The Decentralized Future of Private AI [Illia Polosukhin] – 749

In this episode, Illia Polosukhin, a co-author of the seminal "Attention Is All You Need" paper and co-founder of Near AI, joins us to discuss his vision for building private, decentralized, and user-owned AI. Illia shares his unique journey from developing the Transformer architecture at Google to building the NEAR Protocol blockchain to solve global payment challenges, and now applying those decentralized principles back to AI. We explore how Near AI is creating a decentralized cloud that leverages confidential computing, secure enclaves, and the blockchain to protect both user data and proprietary model weights. Illia also shares his three-part approach to fostering trust: open model training to eliminate hidden biases and "sleeper agents," verifiability of inference to ensure the model runs as intended, and formal verification at the invocation layer to enforce composable guarantees on AI agent actions. Finally, Illia shares his perspective on the future of open research, the role of tokenized incentive models, and the need for formal verification in building compliance and user trust.

🗒️ For the full list of resources for this episode, visit the show notes page: https://twimlai.com/go/749.

🔔 Subscribe to our channel for more great content just like this: https://youtube.com/twimlai?sub_confirmation=1

🗣️ CONNECT WITH US!
===============================
Subscribe to the TWIML AI Podcast: https://twimlai.com/podcast/twimlai/
Follow us on Twitter: https://twitter.com/twimlai
Follow us on LinkedIn: https://www.linkedin.com/company/twimlai/
Join our Slack Community: https://twimlai.com/community/
Subscribe to our newsletter: https://twimlai.com/newsletter/
Want to get in touch? Send us a message: https://twimlai.com/contact/

📖 CHAPTERS
===============================
00:00 – Introduction
5:30 – Near Protocol
8:22 – User-owned AI
9:25 – Confidential computing
11:22 – Importance of confidentiality
15:56 – Challenges for model developers
21:01 – User-data liability to model providers
23:04 – Transition from user feedback to verifiable results
25:05 – Drivers of the shift in cost versus reward
28:15 – Model token rewards for data contribution
31:53 – Challenges of open-source models
35:54 – Barrier to privacy adoption
40:47 – Secure enclave and multi-party computation
43:51 – User interface and product development
45:10 – Computation costs
**46:00 – Barriers
48:28 – Intertia
50:33 – Latency
52:10 – Improving trust and privacy in models
54:19 – Open source vs. open weights
57:17 – Openness, verifiability of inference, formal verification
1:00:16 – Code verification and validation

🔗 LINKS & RESOURCES
===============================
Near AI – https://near.ai/
Decentralized Confidential Machine Learning – https://raw.githubusercontent.com/nearai/por/refs/heads/main/DecentralizedConfidentialMachineLearning.pdf
Autoformalization and Verifiable Superintelligence with Christian Szegedy – 745 – https://twimlai.com/podcast/twimlai/autoformalization-and-verifiable-superintelligence/

📸 Camera: https://amzn.to/3TQ3zsg
🎙️Microphone: https://amzn.to/3t5zXeV
🚦Lights: https://amzn.to/3TQlX49
🎛️ Audio Interface: https://amzn.to/3TVFAIq
🎚️ Stream Deck: https://amzn.to/3zzm7F5