Threat Models

Applied to AI interpretability could be harmful? by Roman Leventov 18d ago

Applied to AGI-Automated Interpretability is Suicide by __RicG__ 18d ago

Applied to Gradient hacking via actual hacking by Max H 19d ago

Applied to A Case for the Least Forgiving Take On Alignment by Thane Ruthenis 1mo ago

Applied to Paths to failure by Karl von Wendt 1mo ago

Applied to On AutoGPT by Charbel-Raphael Segerie 1mo ago

Applied to The basic reasons I expect AGI ruin by Charbel-Raphael Segerie 1mo ago

Applied to Power-seeking can be probable and predictive for trained agents by Victoria Krakovna 1mo ago

Applied to AI Takeover Scenario with Scaled LLMs by simeon_c 1mo ago

Applied to AGI goal space is big, but narrowing might not be as hard as it seems. by Jacy Reese Anthis 2mo ago

Applied to AI x-risk, approximately ordered by embarrassment by Alex Lawsen 2mo ago

Applied to One Does Not Simply Replace the Humans by JerkyTreats 2mo ago

Applied to The Peril of the Great Leaks (written with ChatGPT) by bvbvbvbvbvbvbvbvbvbvbv 2mo ago

Applied to Deep Deceptiveness by Multicore 2mo ago

Applied to What‘s in your list of unsolved problems in AI alignment? by Jacques Thibodeau 3mo ago

Applied to [Linkpost] Some high-level thoughts on the DeepMind alignment team's strategy by Victoria Krakovna 3mo ago

Applied to Contra "Strong Coherence" by Cinera Verinia 3mo ago