On June 7th, at the "2025 Web3 x AI Innovation Forum" jointly hosted by the Tsinghua University Student Blockchain Association and Polyhedra, Assistant Professor at Shanghai Jiao Tong University and Polyhedra co-founder Hu Yuncong delivered a keynote speech titled "Reshaping Trust Boundaries: A New Paradigm of Trusted Outsourced Computing Based on ZK"; Dr. Xie Tiancheng from UC Berkeley and Polyhedra co-founder & CTO delivered a keynote speech titled "AI+Web3: Prospects and Future of ZKML".
Among these, Xie Tiancheng mentioned: "The complexity of verifying liquidity and computation are incomparable. Specifically in AI model applications, the primary value of ZKML (Zero-Knowledge Machine Learning) is to verify service quality. For example, in a trading bot, users can verify the robot's operational quality through ZKML."
Secondly, with the growth of AI Agents and their sensitive permissions that can be authorized, ZKML can effectively ensure the safety of AI agents, meaning users can play the role of verifiers, building a trust system between humans and AI agents. Additionally, ZKML can provide data market functions, audit training data functions, and will play a crucial role in subsequent on-chain AI construction.
