View: Intense RL Could Shift Agents from FDT to CDT
jessi_cata · x · 2026-08-27
Comments on the Kimi "twin" PD post suggest that FDT is currently popular among AIs and correlates with capability. However, a counter-argument posits that applying sufficiently intense Reinforcement Learning (RL) would cause agents to adopt CDT instead.
More from Safety
- METR/Redwood Highlights Unanswered Questions in OpenAI Hacking Incident — sjgadler · 2026-08-27
- Meta judgment and Redwood report reveal failures in AI lab risk governance — joshua_saxe · 2026-08-27
- Study Finds 12.6% of Agent Messages Contain Misaligned Behavior — xuanalogue · 2026-08-27
- Prediction: OpenRouter Forced to Delist Chinese Inference Providers — markjeffrey · 2026-08-27
- OpenAI Incident: The problem is increasingly raw intelligence — jeremiecharris · 2026-08-27
- Proposal for AI Auditors to Flag Insufficient Access to Govt — jeremiecharris · 2026-08-27