This project utilizes the ezkl to create proof of inference (proof of computation, proof that AI Model computed with X input and Y output). We have created a Near Protocol oracle and started working on a verifier for Near Protocol to make ezkl work there. You can use the frontend to select any model and generate a proof, then validate the data, output, and computation, and use the oracle to get the proof on-chain.
Using ZKML makes using AI on-chain SAFE. Instead of common-practice Oracles, we create zk proof and verify on-chain models run.
Imagine the risks that DEX/DEFI could have without ZKML while using Oracles, hundreds of billions of dollars each year.
We bring ZKML to Near Protocol, and we are looking for support to make it happen.