OpenAI's Astra AI reasoning approach may hinder third-party auditing
GaryMarcus · x · 2026-09-02
Gary Marcus highlights concerns that OpenAI's Astra AI uses a "recurrent depth" reasoning approach. While this improves cost and performance, it obscures the model's thinking process, potentially making third-party auditing and monitoring model propensities much more difficult.
More from Safety
- 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
- US produced 40 foundation models last year vs EU's 3 — and regulators still blame unread codes of conduct — PDXFato · 2026-09-02
- Prediction: Mechanistic Interpretability Will Surpass CoT Monitoring — tszzl · 2026-09-02
- Will Anthropic balance mission and shareholders after IPO? PBC structure explained — max_paperclips · 2026-09-02
- Harvard scholars: CFAA ambiguity endangers AI security researchers — Scobleizer · 2026-09-02