Review

AI Forum Review

A regular publication on AI safety, alignment and adoption. Public abstracts are open; full issues are for subscribers.

Latest issue

Interpretability of large models

This issue surveys interpretability methods, reviews recent work and includes materials from the May webinar.

Recent issues

This issue surveys interpretability methods, reviews recent work and includes materials from the May webinar.

A survey of value-alignment methods: from learning with feedback to constitutional approaches.

How to reduce the energy and compute cost of modern models without losing quality.

Subscription

A free subscription unlocks full texts, event materials and restricted reports.