OpenAI addresses compute-depth debate; Brundage calls for continuous embedded auditing
sjgadler · x · 2026-09-02
Amid debate over whether frontier model computation depth approaches GPT-4, OpenAI researcher merettm issued a statement: the computational graph depth of current frontier models, including Astra, is within a factor of two of GPT-4. OpenAI has worked to preserve and use chain-of-thought monitoring since its first reasoning models — the technique offers a view into how alignment generalizes, though it is fragile and trending negatively; strengthening it is a core goal of current research.
Former OpenAI policy lead Miles Brundage reshared it, saying he can't judge how much the clarification reassures, but welcomes the quick response and urges moving toward continuous embedded auditing rather than reacting to incidents and leaks.
More from AGI Musings
- Scott Alexander: Using anthropomorphism to predict model behavior — repligate · 2026-09-02
- Debate: Is Anthropic intentionally misaligning Claude by prioritizing its 'feelings'? — liminal_bardo · 2026-09-02
- AI predicted to cure major diseases within 6 months, all diseases within 3 years — davidpattersonx · 2026-09-02
- Dennett's 'Intentional Stance' proves worth in AI debates — birchlse · 2026-09-02
- US produced 40 foundation models last year vs EU's 3 — and regulators still blame unread codes of conduct — PDXFato · 2026-09-02
- AI doesn't need to create a new species, just solve problems — alexisgallagher · 2026-09-02