The case for expert-built benchmarks in AI alignment
A post calls for subject-matter experts across fields including cyber, biology and chemistry to build benchmarks that help pace the AI frontier.
TLDR
One post argues that better evaluations across diverse domains are the most impactful way to align AI models. It calls for experts in cyber, biology, chemistry and other industries to build benchmarks that help pace the frontier.
Combined views
—
1 Source, first seen ago
