Artificial intelligence (AI) is rapidly evolving, yet two primary concerns persist: the reliability of outputs and the privacy of the underlying data. This dilemma is similarly prominent in blockchain technology, where users desire verification without sacrificing sensitive information or depending on centralized authorities.
Polyhedra, a cryptocurrency initiative developed by researchers from UC Berkeley, Stanford, and Tsinghua University, seeks to address these challenges. The team is transforming zero-knowledge (ZK) proofs into effective tools that can validate machine-learning results, facilitate asset transfers across over 25 blockchains, and protect user data without exposing confidential information.
Polyhedra’s offerings include Expander, a high-performance ZK prover, zkPyTorch for developers, and zkBridge for cross-chain communication. They are also launching EXPchain, a layer-1 blockchain specifically designed for AI applications.
Cointelegraph recently interviewed Tiancheng Xie, Polyhedra’s co-founder and CTO, to delve into the significance of verifiable privacy in AI and Web3. As a cryptographer with a Ph.D. from UC Berkeley, Xie aims to establish ZK as the “trust layer” essential for the forthcoming generation of decentralized and AI-driven applications.
According to Xie, the impetus for privacy-preserving infrastructure in AI comprises security, privacy, and trust, all bolstered by zero-knowledge proofs (ZKPs). As advancements in AI and blockchain occur, safeguarding sensitive data without compromising transparency becomes increasingly crucial. ZKPs allow for AI models to be trained and predictions made while ensuring confidentiality.
He emphasizes that ZKPs can verify computations or predictions of machine-learning models without exposing the data used, making them invaluable for sensitive applications. This process fosters accountability while upholding data privacy, which is vital in various scenarios.
Looking ahead, Xie argues that future zkML engineers will require a unique blend of skills in machine learning and cryptography. Polyhedra is nurturing talent through programs like the Explore Expander Bootcamp and Hackathon, which provide hands-on training and mentorship, helping engineers develop transformative zero-knowledge applications.
