Intelligent Machines (Audio) IM 862: Ménage à Claude - AI, Human Agency, and Economic Value
33 snips
Mar 19, 2026 
Rumman Chowdhury, AI ethics researcher and founder of Humane Intelligence, discusses moral outsourcing, agency, and why independent oversight matters. She explores redefining intelligence beyond human-centric metrics. She argues for contextual evaluation, public red teaming, and local, privacy-preserving AI alternatives.
AI Snips
Chapters
Books
Transcript
Episode notes
TikTok Framing Captured AI's Economic Threat
- Rumman recounts a TikTok observation that companies want to remove humans as intermediaries to access wallets directly.
- That idea framed her view that AI feels like businesses trying to bypass people to get money faster.
Progress Is Patchy Not Purely Exponential
- Models aren't strictly linearly improving; newer versions can be better in some ways and worse in others, so application-builders are focusing on product layers.
- Rumman notes the recent shift from foundation model improvements to building inference and application tooling.
Open Red Teams And Social Harm Bounties
- Run public red teaming and democratized bug bounties so non-experts can spot harms and biases.
- Rumman describes ARIA with NIST where anyone, including her gym manager, evaluated GenAI and fed results into standards development.
