NEAR Co-Founder Illia Polosukhin Presents New AI Research at NVIDIA GTC 2025

This morning, at NVIDIA’s GTC AI Conference in San Jose, California, NEAR co-founder Illia Polosukhin presented new research from NEAR AI to enable confidential, decentralized AI computation. This work provides a technical basis for an AI economy that allows users to harness powerful, customized intelligence without relying on—or giving up data to—the closed, proprietary models of a few massive tech companies.

Leading the Conversation on AI and Blockchain

This was Illia’s second consecutive year speaking at the flagship event of the world leader in AI computing. As one of the co-authors of “Attention Is All You Need,” the landmark 2017 paper introducing the Transformer architecture at the core of the ongoing leap in AI capability, Illia is a key leader in the AI world and is now at the forefront of making AI truly open and user-owned with NEAR. Indeed, Illia is the only Web3 founder speaking at GTC 2025, highlighting NEAR’s unique position at the intersection of blockchain technology and artificial intelligence.

Decentralized Confidential Machine Learning: Enabling User-Owned AI

Illia’s talk unpacked NEAR AI’s recent developments in decentralized confidential machine learning, a cloud approach to training, fine-tuning, and utilizing AI models and agents that addresses several of today’s critical challenges:

• Privacy: Users can use AI models without revealing their personal data, even to the entities hosting the models.
• Open Source: Users can trust these models to be open, transparent, and verifiable.
• Sustainable Economics: Model creators can monetize their work, even in a decentralized environment, ensuring a sustainable business model.
• Community Collaboration: Groups can collectively develop AI models, sharing both costs and profits without requiring absolute trust among participants.

This approach represents a significant departure from the present AI paradigm, in which a handful of powerful corporations own the models, farm user-generated data, and capture all resulting value––or, in the case of open source AI, lack a path to sustainable monetization.

NEAR: The Blockchain for AI

AI isn’t just the future—it’s happening now. But to truly realize its transformative potential, economic and otherwise, AI must be open for all to participate and profit while preserving ownership and privacy of data.

Blockchain technology provides a means to accomplish this, and NEAR is the only platform purpose-built to power AI at scale while preserving privacy and user ownership. Unlike traditional blockchains, NEAR enables intelligent agents to transact, automate workflows, and operate independently while preserving privacy, security, and cross-network interoperability.

Advances like the ones Illia described in this week’s NVIDIA talk are a major part of why the future of AI is NEAR. To learn more about NEAR AI’s research and Decentralized Confidential Machine Learning, visit near.ai.