Decentralized AI from the Academic POV, w/ MIT Media Lab’s Ramesh Raskar Artwork

The People's AI: The Decentralized AI Podcast

Who will own the future of AI? The giants of Big Tech? Maybe. But what if the people could own AI, not the Big Tech oligarchs? This is the promise of Decentralized AI. And this is the podcast for in-depth conversations on topics like decentralized data markets, on-chain AI agents, decentralized AI compute (DePIN), AI DAOs, and crypto + AI. From host Jeff Wilser, veteran tech journalist (from WIRED to TIME to CoinDesk), host of the "AI-Curious" podcast, and lead producer of Consensus' "AI Summit." Presented by Vana.

All Episodes

The People's AI: The Decentralized AI Podcast

Decentralized AI from the Academic POV, w/ MIT Media Lab’s Ramesh Raskar

April 02, 2025 • Jeff Wilser • Season 1 • Episode 7

In this episode of The People’s AI, presented by Vana, we dive deep into the evolution of decentralized AI with Ramesh Raskar, Associate Professor at the MIT Media Lab and advisor to Vana.

We explore why decentralized AI isn’t a rival to centralized AI—but a natural next step. Ramesh outlines a five-phase framework for the evolution of intelligent systems, moving from the “mainframe era” to the emerging “web of AI agents.” Along the way, we discuss how decentralized systems can address problems of data ownership, incentive design, verifiability, and user experience.

Whether you’re interested in AI agents, blockchain, data sovereignty, or the ethics of AI infrastructure, this episode provides a high-level vision with real-world applications... from a credible AI researcher who's far from a kool-aid drinking crypto-bro.

Timestamps:

4:46 – Mainframe vs PC: A mental model for AI evolution

6:01 – From intranet to internet to the “web of AI agents”

9:13 – Limits of centralized AI

12:08 – Market forces, decentralization, and the economy of AI agents

13:21 – The four pillars of decentralized AI

16:41 – Proof of inference, verifiability, and trust in AI

18:14 – Rethinking UX in an agentic future

21:21 – Do we need standards? Or will AI agents just adapt?

25:27 – A case study of how Vana addresses the four pillars

28:21 – A look at the future: agent schools, decentralized labor, and AI socialization

32:58 – How Web3 could accelerate progress in entrenched sectors