Sean Taylor claims fast progress eradicating hallucinations; Andrew Ng: capabilities and safety can align

irinarish · x · 2026-09-04

Sean Taylor weighs in on the Astra discussion: the team is making fast progress on eradicating hallucinations, using evals that realistically capture user experience rather than academic benchmarks. Andrew Ng reshares to argue the perceived capabilities-vs-safety tradeoff is not inevitable — reducing hallucinations is a case where the two align.

Original post →

More from Models

Models channel →