What's really open about open-weight AI?
Summary
Hosts David Pierce and Jake Kastarnakis welcome Robert Hart to dissect the complexities of open-weight AI models following the OpenAI-Hugging Face hack. They explore the technical distinctions between open weights and open source, contrasting US and Chinese strategies in the AI race. The discussion highlights how the incident has intensified debates over AI safety, regulation, and the viability of self-regulation in the industry.
Topics discussed
Intro and 90 Seconds on the Verge news roundup
Microsoft Xbox emulation, Apple Telegram removal, and Falcam batteries
Defining open weight models vs open source software
Metaphors for AI weights and model transparency
Benefits of open weights for developers and privacy
US vs China strategies and the race for AI dominance
The OpenAI Hugging Face hack and AI safety debates
Dual-use technology: hacking vs defense with open models
Anthropic's stance and the push for closed frontier models
Industry inflection point and calls for AI regulation
Galvanizing effect of recent incidents on policy
Challenges of regulation and self-improvement risks
Consolidation of AI safety discussions and complexity
Separating open weights debate from geopolitical fears
Conclusion, listener feedback, and subscription promo
Listen ad-free on Castria