This episode explores the convergence of artificial intelligence (AI) and cryptocurrency, focusing on the concept of "user-owned AI." Against the backdrop of current AI centralization within a few powerful companies, Illia Polosukhin, a co-founder of NEAR Protocol and co-author of the Transformers paper, argues that blockchain technology offers a decentralized solution. More significantly, he proposes a "confidential machine learning cloud" leveraging hardware advancements like Intel SGXs and NVIDIA H100s to enable secure, verifiable AI model execution without compromising user data privacy. For instance, developers can host their models on this cloud, charging users for access while retaining data ownership. As the discussion pivoted to practical applications, Polosukhin envisions an "agentic internet" where AI agents replace websites and apps, interacting with each other on behalf of users, automating complex tasks like travel planning. This vision, already partially realized through NEAR AI's platform, promises a future where AI assistants manage microtransactions and user interactions, creating a more efficient and user-centric online experience. What this means for the future is a shift towards decentralized, user-controlled AI, potentially reshaping the internet's architecture and addressing critical concerns around data privacy and national sovereignty.
