Debating RSI, the US-China Gap, and Jaggedness with JS Denain of Epoch AI
AI Summary
An in-depth technical discussion featuring JS Denain from Epoch AI covering recursive self-improvement (RSI), the timeline gap between US and Chinese frontier AI labs, and the organizational metrics used to evaluate model progress. The conversation examines internal telemetry such as Codex usage spikes among researchers, hardware build-out velocity, and the inherent uncertainties in forecasting AI automation milestones.
Key Takeaways
- Analysis of internal lab data shows rising Codex spending among researchers, though high-level strategic decision-making tasks still require substantial human oversight.
- Chinese frontier models lag behind US counterparts by roughly six to eight months according to current tracking methodologies.
- Architectural innovations diffuse rapidly across global labs due to personnel turnover and shared research ecosystems.
Why it matters
Understanding how frontier labs evaluate internal AI acceleration and resource allocation helps engineers and researchers contextualize real-world efficiency gains versus public marketing hype, providing a clearer lens on true capability growth.
Conversation 0
Be the first to share your perspective.