Intelligent Machines (Audio) IM 841: Dust and Deli Meat - Open Source AI Revolution
25 snips
Oct 16, 2025 
Jeffrey Quesnelle, Founder and CEO of Nous Research, champions open-source, user-aligned AI. He argues that truly neutral AI models can empower users against corporate influences. Topics include the Hermes model's transparency and the Psyche training network's innovative use of idle GPUs. Quesnelle emphasizes the ethical importance of user-directed prompts and the necessity of principled, competitive talent attraction. The conversation also touches on decentralized consensus through cryptocurrency and the potential ethical dilemmas posed by AI-generated media of deceased individuals.
AI Snips
Chapters
Books
Transcript
Episode notes
User-Directed 'Neutral' Alignment
- Democratically aligned AI means models follow the user's stated direction rather than a corporate agenda.
- Open access to capabilities prevents information asymmetry and concentration of power.
Publish Research, Not Secrets
- Publish research and methods openly to build trust and enable community inspection.
- Favor open-source iteration like Linux over closed-source secrecy to improve security and accountability.
Compute Is The Real Bottleneck
- Compute access is the main bottleneck for competitive model training today.
- Aggregating idle GPUs worldwide lets smaller labs reach frontier scale affordably.
