The Stack Overflow Podcast Attention isn’t all we need; we need ownership too
18 snips
Jul 8, 2025 Ilya Polosukhin, co-author of the original Transformers paper and co-founder of NEAR, shares his journey from Ukraine to revolutionizing AI and blockchain. He discusses the evolution of machine learning and the need for decentralized control to foster innovation. Ilya highlights practical blockchain applications in stablecoin transactions and the vision for user-owned AI, which promotes accessibility while ensuring economic viability. The conversation also explores AI-driven commerce's future, shifting towards personalized decision-making powered by scalable blockchain technology.
AI Snips
Chapters
Transcript
Episode notes
Ilya's Early AI Journey
- Ilya Polosukhin started programming at age 10 and was fascinated by gaming and machine learning.
- He eventually co-authored the Transformer paper after prototyping a model that reads entire documents at once, removing recurrence.
AI Progress Is Incremental
- Progress in AI has been slower than expected after deep learning breakthroughs.
- New improvements are incremental tweaks applied to accumulated advances, not radical jumps.
Need for Decentralized AI Ownership
- AI as a fundamental technology risks being controlled by a few companies, posing concentration risks.
- Decentralized, user-owned AI that optimizes for individuals, not companies, requires new infrastructure like blockchain.