facts themselves.

The current probabilistic generation mechanisms and data-driven models almost inevitably lead to AI hallucinations. While this issue may not cause severe consequences in general knowledge or entertainment content, it can have significant impacts in highly rigorous fields such as healthcare, law, aviation, and finance. Therefore, addressing AI hallucinations and biases has become one of the core issues in the development process of AI.

The Mira project is dedicated to addressing issues of AI bias and hallucinations by building a trust layer for AI to enhance its reliability. Mira's core approach is to validate AI outputs through the consensus of multiple AI models. It is essentially a verification network that relies on the consensus of various AI models to validate the reliability of AI outputs. More importantly, Mira employs decentralized consensus for verification.

The key to the Mira network lies in decentralized consensus validation, which is precisely the expertise of the crypto space. At the same time, it also leverages the advantages of multi-model collaboration to reduce bias and hallucination through a collective verification model.

In terms of the verification architecture, the Mira protocol supports the conversion of complex content into independent verification statements. Node operators participate in the verification of these statements, ensuring the honesty of node operators through cryptoeconomic incentive/punishment mechanisms. Different AI models and decentralized node operators collaborate to ensure the reliability of the verification results.

Mira's network architecture includes content transformation, distributed validation, and consensus mechanisms. First, the system decomposes the candidate content submitted by clients into verifiable statements, which are then distributed to nodes for validation, and finally, the results are aggregated to reach a consensus. To protect client privacy, the statements are distributed to different nodes in a randomly sharded manner.

Node operators are responsible for running validator models, processing claims, and submitting verification results. Their motivation to participate in validation comes from the potential earnings, which are derived from the value created for clients. The Mira network aims to reduce the error rate of AI, particularly in fields such as healthcare, law, aviation, and finance, which will generate significant value. To prevent nodes from responding randomly, nodes that continuously deviate from consensus will have their staked tokens deducted.

Overall, Mira offers a new approach to achieving reliability in AI: building a decentralized consensus verification network based on multiple AI models, bringing higher reliability to clients' AI services, reducing AI bias and hallucinations, and meeting clients' demands for higher accuracy and precision. In short, Mira is constructing a trust layer for AI, which will promote the in-depth development of AI applications.

Currently, Mira has partnered with several AI agent frameworks. Users can participate in the Mira public testnet through Klok (a LLM chat application based on Mira), experience validated AI outputs, and have the opportunity to earn Mira points. The future uses of these points have not yet been announced, but they undoubtedly provide additional incentives for user participation.