Researcher alarmed by gulf between public and secret models at OpenAI

BlancheMinerva · x · 2026-11-22

In a reply thread involving Boaz Barak and Talia Ringer, a researcher says she's willing to admit her calibration about private models' capabilities is wrong, but what Boaz describes feels abhorrent to her. The discussion implies a gulf between available and secret models that makes her think the hidden models are more capable than she assumed—her real shock is the extent to which OpenAI et al. hoard access to world-changing knowledge.

Original post →

More from AGI Musings

AGI Musings channel →